DIRECTED SELF-ASSEMBLY FOR

NANOFABRICATION AND DEVICE INTEGRATION

A DISSERTATION
SUBMITTED TO THE DEPARTMENT OF ELECTRICAL ENGINEERING
AND THE COMMITTEE ON GRADUATE STUDIES
OF STANFORD UNIVERSITY
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

He Yi
May 2015

 2015 by He Yi. All Rights Reserved.

Re-distributed by Stanford University under license with the author.

This work is licensed under a Creative Commons AttributionNoncommercial 3.0 United States License.
http://creativecommons.org/licenses/by-nc/3.0/us/

This dissertation is online at: http://purl.stanford.edu/kf745fv6296

ii

I certify that I have read this dissertation and that, in my opinion, it is fully adequate
in scope and quality as a dissertation for the degree of Doctor of Philosophy.
H.S.Philip Wong, Primary Adviser

I certify that I have read this dissertation and that, in my opinion, it is fully adequate
in scope and quality as a dissertation for the degree of Doctor of Philosophy.
Jonathan Fan

I certify that I have read this dissertation and that, in my opinion, it is fully adequate
in scope and quality as a dissertation for the degree of Doctor of Philosophy.
Eric Pop

Approved for the Stanford University Committee on Graduate Studies.

Patricia J. Gumport, Vice Provost for Graduate Education

This signature page was generated electronically upon submission of this dissertation in
electronic format. An original signed hard copy of the signature page is on file in
University Archives.

iii

Abstract
For more than 50 years, the size of the semiconductor devices has been scaling by
approximately a factor of two every 1.5-2 years. This has brought tremendous benefits
for the industry including lower cost per transistor, more computing power and higher
speed. However, it has been recently observed that the scaling of devices is approaching
fundamental (i.e. atomic scale) and economic (i.e. cost per fabrication facility) limits, in
large part because traditional lithography is facing substantial challenges for printing the
shrinking features while maintaining a reasonable cost. In response to this urgent need,
researchers are actively searching for alternative patterning approaches as the next
generation lithography. Potential solutions such as extreme ultraviolet lithography,
electron beam lithography, and multiple patterning lithography have attracted much
attention from the lithography community. However, each one of these solutions has its
own drawbacks, such as extremely high cost or low throughput. Among these solutions,
block copolymer directed self-assembly (DSA) stands out due to its low cost, high
throughput, well-controlled sub-20 nm features, and experimentally demonstrated
potential to scale below 14 nm.
Block copolymers are unique soft materials that can self-assemble through microphase
separation into various periodic nanostructures such as cylinders, spheres and lamellas,
driven by the incompatibility between the different blocks. The feature size of these
nanostructures is dependent on the molecular weight of the block copolymers and
therefore not limited by the same factors that limit optical lithography such as ultraviolet
light wavelength. In addition, the self-assembly could be controlled by a simple thermal
annealing process, which significantly reduces the cost and improves the throughput.
Among all the varieties of nanostructures, the cylindrical self-assembled patterns are
especially suitable for patterning contacts and vias in integrated circuits (ICs).
This dissertation focuses on the application of block copolymer DSA for contact hole
patterning in ICs. This work first demonstrates the flexible control of aperiodic DSA
patterns using small physical guiding templates, using both experiments and
iv

computational simulations. This is followed by the first patterning example of memory

and random logic circuit contacts using DSA. To enable practical technology adoption, I
introduce an alphabet approach that uses a minimal set of small physical templates to
pattern all contact configurations on integrated circuits. This work also illustrates,
through experiments, a general and scalable template design strategy that links the DSA
material properties to the technology node requirements. Last but not least, the
dissertation introduces a method to reduce DSA defectivity by using sub-DSA-resolution
Assist Features (SDRAFs).

Acknowledgement
The research work in this dissertation is achieved with many help from the people who
were part of my graduate life. I wish to acknowledge these people for making my journey
at Stanford University an enjoyable and memorable one.
First, I would like to thank my principle advisor Professor H.-S. Philip Wong, without
which this work would be impossible. Prof. Wong has offered his support, guidance, and
insightful perspectives throughout the years. His knowledge and open-mindedness not
only enlightened me on research, but also have a great inspiration on my viewpoint to life.
He is an exceptional researcher and wonderful mentor, and he is always willing to spend
time with students and figure out the research directions. He listens to students, respects
our opinions, and encourages us with creative ideas. I would not achieve so much without
his guidance.
I would also like to thank Prof. Jonathan Fan and Prof. Eric Pop, not only for their
willingness to be in my reading committee, but also for their kind suggestions regarding
my research and career path. I greatly appreciate Prof. Fabian Pease for serving as the
chair of my oral defense committee, as well as his guidance throughout the years. I also
thank Dr. Bart Laenens for being an examiner in my oral defense and the guidance during
my internship in ASML Brion.
I feel very grateful for Xinyu Bao and Li-Wen Chang, who worked on this DSA project
before I joined Stanford. Without their experience and knowledge, I would not make so
much progress in my research. I would also like to thank my group members in Stanford
and Stanford Nano Shared Facility (SNSF). Special thanks to Dr. Richard Tiberio, Dr.
Jim McVittie and James Conway for their technical assistance and advises to help my
experiments and smooth my research.
In addition, I would like to acknowledge all my collaborators from UIUC and UCSB,
including Prof. Martin Wong, Yuelin Du, Zigang Xiao, Hongbo Zhang, Tatsu Iwama,
vi

Prof. Glenn Fredrickson and Nabil Laachi. Special thanks to all my collaborators in
industry, including Chris Bencher, Huixiong Dai, Yongmei Chen, Jeffery Smith, Xumou
Xu, and Liyan Miao at Applied Materials; Moshe Preil, Azat Latypov, Vito Dai, Gerald
Schmid, Ji Xu, Edward Teoh, and Nicholas Eib at Globalfoundries; Bart Laenens, Yi Zou,
Christopher Spencer, Chenxi Lin, Alek Chen, Huayu Liu, and Sander Wuister at
ASML/Brion; Tamara Druzhinina, Davide Ambesi, Joost Bekaert, Roel Gronheid, Geert
Vandenberghe, Jan Doise, BT Chen, Germain Fenger, and Koen Martens at IMEC;
Lucian Shifen at ARM; Seiji Morita at Toshiba; Andres Torres at Mentor Graphics; Fred
Chen at ITRI; Yan Borodovsky at Intel and Andre Linden at Raith. The industrial
experience inspires me and helps greatly in my research. And I benefit so much from
both conducting collaboration research and doing summer internships.
This work is funded in part by SRC/GRC and NSF. I would like to acknowledge all of
the funding agencies for their financial support.
Most importantly, I would like to thank my parents in China: my mother Rongping Ma
and my father Yanlai Xiong for their unconditional love and support. They encourage me
to strive to become better and teach me to be responsible and consistent. They make me
proud.

vii

Table of Contents
Abstract .............................................................................................................................. iv
Acknowledgement ............................................................................................................. vi
Table of Contents ............................................................................................................. viii
List of Tables ...................................................................................................................... x
List of Figures .................................................................................................................... xi
Chapter 1 Introduction ........................................................................................................ 1
1.1 Context .................................................................................................................. 1
1.2 Preview ................................................................................................................. 2
Chapter 2 Block Copolymer Directed Self-Assembly and Next Generation Lithography . 4
2.1 Directed Self-Assembly of Block Copolymer ...................................................... 4
2.1.1 Block Copolymers ..................................................................................... 4
2.1.2 Directed Self-Assembly ............................................................................. 6
2.2 Next Generation Lithography ............................................................................... 8
Chapter 3 Flexible Control of DSA Aperiodic Pattern Using Small Physical Guiding
Template and an Alphabet Approach ............................................................................... 12
3.1 Introduction ......................................................................................................... 12
3.2 Materials and Process Flow ................................................................................ 13
3.2.1 PS-b-PMMA (Polystyrene-block-Poly(methyl methacrylate)) ............... 13
3.2.2 Self-Assembly Process Flow ................................................................... 14
3.3 Small Physical Guiding Templates ..................................................................... 15
3.4 DSA Design Space.............................................................................................. 19
3.5 DSA Alphabet Concept....................................................................................... 23
Chapter 4 Computational simulation of DSA in small physical guiding templates ......... 25
4.1 Introduction ......................................................................................................... 25
4.2 Self-Consistent Field Theory (SCFT) ......................................................... 26
4.3 Simulation Study......................................................................................... 28
4.3.1 Parameters ................................................................................................ 28
4.3.2 Guiding template affinity parametric studies ................................................... 29
4.4 1-hole DSA pattern design space ........................................................................ 31

viii

4.5 Summary ............................................................................................................. 34

Chapter 5 A General Strategy of Directed Self-Assembly (DSA) Guiding Template
Design for Contact Hole Patterning .................................................................................. 35
5.1 Introduction ......................................................................................................... 35
5.2 Technology Node and Contact Pitch .................................................................. 36
5.2.1 Unidirectional Layouts............................................................................. 36
5.2.2 DSA Design Space Revisited................................................................... 38
5.3 DSA Guiding Templates Design Strategy .......................................................... 39
5.4 DSA Contact Patterning from 14 nm to 7 nm node ............................................ 41
5.4.1 1-bit Half Adder ....................................................................................... 41
5.4.2 Strategy One: DSA Contact Patterning at 14 nm node ............................ 43
5.4.3 Strategy Two: DSA Contact Patterning at 11 nm node ........................... 46
5.4.4 Strategy Three: DSA Contact Patterning at 7nm node ............................ 48
Chapter 6 DSA Interaction Range and Sub DSA resolution Assist Features (SDRAF) .. 53
6.1 Template Density Influence on DSA patterns .................................................... 53
6.2 Materials and Process ......................................................................................... 55
6.3 Study of DSA Interaction Range using Gaussian Convolution .......................... 56
6.3.1 Experimental Observation ....................................................................... 56
6.3.2 Gaussian Convolution .............................................................................. 59
6.3.3 The Impact of Annealing Time ................................................................ 63
6.4 Sub DSA resolution Assist Features (SDRAF) .................................................. 65
6.4.1 Introduction .............................................................................................. 65
6.4.2 Test Guiding Pattern Design .................................................................... 66
6.4.3 Experiment Results .................................................................................. 67
6.4.4 Overfill Rate and SDRAF Sizes .............................................................. 68
6.4.5 Alternative SDRAF Design ..................................................................... 70
Chapter 7 Conclusions ...................................................................................................... 72
7.1 Summary of Contributions.................................................................................. 72
7.2 Future Work ........................................................................................................ 73
Publications ....................................................................................................................... 75
Reference .......................................................................................................................... 79

ix

List of Tables

Table 1. A comparison between different next generation lithographic solutions. .......... 10
Table 2. Pitch length between adjacent contacts versus technology nodes [28]. ............. 38

List of Figures
Figure 2.1 Common structures of block copolymers. ......................................................... 5
Figure 2.2 Theory predicted phase diagram for A-b-B diblock copolymers and
morphologies observed as a function of A block volume fraction (fA) of A-b-B
diblock copolymers [10]. ............................................................................................ 6
Figure 2.3 The process schematic for (a) topographic templates, (b) topographic posts, (c)
chemical templates, and (d) chemical posts. (e)-(h) The SEM images of the BCP
patterns using the corresponding method above [5]. .................................................. 7
Figure 2.4 The price evolution of lithography system since 1985. ..................................... 9
Figure 3.1 Chemical structure of PS-b-PMMA (polystyrene-block-poly
(methyl methacrylate)). ............................................................................................. 13
Figure 3.2 (a) Schematic illustration of PS-b-PMMA and microphase separation of
PS-b-PMMA into hexagonally packed PMMA cylinders embedded in PS matrix. (b)
SEM micrograph showing self-assembled PS-b-PMMA. ........................................ 14
Figure 3.3 Process flow of PS-b-PMMA self-assembly process. ..................................... 15
Figure 3.4 SEM images of DSA patterns (after PMMA cylinder removal) confined by
templates with different sizes and the corresponding hole size distribution. a, single
hole in 75nm squares. b, single hole in 92nm squares. c, 4-hole patterns in 126 nm
square templates. d, hole size distribution of a. e, hole size distribution of b. f, hole
size distribution of c. g, 2-hole patterns in 60 nm 110 nm rectangle templates. h,
3-hole pattern in 70 nm 145 nm rectangle templates. i, hole size distribution of g. j,
hole size distribution of h. Scale bars: 200 nm. ........................................................ 17
Figure 3.5 The pitch and overlay accuracy analysis for 2-hole and 3-hole patterns. a, hole
pitch distribution for the 2-hole pattern. b, hole pitch distribution for the 3-hole
pattern. c, overlay accuracy analysis for upper hole A in the 2-hole pattern. d,
overlay accuracy analysis for lower hole B in the 2-hole pattern. e, overlay accuracy
analysis for upper hole A in the 3-hole pattern. f, overlay accuracy analysis for
middle hole B in the 3-hole pattern. g, overlay accuracy analysis for lower hole C in
the 3-hole pattern. ..................................................................................................... 18
xi

Figure 3.6 2-hole and 3-hole design space for the same BCP film thickness. Each data
point represents one template generating a 2-hole or 3-hole pattern with different
template length and area ratios. Area ratio is the top surface area over the template
area; a larger area ratio means lower template density. As the template length
increases, the distance between holes also increases until the 2-hole patterns turn
into 3-hole patterns. It is shown that if the area ratio is larger, either the hole pitch
will be larger for the same template length or a 3-hole pattern will replace the 2-hole
pattern. ...................................................................................................................... 20
Figure 3.7 (a) Examples of three standard cells [42]. (b) An example of the alphabet set
based on the cells in (a). With the templates in this alphabet set, one can pattern all
the cells in (a) using DSA. ........................................................................................ 24
Figure 4.1 (a) SCFT Simulation domain includes sidewall, substrate of the guiding
template, block copolymer and air. (b) Top view and the cross section view of the
density profile (Phase A volume fraction). ............................................................... 29
Figure 4.2 Majority phase colormap plots of DSA (a) cross-section (b) top-view in an
80x80x40 nm guiding template for various values of the wall and substrate affinity
parameters, wallN and subN. wallN is varied from 0 to -25 and subN is varied from
-25 to 25. Negative N refers to the material that has higher affinity to the minority
polymer phase than majority phase. The majority phase is shown in red while the
minority phase, substrate, sidewall and air are all blue. The green dot line separates
the region where minority phase forms a hanging cylinder and the region that it
forms a cylinder down to the bottom. ....................................................................... 30
Figure 4.3 (a) SEM images of DSA patterns confined by templates with different sizes
from 45 nm to 93 nm. Each number represents the CD of the guiding template in the
image below. (b) SEM images of DSA patterns with different polymer film
thickness. The CD of these guiding templates is 80 nm. The number/rpm represents
the spin speed of block copolymer solution in the image below. ............................. 32
Figure 4.4 Parametric study of DSA (a) cross-section (b) top-view in a 50nm deep square
guiding template for various values of the wall size and polymer film thickness,
which is defined as the distance between the bottom of air and top of sidewall. In
this study wallN and subN are both -25. The majority phase is shown in red while the

xii

minority phase, substrate, sidewall and air are all blue. The green dot line separates
the region where minority phase forms a hanging cylinder and the region that forms
an isolated minority phase cavity in the center of guiding template......................... 33
Figure 5.1 (a) Example of a cell layout designed in unidirectional style [52]. All lines are
designed with the same width and pitch, and the contact holes are positioned only at
pre-determined grid points. The lines of poly-gate (purple) are perpendicular. Some
closely positioned contacts are marked with red circles. (b) Four smallest contact
pitches in (a), which correspond to the contact pairs in red circles except for the first
pair. ........................................................................................................................... 38
Figure 5.2 Demonstration of the three strategies described above. From left to right: (1)
the minimum contact pitch is larger than the resolution of conventional optical
lithography. (2) The minimum contact pitch is larger than the maximum DSA hole
pitch but smaller than the resolution of conventional optical lithography. (3) The
minimum contact pitch falls within the range of DSA hole pitch. ........................... 40
Figure 5.3 (a) Conventional half adder HA-X1 layout from Nangate 45nm Open Cell
Library. (b) Re-designed DSA-aware half adder HA-X1 layout. (c) The two metal
lines switch their positions to remove design rule violation. (d) The contacts (orange
boxes) are moved to new locations so they can be patterned using lithography. ..... 42
Figure 5.4 (a) An axis showing that the maximum DSA hole pitch is smaller than the
smallest pitch length for the 14 nm node. The maximum and minimum pitch values
for the BCP are obtained from the design space in Figure 3. (b) Circuit layout of
one-bit half adder. (c) Template design with only 1-hole templates. (d) Template
design with larger templates for 2 or 3 holes. (e) Successful DSA patterning using
design in (c). (f) Failed DSA patterning using design in (d), as extra holes appear in
over-sized templates. Scale bar: 200nm.................................................................... 44
Figure 5.5 14 nm HA-X1 DSA contact hole patterning analysis. (a) Patterned contact
hole size distribution (red) and the fabricated guiding template size distribution
(blue). Contact holes of 14.3 nm size were generated in guiding templates of 51.3
nm size. (b) Overlay accuracy of DSA contact holes in x and y direction, average
deviation: 1.0 nm. ................................................................................................... 45
Figure 5.6 Different choices of the guiding template design for the contact pair shown on

xiii

the left. The first one uses an elongated elliptical template but generates more DSA
holes than desired. The second one uses two circular templates. However, due to
lithography resolution, the small gap in between could not be resolved. Therefore,
the second strategy actually results in a peanut shape. ............................................. 47
Figure 5.7 Design space of DSA pattern in peanut-shaped template. Each data point
represents a specific peanut-shaped template length and connection width
combination that leads to a 2-hole pair. Different point symbols represent different
template lengths. The standard deviation of the hole-pitch is reflected as the error
bar in y-axis............................................................................................................... 47
Figure 5.8 (a) An axis showing that the maximum BCP pitch is close to the minimum
contact pitch for the 11 nm node but smaller than the lithography resolution (~60
nm). Peanut-shaped templates can be designed for the closely positioned contact
pairs. (b) Circuit layout of one-bit half adder given in Figure 5.3b. (c) 11 nm HA-X1
DSA contact hole patterns with layout given in (b). (d) Size distribution of DSA
contact holes. (e&f) Global overlay accuracy of DSA contact holes in x and y
direction, including ~2nm position error of templates brought by e-beam lithography.
The average global overlay accuracy 2.5nm. While some templates merge together
during fabrication process, the self-assembled holes are still separated with
reasonable size and overlay accuracy, showing high tolerance for template defects.
Scale bar: 200 nm...................................................................................................... 48
Figure 5.9

(a) An axis showing that the BCP pitch range matches with the smallest set

of contact pitches for the 7 nm node. Multiple-hole templates can be designed for all
the contact pairs with pitches within the achievable BCP pitch range. (b) Contact
layout. All the contact pairs with 42 nm pitch are labeled with red circles. (c) 2-hole
and 3-hole design space from Fig. 3. From this design space, we can find the
template length necessary to achieve 2-hole and 3-hole patterns with 42 nm pitch are
120 nm and 160 nm, respectively. (d) Template design based on this strategy.
Templates that generate holes with 42 nm pitches are outlined red. ........................ 50
Figure 5.10 (a) DSA-aware HA-X1 (half adder) layout. (b) A similar DSA-aware HA-X1
layout that cannot be implemented by the first strategy due to the difficulty of
printing the three closely packed contacts (in the red circle) separately. With the

xiv

third strategy, the resolution requirement is relaxed and this layout can be patterned.
(c) Contact layout design based on the layout in (a). (d) Contact layout design based
on the layout in (b). (e) 7 nm HA-X1 DSA contact hole patterns with the layout in
(a). (f) 7 nm HA-X1 DSA contact hole patterns with the layout in (b). Scale bar:
200nm. ...................................................................................................................... 51
Figure 5.11 7 nm HA-X1 DSA contact hole patterning analysis. (a) Size distribution of
DSA contact hole patterned with layout in Figure 5.10e. (b) Size distribution of
DSA contact hole patterned with layout in Figure 5.10f. (c) Global overlay accuracy
of DSA contact holes patterned with layout in Figure 5.10e, including ~2nm position
error of templates brought by e-beam lithography. (d) Global overlay accuracy of
DSA contact holes in x and y direction patterned with layout in Figure 5.10f.
Contact holes ~18nm were generated and the average global overlay accuracy
3nm. .......................................................................................................................... 52
Figure 6.1 For the same size of guiding templates, different template density leads to
different DSA results. The template pitch is: (a) 200 nm; (b) 250 nm; (c) 300 nm; (d)
400 nm. When the template pitch increases, the number of DSA defects (the missing
holes) rise significantly. (e) A cross-section cartoon showing the template is not
overfilled with polymer when the template density is high. (f) A cross-section
cartoon showing the template is overfilled with polymer when the template density
is low. ........................................................................................................................ 54
Figure 6.2 Schematic representation of DSA flow as implemented at IMEC. ................. 56
Figure 6.3 The layout of guiding template design. The numbers on the top of the blocks
as well as the crosses and squares on left and right side are SEM alignment marks.57
Figure 6.4 (a)-(c): SEM image of DSA patterns on the same row (see Figure 6.2). The
CD of the templates in these three cases is 66 nm. Template pitch: (a) 150 nm. (b)
200 nm. (c) 250 nm. (d) Zoom out view of (a). (e) Zoom out view of (b). (f) Zoom
out view of (c). .......................................................................................................... 58
Figure 6.5 (a) Low magnitude SEM image of one design block. (b) Binary image
processed from (a). (c) A binary matrix processed from (b). Each white circle in (b)
is treated as a white pixel in (c). (d) The average of 25 binary images retrieved from
the CDU wafer. (e) The average of 25 binary matrixes processed from (d). ........... 59

xv

Figure 6.6 The gray-scale convolved density map using different sigma values (unit: um).
The input layout is the same as shown in Figure 6.3. ............................................... 60
Figure 6.7 (a) Overfill probability map. (b) 8% percentile band of not overfilled templates.
White pixel means that the template at this location has an 8% chance of not
overfilled. (c) 92% percentile band of not overfilled templates. White pixel means
that the template at this location has a 92% chance of not overfilled. (d) The
convolved result before scaling. (e) The convolved result overlapped with 8%
percentile band. The band is colored in red. (f) The convolved result overlapped
with 92% percentile band. The band is colored in blue. (g) The scaling function. The
convolved density value of the red circle is the average convolved density value of
the red band in (e). The convolved density value of the blue circle is the average
convolved density value of the blue band in (f). ....................................................... 61
Figure 6.8 (a) Convolved result before scaling. (b) Scaled convolved result. (c) Overfill
probability map. (d) Scaled convolved results of different sigma values. It is obvious
that sigma = 0.8 matches the best with (c). (e) Sigma-score curve. The definition of
the score is shown above the curve. At the best case, theres ~10% difference
between scaled density and overfill probability. This difference may come from the
statistical fluctuation of thermodynamics as well as the systematic template CD
variation. ................................................................................................................... 62
Figure 6.9 (a) The overfill probability map retrieved from SEM images of 9 cases. (b)
The convolved density map using sigma = 0.7. ........................................................ 63
Figure 6.10 Low-magnitude SEM images. (a) Template CD 70 nm, anneal 1 min. (b)
Template CD 70 nm, anneal 5 min. (a) Template CD 70 nm, anneal 30 min. (a)
Template CD 60 nm, anneal 1 min. (a) Template CD 60 nm, anneal 5 min. (a)
Template CD 60 nm, anneal 30 min. ........................................................................ 64
Figure 6.11 The comparison between experimental results and convolved density map.
Annealing time is 1 minute for the left two columns. Annealing time is 5 minutes for
the middle two columns. Annealing time is 30 minutes for the right two columns. (a)
Experimental results. (b) Convolved density map. ................................................... 65
Figure 6.12 (a) The layout design of test pattern. Small squares are SDRAFs and
rectangles are targeted DSA templates. (b) The cases that the SDRAFs were not

xvi

resolved due to overexposure. (c) The SEM picture of (a), the CD of SDRAFs is 39
nm. (d) The SEM picture of (b). For target templates, the size is 53 nm by 81 nm
(minor and major axis lengths) in both cases............................................................ 66
Figure 6.13 Comparison between DSA performance with SDRAFs and without SDRAFs.
The top row (a) (d) is the case without SDRAFs, the bottom row (e) (h) is the
case with SDRAFs. (a) SEM image of guiding template without SDRAF. The large
ovals are the target DSA templates, with a major axis of 82 nm and a minor axis of
53 nm. (b) The DSA result of (a). 1-hole DSA patterns are generated inside the
target templates. (c) The low magnification image of the block. This provides a
zoom-out view for us to inspect the location of overfilled templates. Note that (b) is
a zoom-in view of the blue box shown in the center. It is also important to note that
there are some templates at the corner of the block that appear missing but is
actually a result of polymers overfilling, rendering the template less visible under
the SEM (see zoom-in view in (d)). (d) A zoom-in view of the blue box shown in the
left top corner of (c). Red circles label the templates that are overfilled due to low
density. (e) SEM image of guiding template with SDRAFs. The small circular ones
are the SDRAFs with a size of 40 nm. The target ovals have a major axis of 81 nm
and a minor axis of 53 nm. (f) The DSA result of (d). 1-hole DSA patterns are
generated inside the target templates while nothing inside the SDRAFs. (g) The low
magnification image of the block. (f) is a zoom-in view of the blue box shown in the
center. (h) A zoom-in view of the blue box shown in the left top corner of (g). No
oval DSA template is overfilled and they all have 1-hole DSA patterns inside. ...... 67
Figure 6.14 An example of failed SDRAFs. (a) SEM image of guiding templates, the CD
of SDRAF is 55.2 nm. (b) SEM image of DSA result. Inside the red boxes are the
1-hole DSA patterns generated inside the SDRAFs. (c) The low magnification SEM
image of the block. .................................................................................................... 68
Figure 6.15 The relationship between SDRAF sizes and OR. The two stars on the left
marked the OR when there is no SDRAF. Blue data points represent the case where
the target ovals have a major axis of 75 nm and a minor axis of 57 nm. Orange data
points represent the case where the target ovals have a major axis of 80 nm and a
minor axis of 60 nm. Note that although the some of SDRAF CDs shown above

xvii

reached below 30 nm, the actually feature sizes on the wafer suffer from large
variation, which explains some of the noises in this trend. ...................................... 69
Figure 6.16 (a) SEM image of guiding templates without SDRAF. The template CD is 74
nm, and the template pitch is 380 nm. (b) SEM image of DSA results in (a). Some of
the templates were overfilled and DSA defects were formed. (c) Low magnitude
SEM image of the entire contact matrix of (b). The OR is higher than 30%. (d) SEM
image of guiding templates with SDRAFs. The template CD and pitch are same as
(a). The SDRAF size is 41 x 137 nm. (e) SEM image of DSA results in (d). No
templates were overfilled. (f) Low magnitude SEM image of the entire contact
matrix of (e). The OR is 0%...................................................................................... 70
Figure 6.17 (a) SEM image of guiding templates without SDRAF. The template CD is 72
nm, and the template pitch is 600 nm. (b) SEM image of DSA results in (a). Some of
the templates were overfilled and DSA defects were formed. (c) Low magnitude
SEM image of the entire contact matrix of (b). The OR is higher than 50%. (d) SEM
image of guiding templates with SDRAFs. The template CD and pitch are same as
(a). The SDRAF size is 38 x 86 nm. (e) SEM image of DSA results in (d). No
templates were overfilled. (f) Low magnitude SEM image of the entire contact
matrix of (e). The OR is 0.03%, means that only two templates are overfilled. ...... 71
Figure 7.1 (a) Three-dimensional self-assembled nanostructures [63]. (b) SEM of 3D
bilayer nanostructures generated by functionalized posts [64]. ................................ 74

xviii

Chapter 1
Introduction
1.1 Context
Semiconductor device scaling has continued over the last few decades to boost the
cost-performance of the integrated circuits [1]. The practical limit of device scaling
depends on the ability to manufacture billions of semiconductor devices reliably,
economically, and with reasonable process tolerances. As the feature size of the devices
has shrunk down significantly, traditional optical lithography now faces increasing
challenges. Optical lithography, which utilizes the ultraviolet light to print the lines and
holes of the integrated circuits, is approaching its resolution limit. One of the most
demanding challenges lithography facing today is that device scaling not only requires
achieving finer features, but also printing them with tight pitches. If the conventional
lithography were unable to print devices that are both smaller and denser, the scaling of
semiconductor devices will hit a hard stop. In order to extend the resolution limits of
lithography and keep up with the requirement of device scaling, researchers are actively
searching for alternative next generation lithography solutions.
In recent years, directed self-assembly (DSA) of block copolymer has become a subject
of strong interest due to its capability to generate different types of highly packed sub-20
nm nanostructures in a periodic fashion [2]. The size and pitch of the self-assembled
nanostructures rely on the block copolymers molecular parameters instead of ultraviolet
light, therefore providing us with a new way to scale devices without hitting the optical
resolution limit. Researchers and companies are actively seeking solutions to apply
self-assembled nanostructures for patterning integrated circuits, and two of the most
popular self-assembled patterns are lamella and cylindrical nanostructures. The
1

cylindrical nanostructures are especially promising as it could be used to pattern contact

holes of the circuits [3], which is one of the most difficult challenges in optical
lithography.
Before DSA became a promising option for next generation lithography, chemists have
already explored DSA extensively. There has been tremendous progress in using DSA to
achieve various periodic nanostructures over large area [4-6]. However it is important to
note that the contact holes in the integrated circuits are not positioned in a periodic
fashion. In addition, precise control is required of both the size and the position of the
contact holes. Therefore, it is extremely critical to enable DSA for generating and
controlling aperiodic patterns. To achieve such goals, researchers have explored the use
lithographically defined guiding templates to guide the formation of self-assembly
features. Physical guiding templates approaches have proven promising for aperiodic
cylindrical DSA patterns, thus extensively discussed in this dissertation.

1.2 Preview
This dissertation will focus on the application of block copolymer directed self-assembly
(DSA) for patterning contact holes in integrated circuits. We first demonstrated the
flexible control of aperiodic DSA patterns using small physical guiding templates and
introduced the concept of DSA alphabet. Then we showed examples of applying these
aperiodic DSA patterns for patterning the contact holes of memory and random logic
circuits. A general and scalable template design strategy is presented, which links the
technology node requirement to the DSA materials and process. Next, we took a step
further by studying the design rules of DSA and the solution to mitigate DSA defects
brought by the influence of guiding template density.
This thesis is structured as follows. Chapter 2 includes an introduction of block
copolymer directed self-assembly as well as some of other promising next generation
lithography solutions. Chapter 3 presents our directed self-assembly process flow and

explains why we need to use physical guiding templates to generate and control aperiodic
DSA patterns. It will also introduce the new concept of DSA alphabet and explain its
importance. Chapter 4 discusses the computational simulation of DSA process inside
small physical guiding template using Self Consistent Field Theory (SCFT). Chapter 5
develops a general guiding template design strategy and includes examples of DSA
contact hole patterning from 14 nm to 7 nm technology node. Chapter 6 first addresses
the DSA defects brought by the template density influence, and then it presents a
behavioral model based on Gaussian convolution to describe this density influence. This
chapter also provides a solution to reduce DSA defectivity by implementing SDRAF
(Sub-DSA-Resolution Assist Features).

Finally, conclusions are drawn in Chapter 7

along with suggestions for future work.

Chapter 2

Block Copolymer Directed

Self-Assembly and Next Generation
Lithography
2.1 Directed Self-Assembly of Block Copolymer
2.1.1 Block Copolymers
Block copolymers consist of two or more chemically distinct monomers covalently
attached to each other [7]. There are multiple kinds of block copolymers, depending on
the number of chemically distinct monomers and structures. Figure 2.1 shows different
types of block copolymers that are categorized as AB diblock, ABA triblock, ABC
triblock and starblock copolymers. In the following chapters, we will focus on the case of
diblock copolymers that are comprised of A and B blocks (denoted as A-b-B).

AB#diblock#
ABA#triblock#
ABC#triblock#

A#

B#

A#

B#

A#

A#

B#

C#

B#
B#

A#

starblock#

B#

Figure 2.1 Common structures of block copolymers. A, B and C represent different

polymer blocks.

Most distinct monomers in block copolymers are immiscible and therefore they are
spontaneously driven by the repulsion between blocks to segregate into a variety of
ordered morphologies [7]. This process is referred to as self-assembly or microphase
separation as a result of minimization of free energy that drives the two polymer blocks
apart from each other, while the covalent bond between them constraints the maximum
separation distance. The repulsive force between the blocks is characterized by the
product N, where is the Flory-Huggins interaction parameter [7], an approximation of
the energy of mixing of the constituent monomers and N is the degree of polymerization
which stands for the total number of monomers in the block copolymer [7]. The feature
size of microphase separation of block copolymers is typically a few tens of nanometers
[8]. The self-assembled block copolymer films will form periodic microdomains such as
cylinders, spheres, and lamellae. The type of microdomains mainly depends on diblock
polymer type and the volume ratio between the two blocks [9, 10] (Figure 2.2).

Figure 2.2 Theory predicted phase diagram for A-b-B diblock copolymers and
morphologies observed as a function of A block volume fraction (fA) of A-b-B diblock
copolymers [10].

2.1.2 Directed Self-Assembly

Naturally, the self-assembly process of block copolymer films without any guidance will
lead to periodic but highly defective nanostructures. Therefore, researchers have been
searching for both topographical and chemical guiding structures to reduce defects and
improve ordering during microdomain pattern formation [11-18]. The topographic and
chemical template processes are shown schematically in Figure 2.3.

Figure 2.3 The process schematics for (a) topographic templates, (b) topographic posts,
(c) chemical templates, and (d) chemical posts. (e)-(h) The SEM images of the BCP
patterns using the corresponding method above [5].

Topographic templates patterned by the conventional lithography (Figure 2.3a) provide

physical confinement that guides the BCP during self-assembly [4, 19]. The relatively
small area maintains order [20, 21], and the pre-defined template is subdivided into finer
features by the BCP pattern, effectively increasing the spatial resolution. Segalman et al.
[6] have demonstrated using 1 10 m trenches to achieve long range positional order
and orientation of single crystal films of sphere-forming PS-b-P2VP diblock copolymer.
Cheng et al. [13] have demonstrated the pattern density multiplication using PS-b-PFS
diblock copolymer inside sub-500 nm e-beam lithography defined trenches. Gronheid et
al. [22] have also illustrated sub-20 nm self-assembled cylinder using PS-b-PMMA
diblock copolymer inside topographical templates patterned by 193 immersion
lithography, and further transferred the self-assembled patterns to the nitride layer

underneath. In subsequent chapters, graphoepitaxy will be the primary choice for

controlling self-assembly features, especially for generating aperiodic DSA patterns.
Chemical guiding approaches, on the other hand, use chemical patterns of preferential
affinity on the substrate surface to control patterns formation [5] (Figure 2.3 c and d).
Using chemical guiding patterns, researchers have successfully demonstrated not only
just hexagonal cylindrical arrays but also orthogonal arrays using triblock copolymer [23].
Furthermore, as the chemical guiding approach can deliver low-defect self-assembly
results over large areas, it becomes especially ideal for the line-and-space structures in
circuit layout using lamella-forming block copolymer [24-26]. The LiNe DSA flow [27],
for example, has been explored thoroughly and moved readily towards mass volume
manufacturing [28, 29].

2.2 Next Generation Lithography

With the continued scaling of device dimensions, it is increasingly difficult for
conventional optical lithography to meet the requirements for improved resolution while
keeping cost under control. This is especially the case for patterning contact holes in very
large scale integrated (VLSI) circuits, as very precise control is required of both the size
and the position of the cylindrical structures. The current state-of-the-art optical
lithography can reach the 3-sigma values for critical dimension (CD) control and overlay
(registration of one layer to another layer of lithography) accuracy at 2.0 nm and 4.0 nm,
respectively. However, for the 10 nm technology node and beyond, the required
lithography 3-sigma values for critical dimension (CD) control and overlay accuracy are
below 1.0 nm and 3.2 nm [30]. Furthermore, the defectivity of contacts and vias in
semiconductor manufacturing needs to be much less than one defect per billion [31]. In
addition, optical lithography must be extended and next generation lithography developed
not only to maintain the rapid pace of Si complementary metal-oxide-semiconductor
(CMOS) scaling [32], but also to enable fabrication of post-CMOS nanoscale devices .
Therefore, researchers and semiconductor companies have invested tremendous amount
of resources and money to enhance the capability of conventional optical lithography,

and the price of lithography tool has increased exponentially since 1985 (Figure 2.4).

Figure 2.4 The price evolution of lithography system since 1985.*

Because of all these challenges and increasing costs, much attention has been given to
next generation lithographic solutions, including extreme ultraviolet lithography (EUV)
[33], electron beam lithography (e-beam) [34, 35], and multiple patterning lithography
(MPL) [36]. However, each of these solutions has its drawbacks (Table 1). EUV, for
example, could reach much higher resolution than the current 193-immersion lithography
but is overly expensive and suffer from other problems such as limited power source [33].
While one EUV tool costs more than 100 million dollars, the throughput of EUV
lithography is ~50 wafers/hour, only one third of the industry standard today. Compared
with EUV, multiple patterning has satisfying throughput but limited resolution. Though
its cost appears to be lower than EUV lithography, the process complexity required by

*

Source: http://google.brand.edgar-online.com/

multiple patterning increases substantially as device scaling continues. E-beam direct

write, on the other hand, suffers from extremely throughput, as all the features need to be
written by electron beam sequentially.

Cost%

Throughput%

Resolu/on%

EUV%lithography%
Mul/ple%Pa9erning%
E:beam%direct%write%
Directed%Self:Assembly%
(DSA)%
Table 1. A comparison between different next generation lithographic solutions.

Among all the alternative next generation solutions, directed self-assembly stands out due
to low cost, high throughput and high resolution. The ultimate feature size generated by
directed self-assembly is not limited by ultraviolet wavelength, which provides more
potential at smaller technology nodes.
However, with all the advantages, directed self-assembly (DSA) still faces its own
challenges and therefore not ready for semiconductor manufacturing. The challenges
come in three perspectives: 1. The defectivity of DSA is far from industry standards
today. While Intel reported the DSA defectivity of 1/cm2 in 2015 [37], the defectivity
needs to be lower than 0.01/cm2 to ensure the total yield of semiconductor chips . 2. The
overlay requirement for DSA becomes stricter as technology scales down. 3. Not only the
DSA materials and process could be improved, IC layouts could also be further optimized
to be more DSA-friendly. A design methodology connecting DSA process and layout
design is necessary to enable Electronic Design Automation (EDA).
In summary, directed self-assembly is a promising next generation lithography
technology that provides a cost-effective way to extend the resolution of current optical

10

lithography. Before DSA becomes ready to be adopted by semiconductor industry, there

still exist various problems be solved, from a mass production point of view. In the
following chapters, we will focus on addressing some of the challenges of DSA,
especially for the contact hole patterning in integrated circuits.

11

Chapter 3
Flexible Control of DSA Aperiodic
Pattern Using Small Physical Guiding
Template and an Alphabet Approach1
3.1 Introduction
This chapter covers the materials and process flow used in our experiments. The
aperiodic DSA patterns were achieved by using small physical guiding templates and
characterized by the geometric dimension of the guiding templates. Using the aperiodic
1-hole DSA patterns, we for the first time demonstrated the contact hole patterning using
DSA for 22 nm SRAM cell. Then this chapter will introduce the new concept of DSA
alphabet and how this concept could be applied to pattern contact holes using DSA.

2012 John Wiley and Sons. Part of this chapter has been reprinted, with permission, from H. Yi, X.-Y.
Bao, J. Zhang, C. Bencher, L.-W. Chang, X. Chen, R. Tiberio, J. Conway, H. Dai, Y. Chen, S. Mitra and
H.-S. P. Wong, Flexible Control of Block Copolymer Directed Self-Assembly using Small Topographical
Templates: Potential Lithography Solution for Integrated Circuits Contact Hole Patterning, Adv. Mater.,
vol. 24, issue 23, pp. 3107 3114, 2012.
2011 IEEE. Part of this chapter has been reprinted, with permission, from X.-Y. Bao, H. Yi, C. Bencher,
L.-W. Chang, H. Dai, Y. Chen, et al., "SRAM, NAND, DRAM contact hole patterning using block
copolymer directed self-assembly guided by small topographical templates," in Electron Devices Meeting
(IEDM), 2011 IEEE International, pp. 7.7. 1-7.7. 4, 2011.
The work in this chapter has been done in close collaboration with Xinyu Bao, a postdoc at Stanford
University advised by Prof. H.-S. Philip Wong.

12

3.2 Materials and Process Flow

3.2.1 PS-b-PMMA
(Polystyrene-block-Poly(methyl methacrylate))
The block copolymer used in this study is PS-b-PMMA (polystyrene-block-poly(methyl
methacrylate)) shown in Figure 3.1 and is purchased from a commercial vendor. The
PS-b-PMMA has a PS/PMMA volume ratio of ~70:30 (Mn=46.1 kg/mol for PS and
Mn=21.0 kg/mol for PMMA and polydispersity index of 1.09 (the weight-average
molecular weight divided by the number-average molecular weight). This type of
PS-b-PMMA block copolymer is known to naturally generate hexagonally ordered
cylindrical microdomains with diameter of ~20 nm and pitch of ~40 nm (shown in Figure
3.2).
CH3

CH2

CH

CH2

C
C

m
O

O
CH3

Figure

3.1

Chemical

structure

of

(methyl methacrylate)).

13

PS-b-PMMA

(polystyrene-block-poly

PMMA

PS

20 nm

mic rophas e/s eparation

PS
100nm

P MMA

40 nm
(a)

(b)

Figure 3.2 (a) Schematic illustration of PS-b-PMMA and microphase separation of

PS-b-PMMA into hexagonally packed PMMA cylinders embedded in PS matrix. (b)
SEM micrograph showing self-assembled PS-b-PMMA.

3.2.2 Self-Assembly Process Flow

In order to distribute the block copolymer evenly, a polymer solution was made for spin
coating process. PS-b-PMMA powder was dissolved in propylene glycol methyl ether
acetate (PGMEA) (purity: 99.5%) with a concentration of 1 wt.%. PGMEA is a solvent
compatible to industrial manufacturing environment. The PS-b-PMMA solution was
spin-coated onto the substrates with a speed around 1000-3000 rpm. Then the samples
were transferred to an oven and annealed at 180 C in N2 atmosphere for 24 hours. The
ramp rate for heating and cooling is 6 C / min and 2 C / min. After cooling down, the
samples were exposed to deep UV radiation (dose 3.5 J/cm2, wavelength 254 nm) to
degrade the PMMA block and immersed in acetic acid (concentration: 99.7%) for 10
minutes to remove the PMMA segments. Finally the samples were rinsed in deionized
water and dried by N2. The samples were analyzed using Scanning Electron Microscope
(SEM) under 1.2 10-4 bar vacuum and 5 kV acceleration voltage.

14

PS-b-PMMA
Dissolved in PGMEA
Spin coating
Si
PS is left as a resist mask
for pattern transfer

Thermal
Annealing

Deep UV radiation
Soaked in
Acetic Acid
PMMA

PMMA cylinder removal

PS

Figure 3.3 Process flow of PS-b-PMMA self-assembly process.

3.3 Small Physical Guiding Templates

To generate aperiodic block copolymer self-assembly patterns with precise size and
position control, researchers have explored the use of directed self-assembly (DSA) using
lithographically defined guiding templates. Chemical guiding template approaches use
chemical patterns of preferential affinity on the substrate surface to form aperiodic
patterns [19, 24, 25]. However, most studies on chemical guiding templates so far have
been focused on generating lamella self-assembly nanostructures for lines and spaces
patterns [38]. In order to generate aperiodic cylindrical patterns for holes, physical
guiding templates approaches have proven promising [3, 6]. Due to the strong
confinement from physical guiding templates, the block copolymers natural symmetry is
altered and aperiodic cylindrical patterns therefore readily form inside the templates.

15

Because the DSA pattern position is determined by position of the guiding templates,
DSA patterns can be almost arbitrarily placed as required. By altering template shapes
and block copolymer film thickness, various types of cylindrical structures are studied
and placed as required in integrated circuits exactly.
To study the use of small guiding templates we conducted a series of experiments. The
small square templates were fabricated using conventional optical lithography and
etching process on 300 mm Si wafer with sizes ranging from 75 nm to 140 nm and depth
around 50 nm. After performing the PS-b-PMMA DSA process, the DSA patterns were
characterized by high-resolution scanning electron microscope (SEM).

It is noteworthy

that the PS-b-PMMA DSA in small templates does not require the homopolymer that was
used to stabilize the self-assembly as predicted by the SCFT simulation [39], nor the
random copolymer coating that was used to neutralize the substrate surface as the
prevailing method does [40]. The results demonstrated that the aperiodic DSA patterns
can be generated using small physical guiding templates and almost arbitrarily placed as
required by circuit fabrication on one wafer.
Figure 3.4 summarizes some of the major IC-layout relevant aperiodic DSA patterns
obtained in small physical guiding templates. Single-hole patterns (Figure 3.4a&b) can be
formed in square templates from 60 to 100 nm. The four-hole patterns with square
symmetry can be formed in the relatively large template size window between 110~130
nm (Figure 3.4c). In rectangular templates (width from 60 to 80 nm, length from 100 to
150 nm), 2-hole and 3-hole patterns can be formed. Besides hole size distribution, we
also study the overlay accuracy of DSA patterns in Figure 3.5. We calculate the average
position of DSA patterns in guiding templates of the same size, and obtain the overlay
accuracy of the self-assembled features in terms of average positional deviation of
self-assembled holes from the average position. It is shown in figure 3.5 that the average
positional deviations for both 2-hole and 3-hole patterns are less than 2 nm, indicating a
high repeatability and accurate placement. For template size larger than 140 nm, the
pattern formation becomes complicated and uncontrollable due to the reduced
confinement effect from the guiding templates and the internal force of the block

16

copolymers becoming dominating. As shown in Figure 3.4, the average DSA pattern size
deviated from the naturally self-assembled PS-b-PMMA (~ 20 nm). The size deviation
and symmetry change suggested both the PS and PMMA polymer chains were stretched
or compressed by the strong lateral confinement during the self-assembly process.

40%

Mean:(15.2(nm
SD:(2.3(nm

e"

Mean0=028.60nm
30% SD0=05.20nm0

(75(nm

Count

20%

20%

0%
5

10

15

20

25

Hole(Size((nm)

g"

10%

0%
15

30

(126(nm

20%

10%

10%

Mean:(30.3(nm
40% SD:(2.1(nm
30%

Count

30%

f"

8
0920nm

Count

d"

c"

b"

a"

20

25

30

35

40

Hole0Size0(nm)

h"

A
B

45

50

0%
20

25

30

35

Hole(Size((nm)

40

A
B
C

j"

30%
20%

70nmx145nm

20%
10%

10%
0%
8

40% Mean:)14.8)nm
SD:)2.4)nm
30%

60nmx110*nm

Count

50% Mean:*15.0*nm
40% SD:**1.8*nm

Count

i"

10

12

14

16

18

Hole*Size*(nm)

20

22

0%
8 10 12 14 16 18 20 22 24

Hole)Size)(nm)

Figure 3.4 SEM images of DSA patterns (after PMMA cylinder removal) confined by
templates with different sizes and the corresponding hole size distribution. a, single hole
in 75nm squares. b, single hole in 82 nm squares. c, 4-hole patterns in 126 nm square
templates. d, hole size distribution of a. e, hole size distribution of b. f, hole size
distribution of c. g, 2-hole patterns in 60 nm 110 nm rectangle templates. h, 3-hole
pattern in 70 nm 145 nm rectangle templates. i, hole size distribution of g. j, hole size
distribution of h. Scale bars: 200 nm.

17

40%

Mean: 39.7nm
Std: 1.9nm

cd

40%

2-hole pattern

20%

10%

40%

AveDev-x=1.1 nm
40%
AveDev-y=1.8
nm
3-hole: A

36

38

39

Hole

-2

-1

0%

Count
-3

37

38

39

40

0%

-3-2 -2

41

42

AveDev-x=1.3 nm
AveDev-y=1.9 nm
3-hole: C

AveDev-y =1.5 nm
2-hole: B

20%

10%

-1-1 00 1 -31 2 -2 23 -1
Centroid Deviation
Deviation (nm)
Centroid
(nm)

40%

AveDev-x=1.1 nm
AveDev-y=1.8 nm
3-hole: A

30%

36

30%

10%

35

20%

10%

-3

0%
34

45

0%
30-3 1-2 2-1 3 0 -3 1 -3-22 -2 3-1-1 00 1 1 2 23
CentroidDeviation
Deviation (nm)
Centroid
(nm)

AveDev-x=1.5 nm
AveDev-y=1.8 nm
3-hole: B

3-3

-2

-1

AveDev-x=1.3 nm
AveDev-y=1.9 nm
3-hole: C

Count

0%

44

AveDev-y=1.9
nmAveDev-y=1.8 nm
AveDev-y=1.8 nm AveDev-y=1.8
nm
40%
AveDev-x =1.1
nmB
3-hole:
C
AveDev-x
=0.9
nmB
3-hole:
3-hole:
3-hole: A

20%

10%

20%

Hole
Pitch (nm)
nmAveDev-x=1.5
nm
AveDev-x=1.5 nm AveDev-x=1.1AveDev-x=1.3
nm

Count

20%

40 41 42 43
40% (nm)
Pitch

AveDev-y30%
=1.3 nm
2-hole: A

30%

Count

37

Count

30%

3-hole pattern

10%

0%
35

30%

Count

Count

30%

Mean: 37.7nm
Std: 1.0nm

20%

10%

0%

-3

-2

-1

-3

-2

-1

-3

-2

-1

Centroid Deviation (nm)

Figure 3.5 The pitch and overlay accuracy analysis for 2-hole and 3-hole patterns. a, hole
pitch distribution for the 2-hole pattern. b, hole pitch distribution for the 3-hole pattern. c,
overlay accuracy analysis for upper hole A in the 2-hole pattern. d, overlay accuracy
analysis for lower hole B in the 2-hole pattern. e, overlay accuracy analysis for upper
hole A in the 3-hole pattern. f, overlay accuracy analysis for middle hole B in the 3-hole
pattern. g, overlay accuracy analysis for lower hole C in the 3-hole pattern.

18

3.4 DSA Design Space

In section 3.3 we realized that through adjusting the geometric dimensions of guiding
templates, one could flexibly generate and control aperiodic DSA patterns. The feature
size and the pitch can also be compressed or stretched to some extent. However, the
detailed relationship between the geometries of DSA patterns and the template size and
density is not fully explored yet. Therefore, to describe and characterize this relationship,
we introduce the concept of DSA design space. The DSA design space defines the pattern
geometries as a function of a set of conditions such as spin speed, template size and
density, which can be determined to generate a specific pattern. This design space is
determined by the chemical properties of the block copolymer, as well as by the polymer
thickness, template size and template density that result in a specific DSA pattern. It is
important to note that the chemical affinity of the substrate and the depth of the template
also affect the DSA pattern. However, the templates in one contact layer must share the
same depth and substrate material, so for simplicity, these two parameters are not listed in
the design space. All of the templates used in these experiments were fabricated on Si and
etched to a depth of 50 nm.
Figure 3.6 shows an example of the design space of DSA 2-hole and 3-hole patterns.
These two DSA patterns are generated inside elliptical guiding templates, which have
been identified as among the most common DSA aperiodic patterns. From this design
space we observe that there is a clear transition from 2-hole to 3-hole pattern as the
template length increases. We measure the change in pitch for the 2- and 3-hole DSA
patterns resulting from increasing the length of the elliptical templates while keeping the
width unchanged (see Figure 3.6). Proper template designs can avoid conditions at the
boundaries of the design space to ensure a wide process window. The importance of
developing the DSA design space is that it can directly relate the DSA hole pitch to the
polymer film thickness and template size and shape, without requiring knowledge of the
chemical properties or process. This allows the design space, once fully characterized, to
provide reliable guidance for further template design regardless of the type of block

19

copolymer or DSA process used.

!

52
48

A re a !ra tio:!1.5
A re a !ra tio:!2.5
A re a !ra tio:!3.5
A re a !ra tio:!4.5

44
!

D S A -H o le -P itc h -(n m )

56

40
36

2"hole'

32

3"hole'
90

120
150
T em p late-L en g th -(n m )

180

Figure 3.6 2-hole and 3-hole design space for the same BCP film thickness. Each data
point represents one template generating a 2-hole or 3-hole pattern with different
template length and area ratios. Area ratio is the top surface area over the template area; a
larger area ratio means lower template density. As the template length increases, the
distance between holes also increases until the 2-hole patterns turn into 3-hole patterns. It
is shown that if the area ratio is larger, either the hole pitch will be larger for the same
template length or a 3-hole pattern will replace the 2-hole pattern.

3.5 Contact hole patterning of 22 nm SRAM cell using

DSA
The flexible control of DSA pattern afforded by small physical guiding templates enables
patterning contact holes of practical devices for memory and logic circuits. In this section
we will illustrate DSA patterning of contacts holes for nanoscale devices. As a first
example, we chose 22 nm technology node SRAM as a demonstration vehicle to testify

20

our patterning strategy for contact holes using block copolymer DSA since SRAM is an
industry-standard ramp-up vehicle for logic process technology development [3]. Figure
3.7a shows a device layout directly derived from the 22 nm 6T-SRAM cell demonstrated
by IBM [41]. In the layout, the contact holes were designed with multiple pitches and
orders, and elongated holes were used to connect the polysilicon gate and the active
source/drain diffusion region (Figure 3.7a, rectangular connections). To improve the
design regularity, the elongated connections can be replaced by two identical square
contact holes with smaller pitch (~45 nm) (Figure 3.7a, indicated with blue circles),
which can hardly be patterned by the current optical lithography. The DSA process needs
to generate hole patterns with pitches and orders for this SRAM design in a single
process step.

Based on the multiple types of patterns achieved using small topographical templates
shown above, we fabricated small guiding templates (shown in Figure 3.7c) according to
the DSA-friendly SRAM layout in Figure 3.7. The templates were patterned on 300 mm
Si wafers by 193 nm immersion lithography with single exposure and etched with
conventional reactive ion etching process. The fabricated templates have an average size
of about 66 nm (analyzed in Figure 3.7e&f). Larger size elliptical templates were
fabricated to generate contact hole pairs for the connection between polysilicon gate and
active source/drain region. Contact holes with average size of ~25 nm with 3 nm standard
deviation were successfully generated after performing the PS-b-PMMA DSA process
(shown in Figure 3.7d). The DSA generated holes were located at the centroids of the
guiding templates with overlay accuracy about 1 nm (Figure 3.7f). The contact hole pairs

21

were generated in the larger elliptical templates with an average pitch of ~30 nm, which
are suitable for making the connections for contacting the polysilicon gate and active
source/drain region. Thus, contact holes for the 22-nm SRAM devices were successfully
patterned using simple PS-b-PMMA DSA process with small templates of 66 nm
patterned by the current 193 nm immersion optical lithography. The critical dimension
limit of the 193 nm immersion optical lithography has been shrunk down by a factor of
two to three. This is the first time the contact hole pattern for an industry-standard circuit
is demonstrated using DSA combined with 193 nm immersion lithography.

200 nm

Active region

Polysilicon gate

Contact hole

e 40%

Mean: 66.4 nm
SD: 6.9 nm

Count

Mean: 24.8 nm
30% SD: 3.0 nm

Connection

50%

horizontal (mean 0.9 nm)

vertical (mean 1.2 nm)

40%

Count

30%

20%

20%

10%

200 nm

0%

10%

20 30 40 50 60 70 80

Hole Size (nm)

0%
-5 -4 -3 -2 -1 0 1 2 3 4 5

Centroid Deviation (nm)

Figure 3.7 Contact hole patterning for 22 nm SRAM cell. a, contact hole layout derived
directly from the IBM 22 nm 6T-SRAM circuit. b, modified layout by replacing the

22

rectangular connections with square holes and one of the modification is highlighted by
blue circles. c, guiding templates fabricated with 193 nm immersion optical lithography.
d, PS-b-PMMA DSA in the guiding templates. e, statistical results for the patterned
contact hole size distribution (blue) and the fabricated guiding template size distribution
(red). f, the centroid deviation of the patterned features from the guiding templates.

3.6 DSA Alphabet Concept

We have shown that using small physical guiding templates, one could flexibly control
the shape and position of the DSA pattern; this makes it possible to use 1-hole DSA
patterns to pattern contact holes for memory (Section 3.5) and random logic circuits
(Chapter 5). However as technology scales down, the contact/via density scales up, which
simultaneously opens the possibility of using multiple-hole DSA patterns for contact hole
patterning and brings in the challenge of printing guiding templates at a small pitch.
Using DSA for patterning IC contacts requires further knowledge of the placement of
contacts in an IC layout, as the placement of contacts in the IC layout determines the
shape and size of the required templates. For example, an isolated contact could be
patterned using a small circular template generating one single DSA hole, and two
adjacent contacts could be patterned using an elliptical template generating a pair of DSA
holes. While these two examples seem simple and straightforward, the use of these
simple templates alone is not sufficient in a practical situation. More types of contact
combinations and template shapes are needed to decompose a full integrated circuit chip
contact layer with billions of contacts.
We hypothesize that there exists a limited set of guiding templates analogous to the
letters of an alphabet that can cover all the possible contact hole patterns of a full chip
contact layer (Figure 3.8). This alphabet approach would significantly simplify DSA
contact hole patterning when the total number of letters of the alphabet is small and
would allow us to focus on fully characterizing only the design spaces for the letters of

23

the alphabet. By positioning these letters in various locations we would be able to pattern
the full chip contact layer in the same way that the 26 letters of the English alphabet is
sufficient to compose an English newspaper. Some of the most basic letters, such as
circular templates for 1-hole DSA patterns and elliptical templates for 2- and 3-hole DSA
patterns, have been studied extensively [42-44]. To establish a complete alphabet, though,
requires the examination of the entire standard cell library, as well as the optimization of
the layout to further reduce the number of letters in the alphabet as discussed in Du et al
[45].

Figure 3.8 (a) Examples of three standard cells [45]. (b) An example of the alphabet set
based on the cells in (a). With the templates in this alphabet set, one can pattern all the
cells in (a) using DSA.

24

Chapter 4
Computational simulation of DSA in
small physical guiding templates1

4.1 Introduction
Realizing the importance of aperiodic DSA patterns and small physical templates, we
decided to enhance our understanding of the DSA process using computational
simulation. The computational simulation not only identifies the factors that influence the
self-assembled patterns but also provides the guidance for the design of guiding templates
for device fabrication. In this chapter we demonstrate the computational simulation of
BCP DSA in guiding templates based on Self-Consistent Field Theory [46] and present a
comparison of these simulation results with our experimental data [42].
In order to simulate the influence of physical guiding templates on DSA process, a
masking method [47] is used to model confinement of cylinder domain forming BCP in a
3D simulation box. Sidewall and substrate each have the w parameters describing the
attractiveness of different polymer blocks to sidewall or substrate. Negative w refers to
the wall material that has a higher affinity for the minority polymer block compared to
the majority block, and vice versa.

2013 Society of Photo Optical Instrumentation Engineers. Part of this chapter has been reprinted, with
permission, from H. Yi, A. Latypov and H.-S. P. Wong, Computational Simulation of Block Copolymer
Directed Self-Assembly in Small Topographical Guiding Templates, SPIE Advanced Lithography, 8680,
86801L, 2013.
The work in this chapter has been done in close collaboration with Azat Latypov at Globalfoundries.

25

We first studied the influence of w parameters on the DSA result. While sidewall is kept
as minority block affinitive, increasing substrate w parameter from negative to positive
changes the minority block cylinders from hanging in the middle to reaching the bottom.
Another study is also carried out for guiding template sizes from 50nm to 90nm and
different polymer film thicknesses, which is also verified and matched well the
experiments. In both experiments and simulations, the formation of a cylinder is observed,
followed by the formation of an isolated minority phase cavity at the center of guiding
template, as the well size increases. It is found that an increase in polymer film thickness
will also lead to the formation of a cylinder or the formation of an isolated minority phase
cavity at the center of the guiding template. In experiments we observed the same
behavior as the spin coating speed is decreased, which leads to an increase in polymer
film thickness. Also, according to the image analysis results from experiments, as the
guiding template size varies from 68 nm to 78 nm, the average DSA hole size stays
around 19 nm. This agrees with simulation results demonstrating that DSA hole size is
18.9 nm when template size is 65 nm, 70 nm and 75 nm.

4.2 Self-Consistent Field Theory (SCFT)

The SCFT DSA Solver used in this work uses the model based on the following SCFT
Hamiltonian of a compressible diblock copolymer melt [47, 48]:
H [ w A , wB ] =
( 1 " w w %2 " 2
w
2 " w %% wA + wB
A
w
C dx *

$1 '''
$ B
' + $$
*) AB N # 2 & # AB + 2 0 AB + 2 # 0 && 2
2
" w A + wB %
1
w wA ,

.
$
' 2 w w B
AB + 2 # 2 &
AB 0
2 .C V lnQ[wA , wB ],

(4.1.1)
The diblock BCP described by this Hamiltonian has a degree of polymerization N and
an unperturbed radius of gyration R g . This Hamiltonian is a functional of unknown

26

self-consistent fields (potentials) wA = wA (x) and wB = wB (x) , acting on the phases A

and B of the diblock copolymer (BCP), respectively.

The BCP melt is confined in a

guiding template, which is defined by a prescribed wall density w = w (x) . The

parameter AB is the Flory parameter characterizing the interaction of the BCP phases
with each other, w = ( wB wA ) / 2 and w = ( wA + wB ) / 2 are the Flory-like
parameters, where wK (K=A or B) characterize the interaction of the BCP phases with
the wall material [47, 48]. The BCP melt is assumed to be compressible, with its
compressibility

described [47,

48]

by

applying

quadratic

penalty

term

( A + B + w 0 ) 2 , where is a compressibility parameter, 0 is the constant

total density and K = K [wA , wB ], where K is either A or B, are the densities of the BCP
phases A and B, which are known non-local functions of the self-consistent fields. The
constant C = 0 R g3 / N serves to normalize the value of this Hamiltonian. The quantity

is the spatial average of the combined normalized BCP density = ( A + B ) / 0

over the simulation volume V and Q[wA , wB ] is a single polymer chain partition
function, which is a known functional of the self-consistent fields [46].
Using the mean-field approximation, we are looking for the self-consistent fields
providing a saddle point of the SCFT Hamiltonian (4.1.1) [46]. The non-local
dependence of the densities of the BCP phases on the fields is found by solving
Fokker-Planck reaction-diffusion equations for propagators of diblock BCP in given
potential fields [47, 48].
The current version of SCFT DSA Solver implements two options to solve the
Fokker-Planck equations a finite-difference solver or a pseudo-spectral solver based on
an operator splitting (process splitting), as described in G. Fredrickson et al [46]. Two
options have also been implemented for the iterative algorithm for finding the saddle
points of the SCFT Hamiltonian explicit Euler iterations or semi-implicit Seidel
iterations [46, 49].
Although the interaction of the BCP material with the wall is characterized by two
parameters, e.g. w and w in (2.1.1) or wA and wB , the dependence of the DSA result

27

on the value of w should be weak for incompressible, or nearly incompressible

( ) BCP melts, as can be seen from the term with w in (2.1.1).

In the

simulations presented below, we used a simplifying relation between the Flory

parameters, 2 w = AB , for all wall and substrate materials. This assumption is not
expected to strongly affect the solutions with a sufficiently high value of a
compressibility parameter . The parameter w , characterizing the difference in affinities
of the two BCP phases with the wall/substrate materials, has been varied in our
parametric studies, as described below.

4.3 Simulation Study

4.3.1 Parameters
We simulated directed self-assembly of cylinder-forming PS-b-PMMA block
copolymers, the volume fraction of majority phase is set to f = 0.7 to match the
experiments described in chapter 3. The degree of polymerization was set to N = 653 and
unperturbed radius of gyration was set to Rg = 10 nm. The Flory Huggins parameter of
diblock copolymer was chosen as AB N=23.9 . These parameters correspond to a
PS-b-PMMA diblock copolymer with molecular weight around 46.1 21kg/mol. The
lateral spatial grid size in x and y direction was chosen to be 0.25 Rg, while in normal z
direction we chose a smaller grid size as 0.1Rg to attain sufficient resolution of
cross-section morphology.

A masking method [47] is used here to model confinement of block copolymers in a

three-dimensional simulation domain. The simulation domain includes sidewall and
substrate of the guiding template, spin-coated diblock copolymer, and a top surface
representing air. There are three different wall affinity parameters and density functions
are introduced for sidewall, substrate and air, each representing the interaction with PS

28

and PMMA phase differently. Sidewall and substrate each has a wall w factor that
depicts the attractiveness of different polymer blocks to sidewall or substrate. Negative
wall w refers to the material that has stronger affinity to a minority polymer block,
compared to its affinity to a majority block, which in our case corresponds to PS-wetting
materials. Air has no interaction with block polymers but acts as a physical confinement
and defines the thickness of polymer film on top of the sidewall (Figure 4.1). These three
density functions vanish inside the physical domain and rise smoothly away from the
domain boundaries. This method is highly adaptive to different confinement wall
geometries and the topography of block copolymer film.

Top view

1
Phase B

air

0.8

Phase A
0.6

sidewall

Cross section view

air

substrate

Phase B
sidewall

Phase A

Diblock Copolymer

substrate

0.4

0.2
0

Figure 4.1 (a) SCFT Simulation domain includes sidewall, substrate of the guiding
template, block copolymer and air. (b) Top view and the cross section view of the density
profile (Phase A volume fraction).

4.3.2 Guiding template affinity parametric studies

We first studied the influence of w parameters of the confining sidewall and the
substrate. A series of simulations is carried out with varying sidewall and substrate
affinity values. The polymer parameters are the same as stated in section 4.3.1. The CD
of guiding template is 8Rg on the x and the y direction, and 4 Rg on the z direction. The
sidewall has a higher affinity for minority block while the substrate w parameter is
varied between strongly minority block affinitive and strongly majority block affinitive.
It is observed in Figure 4.2 that when the substrate is neutral, the minority block cylinder

29

inside the guiding template can reach down to the guiding template substrate. However,
when sidewall and substrate are both minority block affinitive (i.e. the same material), the
minority block cylinder becomes disconnected from the substrate by the majority block
region.

a"

sub
wall

-25

-20

-15

-10

-5

10

15

20

25

0
-5
-10
-15
-20
-25

b"

sub
wall

-25

-20

-15

-10

-5

10

15

20

25

0
-5
-10
-15
-20
-25

Figure 4.2 Majority phase colormap plots of DSA (a) cross-section (b) top-view in an
808040 nm guiding template for various values of the wall and substrate affinity
parameters, wallN and subN. wallN is varied from 0 to -25 and subN is varied from -25 to
25. Negative N refers to the material that has higher affinity to the minority polymer
phase than majority phase. The majority phase is shown in red while the minority phase,
substrate, sidewall and air are all blue. The green dot line separates the region where

30

minority phase forms a hanging cylinder and the region that it forms a cylinder down to
the bottom.

4.4 1-hole DSA pattern design space

We conducted another parametric study, focusing on 1-hole DSA pattern forming inside
single guiding template. In our experiments, a series of guiding templates were fabricated
using EUV lithography and dry etching on 300 mm Si wafer with sizes ranging from 45
nm to 120 nm and depth around 50 nm. The DSA process is similar to the previous report
[43], and we used different spin speeds to control the thickness of block copolymer film.
As guiding template size increases from 45 nm to 70 nm, DSA pattern changes from
blurred a dark spot to a clear 1-hole pattern. However if the size increases further, a large
dark region or even multiple holes begins to form inside guiding template (Figure 4.3). A
similar trend is also observed when we decrease the spin coating speed, which leads to an
increase in polymer film thickness.
To explain this observation, we used SCFT DSA solver to explore a set of simulation
conditions, with different guiding template sizes from 50 nm to 90 nm and different
polymer film thicknesses. The depth of guiding template was set to 50 nm. The polymer
parameters were the same as stated in 4.3.1. The w N values for sidewall and substrate
were both chosen as -25 (see the green box on the left bottom corner of Figure 4.2),
because the whole guiding template is made of Si (strongly PMMA affinitive). Figure 4.4
shows the cross-sections of SCFT simulation results, which exactly match our
experimental observations. When the guiding template size is less than 55 nm, there is
hardly any PMMA segregation region, except in the center of the top surface. As the well
size becomes large enough, in the center of guiding template an isolated PMMA cavity is
formed, with a PMMA ring on the top surface. The formation of isolated PMMA cavity
also occurs as the polymer film becomes thicker. These results provide us with another
perspective looking into the design space of 1-hole DSA pattern.

31

45nm

57nm

61nm

68nm

73nm

78nm

83nm

88nm

93nm

2500rpm

2000rpm

1500rpm

1000rpm

Figure 4.3 (a) SEM images of DSA patterns confined by templates with different sizes
from 45 nm to 93 nm. Each number represents the CD of the guiding template in the
image below. (b) SEM images of DSA patterns with different polymer film thickness.
The CD of these guiding templates is 80 nm. The number/rpm represents the spin speed
of block copolymer solution in the image below.

32

a"

Thickness

CDwall

10

12

14

16

18

20

10

12

14

16

18

20

50
55
60
65
70
75
80
85
90

b"

Thickness

CDwall

50
55
60
65
70
75
80
85
90

Figure 4.4 Parametric study of DSA (a) cross-section (b) top-view in a 50 nm deep square
guiding template for various values of the wall size and polymer film thickness, which is
defined as the distance between the bottom of air and top of sidewall. In this study wallN

33

and subN are both -25. The majority phase is shown in red while the minority phase,
substrate, sidewall and air are all blue. The green dot line separates the region where
minority phase forms a hanging cylinder and the region that forms an isolated minority
phase cavity in the center of guiding template.

4.5 Summary
We studied DSA of compressible diblock copolymer melt inside small physical guiding
template using Self-Consistent Field Theory (SCFT) computational simulations. In this
chapter, we focused on the 1-hole DSA pattern due to its potential for via/contact hole
patterning in integrated circuits. We first explored the influence of guiding template
wall/substrate affinity parameters on the cross-section of DSA pattern. A connection
between the minority phase cylinder and the substrate is observed when the substrate is
changed from minority phase affinitive to neutral.

Then we further explored the

influence of guiding template sizes and polymer film thickness on DSA graphoepitaxy.
When guiding template size or polymer film thickness increases, a transition of the DSA
pattern from a blurred dark spot to a clear 1-hole pattern occurs. If the size or film
thickness continues to increase, a large dark region or multiple holes appears inside the
guiding template. The experimental SEM results are compared with the results of
simulations obtained from the computational SCFT model of DSA.

34

Chapter 5
A General Strategy of Directed
Self-Assembly (DSA) Guiding
Template Design for Contact Hole
Patterning4
5.1 Introduction
In chapter 3, we discussed using physical guiding templates to generate and control
aperiodic DSA patterns and introduced the concept of DSA alphabet for contact
patterning. For the first time we demonstrated DSA contact patterning for 22 nm SRAM
cell. However we realize that patterning for random logic circuits is more challenging
because contacts are placed more randomly than memory circuits. Although the small
guiding templates are shown to guide the self-assembly off the natural geometry by
strong boundary confinement, it is insufficient to simply surround contact holes with
guiding templates without optimizing the placement and geometry of the guiding
templates.
As the first step towards implementing the alphabet approach, a systematic design

2015 American Chemical Society. Part of this chapter has been reprinted, with permission, from H. Yi,
X.-Y. Bao, R. Tiberio, and H.-S. P. Wong, A General Design Strategy for Block Copolymer Directed
Self-Assembly Patterning Integrated Circuits Contact Holes using an Alphabet Approach, Nano Letters,
15 (2), pp 805812, 2015.

35

strategy is necessary to translate the placement of contacts into template designs. This
design strategy must not only be applicable to one technology generation, but also must
be generally applicable to future technology generations. To develop such a design
strategy, we need to simultaneously consider the material properties of the block
copolymer as well as the size and overlay accuracy requirements for the target
technology nodes.
In this chapter, we describe a general template design strategy that links the DSA
material properties to the target technology node requirements. The design strategy falls
into three categories: 1) use of 1-hole template for every contact, 2) use of peanut-shaped
template for adjacent contact pairs, and 3) use of multiple-hole templates (e.g. 3-hole
elliptical templates or 4-hole square template) for adjacent contact groups. Note that the
peanut-shaped templates are used because they can extend the DSA hole pitch beyond its
natural limits and are easier to print than two separate closely spaced 1-hole templates.
To demonstrate DSA contact hole patterning using each category of the strategies
mentioned above, we use the same block copolymer (PS-b-PMMA) to pattern the same
logic cell (1-bit Half Adder) in three technology nodes: 14 nm, 11 nm, and 7 nm [50, 51].

5.2 Technology Node and Contact Pitch

5.2.1 Unidirectional Layouts
During the course of device scaling in the past decade, industry has realized that simply
shrinking the layout geometries from one technology node to the next is no longer viable
because of the proximity effects in optical lithography. Instead of having bidirectional
structures, such as bent metal lines, IC layouts have now evolved towards gridded
unidirectional layouts. Unidirectional IC layouts represent an important IC development
trend due to their larger process window and higher yield, features for which they have
been widely adopted by the semiconductor industry since the 45 nm technology node.
Such layouts generally involve an IC layout architecture wherein each device layer

36

consists only of unidirectional lines and contact holes positioned at pre-determined grid
points [52, 53]. The bends in the previous bidirectional layouts are implemented by
orthogonal lines on separate layers connected by contacts. The irregularity of the contact
distribution is greatly reduced by aligning the contacts to regular grid points. While
unidirectional IC design is not a must for DSA, the use of unidirectional layout design
could greatly improve the throughput and yield for DSA contact patterning. Because the
adoption of a unidirectional IC design would fix the pitch between adjacent contacts to a
set value determined only by the poly-gate pitch and the metal-line pitch (Figure 5.1).
The three smallest pitches between contacts in different technology nodes are listed in
Table 2 [30, 54]. As printing the closely but irregularly positioned contact groups is one
of the most challenging problems for contact/via printing, it is preferable to limit the
possible distance between contacts to a few fixed numbers. In this way, the guiding
template design only needs to target a few fixed contact pitches and combinations, thus
cutting down the types of templates required to cover all the contact configurations in a
full chip layer. In other words, the size of the alphabet is reduced. A smaller alphabet
could lead to simplified template design, improved template printing quality, and higher
yield for DSA itself [45]. The benefit for adopting DSA for contact patterning is
maximized when the contact pitch matches the natural geometry of the block copolymer
DSA pattern, especially in the scenario that multiple DSA holes are generated inside one
template [43].

37

Figure 5.1 (a) Example of a cell layout designed in unidirectional style [55]. All lines are
designed with the same width and pitch, and the contact holes are positioned only at
pre-determined grid points. The lines of poly-gate (purple) are perpendicular. Some
closely positioned contacts are marked with red circles. (b) Four smallest contact pitches
in (a), which correspond to the contact pairs in red circles except for the first pair.

Node (nm)

20

14

11

10

X-Pitch (PX)

90

82

64

58

46

40

Y-Pitch (PY)

70

64

50

44

36

32

83

76

59

53

43

38

114

104

81

73

58

51

2.0
8.0

1.4
5.1

1.1
3.6

1.0
3.2

0.7
2.3

0.5
1.4

CD control (3)

Overlay (3)

Table 2. Pitch length between adjacent contacts versus technology nodes [30].

5.2.2 DSA Design Space Revisited

Once the distance between closely positioned contacts in certain technology nodes is
determined, it is crucial to know whether these distances can be achieved using block

38

copolymer DSA. If the contact pitch is somewhere close to the natural pitch of the block
copolymer DSA pattern, then multiple-hole templates corresponding to various letters in
the alphabet may be adopted to pattern a group of closely positioned contacts. By varying
the size and the shape of templates, the DSA hole pitch can be adjusted within a certain
range of the natural block copolymer pitch to match contact pitch in the layouts. The
relation between the DSA hole pitch and the size and shape of the templates can be
considered as constituting the DSA design space for the block copolymer self-assembly
into cylindrical patterns. As we discussed in chapter 3, DSA design space provides us
with the limits of DSA hole pitch for a specific guiding template type. From Figure 3.6,
we observe that the pitch of a 2-hole DSA pattern must be between 36 and 50 nm, while
the pitch of a 3-hole DSA pattern cannot exceed 48 nm. Therefore, when the pitch of a
contact pair is larger than 50 nm, it becomes impossible to use 2-hole DSA pattern inside
an elliptical templates to pattern them. On the other hand, a contact pair of pitch 40 nm
could be patterned by 2-hole DSA pattern. For a full-chip design, analogous design
spaces for the other letters in the alphabet must be established as well.

5.3 DSA Guiding Templates Design Strategy

The template design strategies described above are built on knowledge of the contact
pitch and DSA design space. To determine which strategy to use, we must first compare
the contact pitch in a given technology node with the adjustable DSA hole pitch. This
comparison, regardless of the technology node and the polymer type, can lead to three
different results (Figure 5.2): (1) the minimum contact pitch is larger than the resolution
of conventional optical lithography, (2) the minimum contact pitch is larger than the
maximum DSA hole pitch but smaller than the resolution of conventional optical
lithography or (3) the minimum contact pitch falls within the range of the DSA hole pitch.
Each result requires the use of a different template design strategy with its own pros and
cons. There are also various quality metrics to take into account when evaluating a
template design, such as the size distribution and overlay accuracy of the DSA holes as
well as the designs tolerance to variations in the lithography process. It is important to
note that the overlay accuracy, namely the registration of the DSA patterns with features

39

already patterned on the wafers, is one of the factors most strongly impacting the DSA
pattern quality. As the dimensions of ICs are scaled aggressively, the budget for overlay
is shrinking quickly as well.

The overlay requirement for 14 nm and 11 nm logic

circuits are 5.1 nm and 3.6 nm, respectively [30]. Therefore, accommodating the overlay
accuracy of DSA patterns within the overall overlay budget is of paramount importance
for implementing DSA for next generation lithography.
Contact layout

BCP
Max
pitch

Contact
Lithography Min
Resolution
pitch

1st strategy:
1-hole templates
for each contact

BCP
Max
pitch

Contact
Lithography
Min
Resolution
pitch

2nd strategy:
Peanut-shaped templates for
closely positioned contacts

Contact
Min
pitch

BCP
Max
pitch

Lithography
Resolution

3rd strategy:
Multiple-hole templates for
closely positioned contacts

Figure 5.2 Demonstration of the three strategies described above. From left to right: (1)
the minimum contact pitch is larger than the resolution of conventional optical
lithography. (2) The minimum contact pitch is larger than the maximum DSA hole pitch
but smaller than the resolution of conventional optical lithography. (3) The minimum
contact pitch falls within the range of DSA hole pitch.

The overlay accuracy of DSA consists of two parts: the intrinsic position error of the
DSA pattern and the pattern fidelity of the guiding template. The intrinsic position error
of DSA refers to the variation of DSA patterns inside an ideal guiding template and
depends chiefly on the chemical properties of the block copolymer, which will be

40

investigated in the future. The pattern fidelity of the guiding template refers to the
position error of the DSA pattern brought about by a non-ideal template, which may shift
the position of the DSA pattern with respect to the designed location. For instance, there
is a size offset between non-ideal templates and ideal targets caused by the process
variations of conventional lithography, such as focus and dose variations, mask alignment
errors, and various lens aberrations. Since the DSA pattern is sensitive to the shape and
size of the guiding template, such lithography process variations lead to size variations in
the guiding templates and ultimately to nontrivial position error in DSA patterns. In this
paper, we use e-beam lithography to fabricate the guiding templates as an efficient
method for the purpose of concept demonstration. But we must also note that the process
variations of optical/EUV lithography can hardly be represented by e-beam lithography,
and their impact on DSA patterns cannot be ignored. Researchers from IBM and ASML
have addressed studies combining computational lithography and DSA [56-58] and the
results show that conventional lithography is capable of printing the DSA guiding
templates with acceptable errors.

5.4 DSA Contact Patterning from 14 nm to 7 nm node

5.4.1 1-bit Half Adder
To demonstrate our guiding template design strategy, we chose the logic cell 1-bit half
adder (HA-X1) as our testing layout for DSA contact patterning. Figure 5.3a shows
HA-X1 layout from a 45 nm open cell library. During layout optimization process, the
transistor sizes and connections (pin locations) of the logic cells were kept unchanged to
minimize any impact on the VLSI (physical) design flow at the full-chip level (Figure
5.3b). We shrank the cell layouts down to different technology node with poly gate pitch
and Metal-1 pitch lengths indicated in Table 1 [30, 54]. In addition, the size and pitch of
guiding templates must be consistent with the capabilities of conventional optical
lithography that has larger feature sizes. For example, Figure 5.3c shows an improper
contact layout design, where the corresponding guiding templates are too close to be

41

printed. To enable patterning of the guiding templates, the positions of two metal lines
are switched while maintaining the same circuit function and the same cell area as the
original layout. Thus layout optimization must ensure that contact holes can be patterned
by DSA and guiding templates are printable by conventional optical lithography (shown
in Figure 5.3d). Overall, we re-designed a total of 85 logic cells layouts (including
inverters, buffers, NAND, NOR, AOI, OAI, adders, and flip-flops) from the above cell
library into DSA-aware layouts. No area penalty is observed for any of these cells and the
full-chip VLSI (physical) design flow is largely unchanged, because all necessary
changes occur at the library cell level.

Metal&1

Poly

Ac-ve&Region

Contact

Switch(
Lines(
d

Design(Rule(
viola5on(
removal(
Figure 5.3 (a) Conventional half adder HA-X1 layout from Nangate 45 nm Open Cell
Library. (b) Re-designed DSA-aware half adder HA-X1 layout. (c) The two metal lines
switch their positions to remove design rule violation. (d) The contacts (orange boxes) are
moved to new locations so they can be patterned using lithography.

42

5.4.2 Strategy One: DSA Contact Patterning at 14 nm node

When the contact pitch is larger than the resolution of conventional lithography, one-hole
templates for each individual contact should be used. Since the minimum contact pitch in
this case is larger than the maximum DSA hole pitch, an attempt to use larger templates
to generate multiple holes will either lead to extra DSA holes or DSA holes with the
wrong pitch (Figure 5.4). The sole use of one-hole templates benefits from simplicity and
results in high tolerance to process variation. As shown in a previous study [42, 43],
one-hole DSA patterns can be generated inside circular templates ranging from 55 nm to
75 nm in diameter, where the DSA hole diameter remains approximately constant at 20
nm and the overlay accuracy at 1 nm. This means that even if the template size varies due
to process variations such as focus and dose variations, a 1-hole DSA pattern can still be
generated successfully. On the other hand, as this strategy requires that every template be
separated, thereby necessitating a design rule check to ensure templates do not merge due
to lithography resolution limitations and cause defects [59]. In our HA-X1 example, the
contacts were successfully patterned with DSA using strategy (1) described above. The
DSA contact holes were achieved with uniform dimensions and the positional accuracy
of 1 nm on both horizontal and vertical directions. (Figure 5.5) The positional accuracy is
measured as the centroid offset between DSA holes and guiding templates.

43

Figure 5.4 (a) An axis showing that the maximum DSA hole pitch is smaller than the
smallest pitch length for the 14 nm node. The maximum and minimum pitch values for
the BCP are obtained from the design space in Figure 3. (b) Circuit layout of one-bit half
adder. (c) Template design with only 1-hole templates. (d) Template design with larger
templates for 2 or 3 holes. (e) Successful DSA patterning using design in (c). (f) Failed
DSA patterning using design in (d), as extra holes appear in over-sized templates. Scale
bar: 200nm.

44

50%

20%

10%

0%

12

13

14

15

16

17

18

HHole)size)(nm))
o le1S iz! e1(n m )

20%

10%

@4

@3

@2

@1

10%

46

48

50

52

54

56

58

60

62

64

40%

30%

0%

20%

TTemplate)size)(nm))
em p late1S iz e1(n m )

Mean 1D eviatio n @X :11.0n m

b))))

30%

0%

19

Mean :151.3n m
S td :12.6n m

40%

Count)Frequency)
C ount

Mean :114.3n m
S td :11.7n m

30%

a))))
b) 40%
Count)Frequency)
C ount

40%

Count)Frequency)
C ount

Count)Frequency)
C ount

a)

Mean 1D eviatio n @Y :11.0n m

30%

20%

10%

0%

@4

@3

@2

@1

CCentroid)Devia-on)(nm))
en tro id 1D eviatio n 1(n m )

CCentroid)Devia-on)(nm))
en tro id 1D eviatio n 1(n m )

Figure 5.5 14 nm HA-X1 DSA contact hole patterning analysis. (a) Patterned contact
hole size distribution (red) and the fabricated guiding template size distribution (blue).
Contact holes of 14.3 nm size were generated in guiding templates of 51.3 nm size. (b)
Overlay accuracy of DSA contact holes in x and y direction, average deviation: 1.0 nm.

The 1-hole DSA patterns are not only useful for patterning circuits at 14 nm node, but
also can be a solution for printing isolated contacts for even smaller technology nodes. As
there always exists isolated contacts in most circuits regardless of technology nodes,
single DSA contacts will always be applicable. The 1-hole DSA pattern have the benefits
of shrinking the guiding template size from a larger printed pattern, improving contact
CD uniformity (CDU), as well as maintaining and satisfying overlay accuracy [42, 60].
While it is very challenging but still possible to print 20 nm contacts using conventional
lithography at present, as the technology feature size scales down further it would
become extremely difficult to pattern contacts much smaller. By that time, it would be
crucial to have the capability to shrink contact sizes from a template that can be printed
by optical lithography. This is the main application of 1-hole DSA patterns.

45

5.4.3 Strategy Two: DSA Contact Patterning at 11 nm node

Alternatively, if the minimum contact pitch is larger than the maximum DSA hole pitch
but smaller than the resolution of conventional optical lithography, one faces a dilemma
between using one-hole templates and multiple-hole templates, such as 2-hole or 3-hole
elliptical templates. A very simple example involves patterning a pair of contacts with the
pitch of 64 nm, as required in the 11 nm node (Figure 5.6). If a 2-hole elliptical template
is adopted, the required contact pitch will be larger than the maximum DSA pitch, thus
resulting in an extra hole in the middle. On the other hand, if two one-hole templates are
adopted with each template around 60 nm in diameter, then conventional optical
lithography will be unable to resolve the 4 nm gap between these two templates. One
possible solution to this conundrum is to instead use a special kind of guiding templates
that is peanut-shaped (Figure 5.6). The peanut shapes are specifically important, as they
represent the case where two circular templates designed to be separate merge due to the
limited pitch resolution of conventional lithography. Because a two-hole pair can still
form inside the merged template as a useful pattern, the peanut-shaped templates
effectively extend the DSA hole pitch. Knowing that the DSA hole pitch inside the
peanut-shaped templates is strongly influenced by the neck of the peanut, one can use
three parameters to characterize the peanut shape: the template length and width, and the
width of the connection part. Among these three parameters, the connection width has the
most influence on the distance variation between the two holes. The DSA design space of
peanut-shaped templates is shown in Figure 5.7. Based on this design space, the
templates for DSA contact patterning for the 11 nm HA-X1 are re-designed and the DSA
result is shown in Figure 5.8. As the templates turn into peanut shapes, the centroid of
each template does not correspond to each DSA contact holes any more. Therefore,
instead of measuring the positional accuracy defined above, we consider global overlay
accuracy, which measures the position of each DSA hole with respect to the contact
positions in the layout. DSA contact holes were achieved with the global overlay
accuracy of 1.9 nm on horizontal direction and 2.5 nm on vertical direction, including ~2
nm position error of templates brought by e-beam lithography.

46

Guiding'template'design'
64'nm'

DSA'result'

X"

'
''4'nm'

Contact'layout'
'

a!
64'nm'

Figure 5.6 Different choices of the guiding template design for the contact pair shown on
the left. The first one uses an elongated elliptical template but generates more DSA holes
than desired. The second one uses two circular templates. However, due to lithography
resolution, the small gap in between could not be resolved. Therefore, the second strategy
actually results in a peanut shape.

90
135n m
140n m
145n m
150n m

80
75
70
connection
width

65

Template
width

60
55
50

D S A H o le P itc h (n m )

85

Template length

20

30

40

50

60

70

C o n n ec tio n Wid th (n m )
Figure 5.7 Design space of DSA pattern in peanut-shaped template. Each data point
represents a specific peanut-shaped template length and connection width combination
that leads to a 2-hole pair. Different point symbols represent different template lengths.
The standard deviation of the hole-pitch is reflected as the error bar in y-axis.

47

Mean :.21.1n m !! S td :.2.0n m

30%

20%

10%

0%

16

18

20

22

H o le.S iz e.(n m )

Hole&Size&(nm)&

24

26

f
30%

Mean ,D ev iatio n !X :,1.9n m

20%

10%

0%

!4

!2

G lo b al,O verlay,A c c u rac y,(n m )

Global&Overlay&Accuracy&(nm)&

Count&Frequency&
C o u n t/F re q u e n c y

40%

Count&Frequency&
C o u n t,F re q u e n c y

Count&Frequency&
C o u n t.F re q u e n c y

50%

Mean /D ev iatio n !Y :/2.5n m

40%
30%
20%
10%
0%

!9

!6

!3

G lo b al/O verlay/A c c u rac y/(n m )

Global&Overlay&Accuracy&(nm)&

Figure 5.8 (a) An axis showing that the maximum BCP pitch is close to the minimum
contact pitch for the 11 nm node but smaller than the lithography resolution (~60 nm).
Peanut-shaped templates can be designed for the closely positioned contact pairs. (b)
Circuit layout of one-bit half adder given in Figure 5.3b. (c) 11 nm HA-X1 DSA contact
hole patterns with layout given in (b). (d) Size distribution of DSA contact holes. (e&f)
Global overlay accuracy of DSA contact holes in x and y direction, including ~2nm
position error of templates brought by e-beam lithography. The average global overlay
accuracy 2.5nm. While some templates merge together during fabrication process, the
self-assembled holes are still separated with reasonable size and overlay accuracy,
showing high tolerance for template defects. Scale bar: 200 nm.

5.4.4 Strategy Three: DSA Contact Patterning at 7 nm

node

48

For the third type of result, namely the minimum contact pitch falls within the range of
DSA hole pitch, one can use multiple-hole templates for generating multiple holes. In this
scenario, the pitch of the DSA holes must match the contact pitch in the layout to avoid
positional errors. To produce the targeted hole pitch, the geometry of the templates are
determined from the corresponding DSA design space. An example of using this strategy
to pattern a 7 nm node one-bit half adder is shown in Figure 5.9. First, the contact pitch is
obtained from the targeted technology node. The appropriate template size and shape can
then be found in the DSA design space (Figure 5.9c) using the targeted contact pitch. The
advantage of this strategy is that the resolution requirement for lithography is relaxed,
because the templates can merge as long as their dimensions and shapes can be controlled.
Indeed, the stringent resolution requirement for the first design strategy necessitates that
we switch the position of two metal lines in the original half adder layout to prevent the
templates from merging due to the lithography resolution limit (see chapter 5.4.1). When
the third design strategy is adopted for 7 nm node, the resolution requirement for the
template is relaxed and both the original and modified half adder layouts can be patterned
(Figure 5.10).

The global overlay accuracy in Figure 5.10e and 4.10f are both 2.5 nm

on horizontal direction and 2.8 nm and 2.9 nm on vertical direction, including ~2 nm

position error of templates brought by e-beam lithography (Figure 5.11). Though this
strategy can flexibly adapt to different layout designs, it is important to note that this
strategy also needs to satisfy higher overlay accuracy requirement, as it is often targeted
towards smaller technology nodes such as 7 nm node (Table 1). Therefore, the pattern
fidelity of the guiding templates is very important and must be controlled carefully.

49

[+33!

BCP+
+(Px/2)2+++PY2+ PX! Max+pitch! +Px2+++PY2+
42!

36!

Pitch:+42nm+

46!
56

52
48

Figure 5.9

58!

A rea.ratio :.1.5
A rea.ratio :.2.5
A rea.ratio :.3.5
A rea.ratio :.4.5

44
40
36
32

Length:+
120nm+

Length:++160nm+++++++++120nm+

50+]!

BCP+
Min+pitch! PY!

D S A .H o le .P itc h .(n m )

90

120

150

180

T em p late.L en g th .(n m )

29hole+
Pitch:+42nm+
Length:120nm+

39hole+
Pitch:+42nm+
Length:+160nm+

(a) An axis showing that the BCP pitch range matches with the smallest set

of contact pitches for the 7 nm node. Multiple-hole templates can be designed for all the
contact pairs with pitches within the achievable BCP pitch range. (b) Contact layout. All
the contact pairs with 42 nm pitch are labeled with red circles. (c) 2-hole and 3-hole
design space from Fig. 3. From this design space, we can find the template length
necessary to achieve 2-hole and 3-hole patterns with 42 nm pitch are 120 nm and 160 nm,
respectively. (d) Template design based on this strategy. Templates that generate holes
with 42 nm pitches are outlined red.

50

Figure 5.10 (a) DSA-aware HA-X1 (half adder) layout. (b) A similar DSA-aware HA-X1
layout that cannot be implemented by the first strategy due to the difficulty of printing the
three closely packed contacts (in the red circle) separately. With the third strategy, the
resolution requirement is relaxed and this layout can be patterned. (c) Contact layout
design based on the layout in (a). (d) Contact layout design based on the layout in (b). (e)
7 nm HA-X1 DSA contact hole patterns with the layout in (a). (f) 7 nm HA-X1 DSA
contact hole patterns with the layout in (b). Scale bar: 200nm.

51

50%
40%
30%
20%
10%

20

H o le/S iz e/(n m )
Hole&size&(nm)&

22

20%
10%

d 60%

C6

C3

30%
20%
10%
C6

C3

50%

20

22

Mean /D ev iatio n CY :/2.8n m

C9

C6

C3

G lo b al/O verlay/A c c u rac y/(n m )

Global&Overlay&Accuracy&(nm)&

Mean /D ev iatio n CY :/2.9n m

40%
30%
20%
10%
0%

18

10%

60%

40%

16

H o le/S iz e/(n m )
Hole&size&(nm)&

20%

0%

Mean /D ev iatio n CX :/2.5n m

14

30%

G lo b al/O verlay/A c c u rac y/(n m )

Global&Overlay&Accuracy&(nm)&

C9

10%

40%

50%

0%
C12

20%

50%

Mean /D ev iatio n CX :/2.5n m

Mean :/17.5n m !! S td :/2.0n m

30%

0%

24

30%

0%

CCount&Frequency&
o u n t/F re q u e n c y

18

CCount&Frequency&
o u n t/F re q u e n c y

Count&Frequency&
C o u n t/F re q u e n c y

40%

16

CCount&Frequency&
o u n t/F re q u e n c y

0%

Mea n :/18.5n m !! S td :/1.5n m

Count&Frequency&
C o u n t/F re q u e n c y

a
Count&Frequency&
C o u n t/F re q u e n c y

C10

C5

10

G lo b al/O verlay/A c c u rac y/(n m )

G lo b al/O verlay/A c c u rac y/(n m )

Global&Overlay&Accuracy&(nm)&

Global&Overlay&Accuracy&(nm)&

Figure 5.11 7 nm HA-X1 DSA contact hole patterning analysis. (a) Size distribution of
DSA contact hole patterned with layout in Figure 5.10e. (b) Size distribution of DSA
contact hole patterned with layout in Figure 5.10f. (c) Global overlay accuracy of DSA
contact holes patterned with layout in Figure 5.10e, including ~2nm position error of
templates brought by e-beam lithography. (d) Global overlay accuracy of DSA contact
holes in x and y direction patterned with layout in Figure 5.10f. Contact holes ~18 nm
were generated and the average global overlay accuracy 3 nm.

52

Chapter 6
DSA Interaction Range and Sub DSA
resolution Assist Features (SDRAF)5
6.1 Template Density Influence on DSA patterns
In the previous chapters, we have demonstrated and explored the flexible control of DSA
patterns through varying the geometric parameters of the guiding templates. However,
the formation of DSA patterns is also strongly influenced by the template density, which
has not been studied in a systematic and quantitative manner. The template density
influence could cause different DSA outcome and lead to increased DSA defectivity.
An example of template density influence is shown in Figure 6.1. This density influence
exists due to the fact that different template densities lead to different film thicknesses
during the spin coating process, and thereby influencing the DSA outcome (Figure
6.1e&f). Spin coating process is extensively used to dispense and coat BCP materials on
the planar surface, because it is capable of achieving highly uniform thin polymer film on
the flat surface. However it is important to note that with spin coating process, the coated

2015 Society of Photo Optical Instrumentation Engineers. Part of this chapter has been reprinted, with
permission, from H. Yi, J. Bekaert, R. Gronheid, G. Fenger, K. Nafus, H.-S. P. Wong, Study of DSA
Interaction Range using Gaussian Convolution, SPIE Advanced Lithography, 9423, 94232A, 2015.
2015 Society of Photo Optical Instrumentation Engineers. Part of this chapter has been reprinted, with
permission, from H. Yi, J. Bekaert, R. Gronheid, G. Vandenberghe, K. Nafus, H.-S. P. Wong,
Experimental Study of Sub DSA resolution Assist Features (SDRAF), SPIE Advanced Lithography, 9423,
94231F, 2015.
The work in this chapter has been done in close collaboration with IMEC.

53

film thickness on the patterned area is very different from the non-patterned flat area.
Furthermore, the density of the pattern will strongly influence the local film thickness.
Compared with the high pattern density region, lower pattern density region will end up
with a thicker coat of polymer film. This condition is described as overfilled (Figure
6.1f). The local polymer film thickness is a key factor of the DSA process, as the polymer
volume inside the guiding templates determines whether useful DSA patterns or defects
will form [42].

200#nm#

Polymer#not#overlled#

200#nm#

200#nm#

200#nm#

Polymer#overlled#

Template#

Template#

Figure 6.1 For the same size of guiding templates, different template density leads to
different DSA results. The template pitch is: (a) 200 nm; (b) 250 nm; (c) 300 nm; (d) 400
nm. When the template pitch increases, the number of DSA defects (the missing holes)
rise significantly. (e) A cross-section cartoon showing the template is not overfilled with
polymer when the template density is high. (f) A cross-section cartoon showing the
template is overfilled with polymer when the template density is low.

Realizing the fact that the contact density in integrated circuits could hardly be uniform
across the entire chip, it is critical to determine which region in the layouts may end up
with defects brought by the density influence. To characterize this density influence we
introduce the concept of DSA Interaction Range (DSAIR). DSAIR is akin to the concept
of interaction range in optical lithography, describing a certain range that template
density starts to influence DSA patterns. It is not fixed for a given block copolymer and is
also process-related.

54

There are several challenges in terms of systematically studying the density influence on
patterned area and finding DSAIR. It would be ideal to have a theoretical physics-based
model to predict the local film thickness variation brought by the templates. However,
this model must consider the fluid mechanics on nano-scale features as well as the
solvent evaporation process during the thermal annealing, which are overly complex and
difficult to model. In practice, a direct measurement of the film thickness through
cross-section SEM or TEM is more applicable, though the throughput is relatively low
and obtaining a cross-section is destructive. To quantify the DSAIR requires a huge
amount of data in order to be able to map the local film thickness with all kinds of
template density. Therefore, it is strongly desirable to create a behavioral model to
describe the influence of pattern density on DSA pattern formation. The model does not
need to account for all the detailed physics involved in the spin coating and annealing
process, but it should be able to determine whether the distribution of the contact density
will play a negative effect on the DSA pattern formation for a given DSA process.
In this chapter, we adopt a Gaussian Convolution model to simulate the density influence.
This model could be used to predict the location in the circuit layout where the overfilled
condition would happen due to low contact density. The degree of the overfilled
condition is correlated with the sigma value in the Gaussian function, which we use to
represent DSAIR. This convolution model would provide an important input to the DSA
models for predicting density influence for a given circuit layout. Then we
experimentally demonstrate an effective solution to mitigate the template density
influence and reduce DSA defectivity using sub-DSA resolution Assist Features
(SDRAF). These SDRAFs are small templates that cannot generate any transferrable
DSA patterns and they serve as reservoirs to the polymer to reduce the overfilled
conditions brought by the low contact density. SDRAFs of various sizes and designs are
demonstrated and proven to be highly effective.

6.2 Materials and Process

55

All the experimental data shown in this chapter were retrieved in IMEC using IMEC
DSA flow [22] and therefore it is different with the flow described in chapter 3. Figure
6.2 demonstrates the IMEC DSA flow. The guiding template substrate is made of silicon
and the sidewall is made of 100 nm spin-on carbon (SOC) and 30 nm spin-on glass
(SOG). We use PS-b-PMMA block copolymer provided by AZ Electronic Materials
(AZEMBLYTM PME585). This block copolymer has a natural period of 37.1 nm.

Dry%etch%

85 nm NTD resist
30 nm SOG
100 nm SOC
Substrate

193i%litho%

Coat%BCP%
and%anneal%
Remove%PMMA%

Figure 6.2 Schematic representation of DSA flow as implemented at IMEC.

6.3 Study of DSA Interaction Range using Gaussian

Convolution
6.3.1 Experimental Observation
It is critical to have a working test pattern design that is capable of reflecting the
influence of pattern density and enabling its extraction. It has been reported before that
the overfilled 1-hole templates lead to DSA defects [22]. To simplify the discussion we
focus on the 1-hole DSA guiding templates only. Our test pattern layout is shown in
Figure 6.3. This set of patterns consists of 87 array of 8 m 8 m blocks; each block

56

consists of a matrix of templates designed with the same size and pitch. All the blocks on
the same row share the same template size, while all the blocks on the same column share
the same template pitch. Therefore each block has a unique template size/pitch
combination, making it convenient to observe using SEM imaging the influence of
template density as well as the influence of template size on the DSA pattern.

Design CD
pitch

Mask design

Figure 6.3 The layout of guiding template design. The numbers on the top of the blocks
as well as the crosses and squares on left and right side are SEM alignment marks.

A typical trend of our experimental observations is shown in Figure 6.4. It shows DSA
results with the same template CD but different template pitch. As the pitch is increased
from 150 nm to 250 nm, the number of overfilled templates increases dramatically,
resulting higher defectivity. It is observed that the overfilled templates appear mostly at
the corners and edges of our design block, which create the illusion that the templates
have disappeared from a zoom-out view (Figure 6.3e&f), while the not-overfilled
templates have much higher contrast around the template edge. The distribution of the
overfilled templates is caused by the fact that the density at the block corner is much
lower than the density in the center. A lower template density leads to more reflow into
the templates and tends to make them overfilled. It is important to note that although
overfilled templates would result in DSA defects, not-overfilled templates do not
guarantee a non-defect condition. Whether a useful 1-hole DSA pattern will form also

57

depends on other factors such as the size of the templates. If the size of the template is
too small or too large, no valid 1-hole pattern will form even if it is not overfilled.
Realizing the fact that the overfilled and not-overfilled templates have an obvious
contrast difference in the zoom-out view, we consider each template as a binary pixel (0
is overfilled) and transform the low-magnitude SEM images into binary matrix (Figure
6.5). And through these binary images we observed there exists a certain similarity
between the distribution of the not-overfilled templates and the Gaussian convolution of
the template block.
c

150$nm$

f$

250$nm$

200$nm$

2$um$

2$um$

2$um$

Figure 6.4 (a)-(c): SEM image of DSA patterns on the same row (see Figure 6.2). The
CD of the templates in these three cases is 66 nm. Template pitch: (a) 150 nm. (b) 200
nm. (c) 250 nm. (d) Zoom out view of (a). (e) Zoom out view of (b). (f) Zoom out view
of (c).

58

a"

c"

Figure 6.5 (a) Low magnitude SEM image of one design block. (b) Binary image
processed from (a). (c) A binary matrix processed from (b). Each white circle in (b) is
treated as a white pixel in (c). (d) The average of 25 binary images retrieved from the
CDU wafer. (e) The average of 25 binary matrixes processed from (d).

6.3.2 Gaussian Convolution

The convolution operation was carried out using DENSITY_CONVOLVE function in
Calibre software [61]. There are two key input parameters in this function: the sigma of
the Gaussian function and the template layout. This function will conduct the convolution
and output the convoluted density map. Figure 6.6 shows the convolution results using
different sigma values. The sigma value corresponds to the concept of DSAIR, which
characterizes the distance that polymers reflow across during thermal annealing. To avoid
confusion, we still use the term sigma instead of DSAIR in the discussion below. As
the sigma value grows, the convolved density map appears more blurred, meaning more
polymer reflow and thus more templates become overfilled.

59

Sigma&=&0.2&

Sigma&=&0.6&

Sigma&=&1&

Sigma&=&1.6&

Sigma&=&2&

Figure 6.6 The gray-scale convolved density map using different sigma values (unit: um).
The input layout is the same as shown in Figure 6.3.

The sigma value directly determines the contour of the convolved density map and
provides an estimate of the number of overfilled templates. Also, this sigma should be
entirely dependent on the DSA process and is independent of the input test layout pattern.
Therefore, finding the correct sigma is critical in order to establish the Gaussian
convolution model. Take the block shown in Figure 6.5 as an example: the simplest way
to find the sigma is that for each sigma value we establish a score measuring the degree
of similarity between the experiment (Figure 6.5c) and the convolved results, and then we
choose the sigma which gives the highest score among all. However in order to do so we
need to overcome two problems: 1) All the convolved results are continuous while the
Figure 6.5c is binary and noisy; 2) The value of the density map is scaled between 0 and
1, while there are blocks whose densities are below 1 but not overfilled at all and blocks
whose densities are above 0 but entirely overfilled.
Averaging several SEM images of the same situation together could solve the first
problem. We experimented with a 300 mm CDU wafer and chose 25 dies with the same
template CD. They shared the same distribution of overfilled templates and suffered from
the similar noise pattern. The averaged result is no longer a binary matrix but a gray-scale
overfill probability map (Figure 6.5d&e). A pixel of intensity 0.04 means that the
template at this location is not overfilled in 1 out of 25 cases. Using this method we can
significantly reduce the noises inside the block and achieve higher model prediction
accuracy.

60

To solve the second problem we need to realize that the convolved density map needs to
be appropriately scaled in order to match the overfill probability map and there are two
related thresholds. When the template density is above the upper threshold, the template
wont become overfilled. On the other hand, if the template density is below the lower
threshold, the template will definitely become overfilled. And if the template density falls
between these two thresholds, this template has a certain probability of becoming
overfilled. We assume there is a linear relationship between the convolved density value
and the overfilled probability. Figure 6.7 describes the procedures to find these two
thresholds and the corresponding scaling function. As it would be difficult to directly
extract the thresholds from the convolved result, we choose the 8% and 92% percentile
band and use linear regression to determine the threshold (Figure 6.7b&c). We extracted
the convolved density value of these two levels by overlapping the 8% and 92%
percentile band with the before-scaled convolved result (Figure 6.7e&f).
Probability(of(not(overlled(
(a)(

(b)(

(g)(

(c)(

1(

92%(

(d)(

(e)(

(f)(

8%(
0(
Convolved(density(value(

Figure 6.7 (a) Overfill probability map. (b) 8% percentile band of not overfilled
templates. White pixel means that the template at this location has an 8% chance of not
overfilled. (c) 92% percentile band of not overfilled templates. White pixel means that
the template at this location has a 92% chance of not overfilled. (d) The convolved result
before scaling. (e) The convolved result overlapped with 8% percentile band. The band is
colored in red. (f) The convolved result overlapped with 92% percentile band. The band
is colored in blue. (g) The scaling function. The convolved density value of the red circle

61

is the average convolved density value of the red band in (e). The convolved density
value of the blue circle is the average convolved density value of the blue band in (f).

After we have determined the scaling function, we scale the convolved results of different
sigma values and calculate the score by measuring the similarity between the scaled
results and the overfill probability map (Figure 6.8). The score equation is shown in
Figure 6.8e, with lower score indicating a better match. From Figure 6.8e we observed
that sigma of 0.7 resulted in the best match. Therefore, we found the optimal sigma.
Figure 6.9b shows another 9 cases (including the one in Figure 6.8) that also chose sigma
as 0.7, and they all matched well with the experimental results (Figure 6.9a).

1(b)

(a)

(a)$

(b)$

(d)

(c)

(c)$

(e)
(d)$

Sigma&=&0.2

r
(e)$

Sigma&=&0.8

Sigma&=&1.4

0.5$

1.0$ Sigma$ 1.5$

Sigma&=&2.0

Sigma&=&2.6

2.0$

2.5$

Figure 6.8 (a) Convolved result before scaling. (b) Scaled convolved result. (c) Overfill
probability map. (d) Scaled convolved results of different sigma values. It is obvious that
sigma = 0.8 matches the best with (c). (e) Sigma-score curve. The definition of the score

62

is shown above the curve. At the best case, theres ~10% difference between scaled
density and overfill probability. This difference may come from the statistical fluctuation
of thermodynamics as well as the systematic template CD variation.

(a)$

(b)$

Figure 6.9 (a) The overfill probability map retrieved from SEM images of 9 cases. (b)
The convolved density map using sigma = 0.7.

6.3.3 The Impact of Annealing Time

As we mentioned above, the sigma value is dependent on the DSA process parameters.
Here we studied the influence of annealing time on the sigma value and tried three
annealing time: 1 min, 5 min, and 30 min. The standard annealing time we used in the
section above is 5 minutes. The low-magnification SEM images of three annealing
condition is shown in Figure 6.10, and it is clearly shown that longer annealing time leads
to much more overfilled defects. Using the method in section 3.2, we find the sigma
values for 1 min and 30 min annealing are 0.3 and 2.5 respectively (Figure 6.11). This
also meets the expectation we received from Figure 6.10. However, it is important to note
that in the last case of Figure 6.11, the convolved result does not match so well with the
experiments. The possible reasons are: 1) The overfill probability is close to zero
everywhere, which makes it not very useful in terms of searching the optimal sigma value;
2) As the sigma increases to 2.5 (or maybe even larger), not only does our own layout is
influencing the DSA patterns, other neighboring layouts will have an impact too.

63

However, the information of other neighboring layouts is unknown, therefore rendering

the prediction not accurate enough.
c

70'nm'

f'

60'nm'

1'min'

5'min'

30'min'

Figure 6.10 Low-magnitude SEM images. (a) Template CD 70 nm, anneal 1 min. (b)
Template CD 70 nm, anneal 5 min. (a) Template CD 70 nm, anneal 30 min. (a) Template
CD 60 nm, anneal 1 min. (a) Template CD 60 nm, anneal 5 min. (a) Template CD 60 nm,
anneal 30 min.

64

1"min"

5"min"

30"min"

(a)"

(b)"

Figure 6.11 The comparison between experimental results and convolved density map.
Annealing time is 1 minute for the left two columns. Annealing time is 5 minutes for the
middle two columns. Annealing time is 30 minutes for the right two columns. (a)
Experimental results. (b) Convolved density map.

6.4 Sub DSA resolution Assist Features (SDRAF)

6.4.1 Introduction
Sub-DSA resolution Assist Features (SDRAF) are small openings in the template layer
that are designed to artificially create a more uniform template density, leading to a
reduced number of overfilled templates and therefore improved DSA performance.
Unlike the Sub-Resolution Assist Features (SRAF) for optical lithography that could not
be printed on the wafer, SDRAFs need to be printed to be functional. There are two
advantages of SDRAF: First, allowing these small features to be printed can increase the
process window of lithography [62]. Second, when placing these SDRAFs around the
low-density guiding templates, the SDRAFs become the polymer reservoirs to store the
extra polymers, preventing the low-density guiding templates from becoming overfilled,
thereby improving the quality of DSA and reduce defects. However, it is very important
to note that the size of the SDRAFs need to be controlled within a certain range to make
sure that no etch-transferrable DSA patterns are generated inside the SDRAF. On the
other hand, the size of the SDRAFs also cannot be too small such that they cannot store

65

enough polymers to reduce the overfilled conditions.

6.4.2 Test Guiding Pattern Design

The basic test patterns of guiding templates are shown in Figure 6.11a. The rectangles
represent the target templates that generate useful DSA patterns, while the squares
represent the SDRAFs. This pattern is placed periodically in an 8 m 8 m test block.
We designed a series of these patterns with various sizes of the rectangles as well as the
size of the squares, and we printed them on a FEM wafer. Due to the different exposure
doses and focuses of each die, we could find two dies with the rectangles of the same size,
and the squares on one of them were not resolved (Figure 6.12c&d). In this way we could
compare these two cases and find out if the SDRAFs really help to eliminate DSA defects
caused by overfilled conditions (Figure 6.12).

No%SDRAFs%

With%SDRAFs%
(a)%

(b)%

(c)%

(d)%

300#nm#

300#nm#

Figure 6.12 (a) The layout design of test pattern. Small squares are SDRAFs and
rectangles are targeted DSA templates. (b) The cases that the SDRAFs were not resolved
due to overexposure. (c) The SEM picture of (a), the CD of SDRAFs is 39 nm. (d) The
SEM picture of (b). For target templates, the size is 53 nm by 81 nm (minor and major
axis lengths) in both cases.

66

6.4.3 Experiment Results

Figure 6.13 shows the comparison between DSA performance with SDRAFs and without
SDRAFs. It is observed that under the zoom-out view, the not-overfilled templates have
much higher contrast around the template edge and could be identified clearly, while the
overfilled templates look like they disappeared (Figure 6.13c&d). The comparison
clearly shows that SDRAFs help prevent the overfilled conditions, especially at the
corner of the design pattern, where the template density is the lowest. In Figure 6.13h,
some of the SDRAFs are overfilled at the corner as well. After adopting SDRAFs, the
number of defects caused by overfilled condition is dropped from ~ 40 to zero. However,
as the SDRAFs size continues to increase, DSA holes start to appear inside the SDRAFs,
which are counted as failure as these extra holes may be transferred to the bottom layer
and result in unnecessary contacts (Figure 6.14).
a

Figure 6.13 Comparison between DSA performance with SDRAFs and without SDRAFs.
The top row (a) (d) is the case without SDRAFs, the bottom row (e) (h) is the case
with SDRAFs. (a) SEM image of guiding template without SDRAF. The large ovals are
the target DSA templates, with a major axis of 82 nm and a minor axis of 53 nm. (b) The
DSA result of (a). 1-hole DSA patterns are generated inside the target templates. (c) The
low magnification image of the block. This provides a zoom-out view for us to inspect
the location of overfilled templates. Note that (b) is a zoom-in view of the blue box
shown in the center. It is also important to note that there are some templates at the corner

67

of the block that appear missing but is actually a result of polymers overfilling,
rendering the template less visible under the SEM (see zoom-in view in (d)). (d) A
zoom-in view of the blue box shown in the left top corner of (c). Red circles label the
templates that are overfilled due to low density. (e) SEM image of guiding template with
SDRAFs. The small circular ones are the SDRAFs with a size of 40 nm. The target ovals
have a major axis of 81 nm and a minor axis of 53 nm. (f) The DSA result of (d). 1-hole
DSA patterns are generated inside the target templates while nothing inside the SDRAFs.
(g) The low magnification image of the block. (f) is a zoom-in view of the blue box
shown in the center. (h) A zoom-in view of the blue box shown in the left top corner of
(g). No oval DSA template is overfilled and they all have 1-hole DSA patterns inside.
(a)$

(b)$

(c)$

Figure 6.14 An example of failed SDRAFs. (a) SEM image of guiding templates, the CD
of SDRAF is 55.2 nm. (b) SEM image of DSA result. Inside the red boxes are the 1-hole
DSA patterns generated inside the SDRAFs. (c) The low magnification SEM image of
the block.

6.4.4 Overfill Rate and SDRAF Sizes

In the previous section we have shown that SDRAFs could help prevent overfilled
conditions and reduce DSA defects. As shown in Figure 6.13c&g, each test pattern
design corresponded to an 8 8 m matrix block on the wafer. It is observed that under
the zoom-out view, the not-overfilled templates have much higher contrast around the
template edge and could be seen clearly, while the overfilled templates look like they
disappeared (Figure 6.13c&d). Therefore, from the low-magnitude zoom-out SEM
image we could easily extract the number of overfilled templates in the whole block. In
order to further evaluate the effectiveness of different SDRAFs, we introduce the concept

68

of the overfill rate (OR), which is defined as the percentage of overfilled templates in the
whole block. After adopting SDRAFs, the OR will decrease due to limited occurrence of
overfill. And the effect of SDRAFs is measured as the reduction of OR. For example, in
the case shown above (Figure 6.13), the OR in Figure 6.12c is 4%, which means in this
case, 4% of the oval templates were overfilled. Similarly, the OR in Figure 6.13g is 0%,
which means no oval templates were overfilled. Thus we can conclude that the SDRAFs
caused an OR drop of 4%.
Then we continue to explore the effects of SDRAFs based on their sizes. Intuitively we
would assume that larger SDRAFs would lead to more reduction in OR. However it is
important to note that if the OR w/o SDRAFs were already very close to 0, the effect of
the difference between small and large SDRAFs would not be obvious since the OR
could easily be pulled to zero. Therefore we picked the case where the OR w/o SDRAFs
was larger than 2% so that we could observe the influence brought by SDRAF size.
In Figure 6.15 we showed the relationship between SDRAF sizes and the corresponding
OR after using SDRAFs. The ones without SDRAFs were labeled. It is clearly shown that
larger SDRAFs will cause larger OR drop, and the effect would saturate when the OR
reached 0 after a certain SDRAF size.

Overll$Rate$(OR)$

5.0%$
Oval$size:$75$nm$x$57$nm$

4.0%$

Oval$size:$80$nm$x$60$nm$

3.0%$
2.0%$
1.0%$
0.0%$
0$

10$

20$
30$
SDRAF$CD$(nm)$

40$

50$

Figure 6.15 The relationship between SDRAF sizes and overfill rate (OR). Overfill rate is
defined as the percentage of overfilled templates in the whole block. The two stars on the
left marked the OR when there is no SDRAF. Blue data points represent the case where

69

the target ovals have a major axis of 75 nm and a minor axis of 57 nm. Orange data
points represent the case where the target ovals have a major axis of 80 nm and a minor
axis of 60 nm. Note that although the some of SDRAF CDs shown above reached below
30 nm, the actually feature sizes on the wafer suffer from large variation, which explains
some of the noises in this trend.

6.4.5 Alternative SDRAF Design

Here we adopted a SDRAF design that the shape is very similar with the traditional
SRAF design in optical lithography [63]. The experimental procedures were the same as
section 6.2. It is also worthy to note that some of the SDRAFs were overfilled that they
look like disappeared in both zoom-in and zoom-out views. In experiments we tried
different template pitch with different sizes of SDRAFs, the results are shown in Figure
6.16 and Figure 6.17. This set of results prove that the applicability of SDRAFs using
different designs.

a"

c"

500"nm"

500"nm"

500"nm"

5"um"

f"

500"nm"

5"um"

Figure 6.16 (a) SEM image of guiding templates without SDRAF. The template CD is 74
nm, and the template pitch is 380 nm. (b) SEM image of DSA results in (a). Some of the
templates were overfilled and DSA defects were formed. (c) Low magnitude SEM image
of the entire contact matrix of (b). The OR is higher than 30%. (d) SEM image of guiding

70

templates with SDRAFs. The template CD and pitch are same as (a). The SDRAF size is
41 x 137 nm. (e) SEM image of DSA results in (d). No templates were overfilled. (f) Low
magnitude SEM image of the entire contact matrix of (e). The OR is 0%.

a"

c"

500"nm"

500"nm"

500"nm"

5"um"

f"

500"nm"

5"um"

Figure 6.17 (a) SEM image of guiding templates without SDRAF. The template CD is 72
nm, and the template pitch is 600 nm. (b) SEM image of DSA results in (a). Some of the
templates were overfilled and DSA defects were formed. (c) Low magnitude SEM image
of the entire contact matrix of (b). The OR is higher than 50%. (d) SEM image of guiding
templates with SDRAFs. The template CD and pitch are same as (a). The SDRAF size is
38 x 86 nm. (e) SEM image of DSA results in (d). No templates were overfilled. (f) Low
magnitude SEM image of the entire contact matrix of (e). The OR is 0.03%, means that
only two templates are overfilled.

71

Chapter 7

Conclusions
7.1 Summary of Contributions
There are three fundamental contributions established by our work and discussed in this
thesis: 1. We identify that the aperiodic DSA patterns are critical for contact hole
patterning in integrated circuits, and that small physical guiding templates can provide
the flexible control of placement required for aperiodic DSA patterns. 2. We introduce
the concept of DSA alphabet and establish the design rules for DSA contact hole
patterning. 3. We experimentally show that sub-DSA-resolution assist features can
effectively reduce defects due to uneven template densities across the wafer.
Realizing that contact hole patterning does not require long range order, first we adopted
small physical guiding templates to flexibly control and generate aperiodic DSA patterns.
Both experiments and simulations were carried out to verify the applicability of small
physical guiding templates. Then we realized that there exists a limited set of guiding
templates that can cover all the possible contact patterns of a full chip contact layer,
which we named as the alphabet of DSA. Based on the DSA alphabet concept, we
discovered a general and scalable template design strategy that links the DSA material
properties to the technology node requirements. Three different design solutions are
presented and experimentally demonstrated for using PS-b-PMMA in DSA contact
patterning for 14, 11 and 7 nm node one-bit half adders. This strategy is general and can
be applied to technology nodes beyond the experimentally demonstrated 7 nm node,
provided the block copolymer used meets the necessary size and pitch requirements. Last
but not the least, we extensively studied the DSA Interaction Range and built a Gaussian
Convolution model to describe the template density influence on DSA process. For the
first time, we experimentally proved that Sub-DSA Resolution Assist Features (SDRAFs)
72

were effective in mitigating the template density influence and reduce the overfilled
defects. Our work provides a viable path for low-cost DSA technology to extend optical
lithography beyond its current limits.

7.2 Future Work

In the face of rapid device scaling, challenges such as optimizing and tuning the template
design based on overlay, defectivity, and lithography requirements will need to be further
investigated in order for practical implementation of DSA in industry. Our DSA guiding
template design strategy points to an important research direction: IC layouts can be
further optimized to improve the yield and accuracy of DSA process. Topics including
the minimization and the optimization of the alphabet set [45, 64], DSA hotspot detection
[65], DSA-aware routing strategy [66], as well as the combination of lithography
simulation and template design [58] are receiving attention from researchers and waiting
to be explored thoroughly.
Most of the research up-to-date has been focusing on the planar DSA patterns and
achieved great success in either controlling the patterns or their applications. In 10 or 15
years, after we reach the ultimate device scaling limits, non-traditional devices based on
new physical phenomena will become the lead and this requires extended patterning
capabilities that are beyond the limits of conventional optical lithography. The more
complex pattern generation, such as three-dimensional self-assembled structures may
bring new opportunities and even revolutionize the field of nanofabrication, since 3D
structures do not require layer-by-layer process steps like traditional planar process.
There exist many possibilities, such as fabricating 3D devices like cross-point memory
arrays in a single step, rather than building up the structure sequentially.

73

(a)

Self-assembly

Selective etch

Metal deposition

Polymer etch

(b)

Figure 7.1 (a) Three-dimensional self-assembled nanostructures [67]. (b) SEM of 3D

bilayer nanostructures generated by functionalized posts [68].

Researchers have explored 3D nanostructures self-assembled by either sequential

stacking of block copolymer films or functionalized resist posts as the templates (Figure
7.1). However, there have been few reports about how to utilize these complex
nanostructures, the difficulties of which come into many perspectives. First, it is very
challenging to transfer the three dimensional structures to other materials or substrates.
The conventional methods of pattern transfer rely mostly on dry or wet etching, which
works mostly for the two-dimensional patterns. In order to use the three-dimensional
structures, it is extremely important to be able to fully preserve the structures. Secondly,
researchers are still exploring how to generate, control and align these 3D self-assembled
structures in an accurate and flexible fashion. It requires additional efforts to match these
structures with specific device structures and dimensions. Therefore, it would become an
interesting and promising scientific research field to explore how to achieve flexible and
accurate control of 3D self-assembled structures, with the capability to transfer the
structures and fabricate emerging devices that are beyond todays imagination.

74

Publications
[1]

M. C. Tung, H. Yi, H.-S. P. Wong, Optimization of Peanut-Shaped Template

Geometry for Block Copolymer Directed Self-Assembly, 59th International
Conference on Electron, Ion, and Photon Beam Technology and Nanofabrication
(EIPBN), paper P04-07, San Diego, CA, May 26 May 29, 2015

[2]

Z. Xiao, D. Guo, M.D.F. Wong, H. Yi, M.C. Tung, H.-S. P. Wong, Layout
Optimization and Template Pattern Verification for Directed Self-Assembly
(DSA), Design Automation Conference (DAC), San Francisco, June 7 11, 2015.
Invited paper in Special Session 75, Design for Manufacturability for Sub-10nm
Technologies: Challenges and Solutions

[3]

H.-S. P. Wong, H. Yi, M.C. Tung, K. Okabe, Physical Layout Design of

Directed Self-assembly Guiding Alphabet for IC Contact Hole/via Patterning,
invited paper, 2015 ACM International Symposium on Physical Design (ISPD), pp.
65 66, Monterey, California, March 29-April 1, 2015.

[4]

H. Yi, X.-Y. Bao, R. Tiberio, and H.-S. P. Wong, A General Design Strategy for
Block Copolymer Directed Self-Assembly Patterning Integrated Circuits Contact
Holes using an Alphabet Approach, Nano Letters, 15 (2), pp 805812, 2015.

[5]

H. Yi, J. Bekaert, R. Gronheid, G. Fenger, K. Nafus, H.-S. P. Wong, Study of

DSA Interaction Range using Gaussian Convolution, SPIE Advanced
Lithography, 9423, 94232A, 2015.

[6]

H. Yi, J. Bekaert, R. Gronheid, G. Vandenberghe, K. Nafus, H.-S. P. Wong,

Experimental Study of Sub DSA resolution Assist Features (SDRAF), SPIE
Advanced Lithography, 9423, 94231F, 2015.

[7]

Z. Xiao, Y. Du, M.D.F Wong, H. Yi, H.-S. P. Wong, H. Zhang, "Contact pitch
and location prediction for Directed Self-Assembly template verification," Design
Automation Conference (ASP-DAC), January 19 22, 2015.

[8]

Z. Xiao, Y. Du, H. Tian, M.D.F Wong, H. Yi, H.-S. P. Wong, H. Zhang,

"Directed Self-Assembly (DSA) Template Pattern Verification," Design
Automation Conference (DAC), June 1 5, San Francisco, 2014.
75

[9]

S. Wuister, T. Druzhinina, D. Ambesi, B. Laenens, H. Yi, J. Finders, Influence

of litho patterning on DSA placement errors, SPIE Advanced Lithography, 9049,
2014.

[10]

Y. Du, Z. Xiao, M.D.F. Wong, H. Yi, H.-S. P. Wong, Via layer DSA hot spot
removal through local rerouting, SPIE Advanced Lithography, 9049, 2014.

[11]

Z. Xiao, Y. Du, H. Tian, M.D.F Wong, H. Yi, H.-S. P. Wong, "DSA template
optimization for contact layer in 1D standard cell design," SPIE Advanced
Lithography, 9049, 2014.

[12]

Y. Wu, H. Yi, Z. Zhang, Z. Jiang, J. Sohn, S. Wong, H.-S. P. Wong, First

demonstration of RRAM patterned by block copolymer self-assembly,
International Electron Devices Meeting (IEDM), pp.20.8.1,20.8.4, 9-11, 2013.

[13]

H. Yi, Y. Wu, Z. Zhang, H.-Y. Chen, S. Yu, H.-S. P. Wong, Metal Oxide
Resistive Switching Memory (RRAM): Devices, Fabrication, and Self-Assembly
Patterning for Random Logic and Memory Devices (SRAM, NAND, RRAM),
invited plenary talk, 26th International Microprocesses and Nanotechnology
Conference, Hokkaido, Japan, November 5 8, 2013.

[14]

Y. Du, D. Guo, M.D.F. Wong, H. Yi, H. -S.P. Wong, H. Zhang, Q. Ma, Block
copolymer directed self-assembly (DSA) aware contact layer optimization for 10
nm 1D standard cell library, Computer-Aided Design (ICCAD), pp.186, 193,
18-21, 2013.

[15]

H. Yi, H.-S. P. Wong, "Alphabet-Based Template Design Rules - A Key Enabler

for a Manufacturable DSA Technology," invited paper, AVS 60th International
Symposium & Exhibition (AVS), 2013.

[16]

H. Yi, H.-S. P. Wong, Directed Self-Assembly for Random Logic Circuits An

Opportunity for Design Tool Development, keynote talk, China Semiconductor
Technology

International

Conference

(CSTIC),

Symposium

I:

Device

Engineering and Technology, 2013.

[17]

H. Yi, X.-Y. Bao, R. Tiberio, and H.-S. P. Wong, Design Strategy of Small
Topographical Guiding Templates for sub-15 nm Integrated Circuits Contact Hole
Patterns using Block Copolymer Directed Self-Assembly, SPIE Advanced
Lithography, 8680, 868010, 2013.

76

[18]

H. Yi, A. Latypov and H.-S. P. Wong, Computational Simulation of Block

Copolymer Directed Self-Assembly in Small Topographical Guiding Templates,
SPIE Advanced Lithography, 8680, 86801L, 2013.

[19]

A. Latypov, M. Preil, G. Schmid, J. Xu, H. Yi, K, Yoshimoto and Y. Zou,

Exploration of the directed self-assembly based nano-fabrication design space
using computational simulations, SPIE Advanced Lithography, 868013, 2013.

[20]

H. Yi and H.-S. P. Wong, Block Copolymer Directed Self-Assembly Two-Hole

Pattern inside Peanut-Shaped Templates, The International Conference on
Electron, Ion, Photon Beam Technology and Nanofabrication (EIPBN), oral
presentation, 2013.

[21]

H. Yi, X.-Y. Bao, J. Zhang, C. Bencher, L.-W. Chang, X. Chen, R. Tiberio, J.

Conway, H. Dai, Y. Chen, S. Mitra and H.-S. P. Wong, Flexible Control of
Block Copolymer Directed Self-Assembly using Small Topographical Templates:
Potential Lithography Solution for Integrated Circuits Contact Hole Patterning,
Adv. Mater., vol. 24, issue 23, pp. 3107 3114, 2012.

[22]

H. Yi, X.-Y. Bao, J. Zhang, R. Tiberio, J. Conway, L.-W. Chang, S. Mitra and
H.-S. P. Wong, Contact Hole Patterning for Random Logic Circuits using Block
Copolymer Directed Self-Assembly, SPIE Advanced Lithography, 8323,
83230W, 2012.

[23]

C. Bencher, H. Yi, J. Zhou, M. Cai, H. Dai, J. Smith, J. Cheng, D. Sanders, M.

Tijo and S. Holmes ,Directed Self-Assembly Defectivity Assessment, SPIE
Advanced Lithography, 8323, 83230N, 2012.

[24]

H.-S. P. Wong, C. Bencher, H. Yi, X.-Y. Bao and L.-W. Chang, Block
Copolymer Directed Self-Assembly Enables Sub-lithographic Patterning for
Device Fabrication, SPIE Advanced Lithography, 8323, 832303, 2012.

[25]

H. Yi, X.-Y. Bao, C. Bencher, H. Dai, Y. Chen and H.-S. P. Wong, Design
Space for One-hole Pattern using Block Copolymer Directed Self-Assembly, The
International Conference on Electron, Ion, Photon Beam Technology and
Nanofabrication (EIPBN), oral presentation, 2012.

[26]

H.-S. P. Wong, C. Bencher, H. Yi, X.-Y. Bao and L.-W. Chang, Directed
Self-Assembly for the Semiconductor Industry, The International Conference on

77

Electron, Ion, Photon Beam Technology and Nanofabrication (EIPBN), invited

oral presentation, 2012.
[27]

X.-Y. Bao, H. Yi, C. Bencher, L.-W. Chang, H. Dai, Y. Chen, P.-T. J. Chen and
H.-S. P. Wong, SRAM, NAND, DRAM Contact Hole Patterning using Block
Copolymer Directed Self-assembly Guided by Small Topographical Templates,
International Electron Devices Meeting (IEDM), pp. 167-170, 2011.

[28]

C. Bencher, H. Dai, L. Miao, Y. Chen, P. Xu, Y. Chen, S. Oemardini, J. Sweis, V.

Wiaux, J. Hermans, L.-W.Chang, X. Bao, H. Yi, H.-S. P. Wong, Mandrel Based
Patterning: Density Multiplication Techniques for 15 nm Nodes, SPIE Advanced
Lithography, 7973, 79730K, 2011.

78

Reference
[1]

R. R. Schaller, "Moore's law: past, present and future," Spectrum, IEEE, vol. 34,
pp. 52-59, 1997.

[2]

C. T. Black, R. Ruiz, G. Breyta, J. Y. Cheng, M. E. Colburn, K. W. Guarini, et al.,

"Polymer self assembly in semiconductor microelectronics," IBM J. Res. Dev.,
vol. 51, pp. 605-633, 2007.

[3]

X.-Y. Bao, H. Yi, C. Bencher, L.-W. Chang, H. Dai, Y. Chen, et al., "SRAM,
NAND, DRAM contact hole patterning using block copolymer directed
self-assembly guided by small topographical templates," in Electron Devices
Meeting (IEDM), 2011 IEEE International, 2011, pp. 7.7. 1-7.7. 4.

[4]

I. Bita, J. K. Yang, Y. S. Jung, C. A. Ross, E. L. Thomas, and K. K. Berggren,

"Graphoepitaxy of self-assembled block copolymers on two-dimensional periodic
patterned templates," Science, vol. 321, pp. 939-943, 2008.

[5]

R. Ruiz, H. Kang, F. A. Detcheverry, E. Dobisz, D. S. Kercher, T. R. Albrecht, et

al., "Density multiplication and improved lithography by directed block
copolymer assembly," Science, vol. 321, pp. 936-939, 2008.

[6]

R. A. Segalman, H. Yokoyama, and E. J. Kramer, "Graphoepitaxy of spherical

domain block copolymer films," Adv. Mater., vol. 13, pp. 1152-1155, 2001.

[7]

I. W. Hamley, The physics of block copolymers vol. 329: Oxford University Press
New York, 1998.

[8]

E. L. Thomas, R. L. Lescanec, F. C. Frank, J. S. Higgins, A. Klug, and I. W.

Hamley, Phase Morphology in Block Copolymer Systems [and Discussion] vol.
348, 1994.

[9]

F. S. Bates and G. H. Fredrickson, "Block Copolymer Thermodynamics: Theory

and Experiment," Annual Review of Physical Chemistry, vol. 41, pp. 525-557,
1990.

[10]

F. S. Bates and G. H. Fredrickson, "Block copolymersdesigner soft materials,"

Physics today, vol. 52, pp. 32-38, 2008.

[11]

L. Rockford, Y. Liu, P. Mansky, T. P. Russell, M. Yoon, and S. G. J. Mochrie,

"Polymers on nanoperiodic, heterogeneous surfaces," Physical Review Letters,

79

vol. 82, pp. 2602-2605, 1999.

[12]

R. A. Segalman, H. Yokoyama, and E. J. Kramer, "Graphoepitaxy of Spherical

Domain Block Copolymer Films," Advanced Materials, vol. 13, pp. 1152-1155,
2001.

[13]

J. Y. Cheng, A. M. Mayes, and C. A. Ross, "Nanostructure engineering by

templated self-assembly of block copolymers," Nature Materials, vol. 3, pp.
823-828, 2004.

[14]

M. P. Stoykovich, M. Muller, S. O. Kim, H. H. Solak, E. W. Edwards, J. J. de

Pablo, et al., "Directed assembly of block copolymer blends into nonregular
device-oriented structures," Science, vol. 308, pp. 1442-1446, Jun 2005.

[15]

I. Bita, J. K. W. Yang, Y. S. Jung, C. A. Ross, E. L. Thomas, and K. K. Berggren,

"Graphoepitaxy of Self-Assembled Block Copolymers on Two-Dimensional
Periodic Patterned Templates," Science, vol. 321, pp. 939-943, August 15, 2008
2008.

[16]

M. P. Stoykovich and P. F. Nealey, "Block copolymers and conventional

lithography," Materials Today, vol. 9, pp. 20-29, 2006.

[17]

S. O. Kim, H. H. Solak, M. P. Stoykovich, N. J. Ferrier, J. J. de Pablo, and P. F.

Nealey, "Epitaxial self-assembly of block copolymers on lithographically defined
nanopatterned substrates," Nature, vol. 424, pp. 411-414, Jul 2003.

[18]

J. Y. Cheng, C. A. Ross, E. L. Thomas, H. I. Smith, and G. J. Vancso, "Templated

Self-Assembly of Block Copolymers: Effect of Substrate Topography," Advanced
Materials, vol. 15, pp. 1599-1602, 2003.

[19]

F. A. Detcheverry, P. F. Nealey, and J. J. de Pablo, "Directed Assembly of a

Cylinder-Forming Diblock Copolymer: Topographic and Chemical Patterns,"
Macromolecules, vol. 43, pp. 6495-6504, 2010.

[20]

X.-Y. Bao, H. Yi, C. Bencher, L.-W. Chang, H. Dai, Y. Chen, et al., "SRAM,
NAND, DRAM contact hole patterning using block copolymer directed
self-assembly guided by small topographical templates," in Electron Devices
Meeting (IEDM), 2011 IEEE International, 2011, pp. 7.7. 1-7.7. 4.

[21]

C. Li-Wen, B. Xinyu, C. Bencher, and H. S. P. Wong, "Experimental

demonstration of aperiodic patterns of directed self-assembly by block copolymer

80

lithography for random logic circuit layout," in Electron Devices Meeting (IEDM),
2010 IEEE International, 2010, pp. 33.2.1-33.2.4.
[22]

R. Gronheid, J. Bekaert, V.-K. Murugesan Kuppuswamy, N. Vandenbroeck, J.

Doise, Y. Cao, et al., "Process optimization of templated DSA flows," 2014, pp.
90510I-90510I-7.

[23]

C. Tang, E. M. Lennon, G. H. Fredrickson, E. J. Kramer, and C. J. Hawker,

"Evolution of block copolymer lithography to highly ordered square arrays,"
Science, vol. 322, pp. 429-432, 2008.

[24]

J. Y. Cheng, C. T. Rettner, D. P. Sanders, H. C. Kim, and W. D. Hinsberg, "Dense

SelfAssembly on Sparse Chemical Patterns: Rectifying and Multiplying
Lithographic Patterns Using Block Copolymers," Adv. Mater., vol. 20, pp.
3155-3158, 2008.

[25]

S.-M. Park, O.-H. Park, J. Y. Cheng, C. T. Rettner, and H.-C. Kim, "Patterning
sub-10 nm line patterns from a block copolymer hybrid," Nanotechnology, vol. 19,
p. 455304, 2008.

[26]

M. P. Stoykovich, M. Mller, S. O. Kim, H. H. Solak, E. W. Edwards, J. J. De

Pablo, et al., "Directed assembly of block copolymer blends into nonregular
device-oriented structures," Science, vol. 308, pp. 1442-1446, 2005.

[27]

P. A. Rincon Delgadillo, R. Gronheid, C. J. Thode, H. Wu, Y. Cao, M. Somervell,

et al., "All track directed self-assembly of block copolymers: process flow and
origin of defects," 2012, pp. 83230D-83230D-9.

[28]

P. A. R. Delgadillo, R. Gronheid, C. J. Thode, H. Wu, Y. Cao, M. Neisser, et al.,

"Implementation of a chemo-epitaxy flow for directed self-assembly on 300-mm
wafer processing equipment," Journal of Micro/Nanolithography, MEMS, and
MOEMS, vol. 11, pp. 031302-1-031302-5, 2012.

[29]

P. Rincon Delgadillo, R. Harukawa, M. Suri, S. Durant, A. Cross, V. R.

Nagaswami, et al., "Defect source analysis of directed self-assembly process
(DSA of DSA)," 2013, pp. 86800L-86800L-9.

[30]

(September 9). International Technology Roadmap for Semiconductors. Available:

http://public.itrs.net

[31]

C. Bencher, H. Yi, J. Zhou, M. Cai, J. Smith, L. Miao, et al., "Directed

81

self-assembly defectivity assessment (Part II)," in SPIE Advanced Lithography,

2012, pp. 83230N-83230N-10.
[32]

M. M. Shulaker, J. Van Rethy, T. F. Wu, L. Suriyasena Liyanage, H. Wei, Z. Li,

et al., "Carbon Nanotube Circuit Integration up to Sub-20 nm Channel Lengths,"
ACS Nano, vol. 8, pp. 3434-3443, 2014/04/22 2014.

[33]

B. Wu and A. Kumar, "Extreme ultraviolet lithography: a review," J. Vac. Sci.

Technol. B, vol. 25, pp. 1743-1761, 2007.

[34]

V. R. Manfrinato, L. Zhang, D. Su, H. Duan, R. G. Hobbs, E. A. Stach, et al.,

"Resolution limits of electron-beam lithography toward the atomic scale," Nano
Lett., vol. 13, pp. 1555-1558, 2013.

[35]

R. Pease, "Electron beam lithography," Contemporary Physics, vol. 22, pp.

265-290, 1981.

[36]

C.-s. Koay, S. Halle, S. Holmes, K. Petrillo, M. Colburn, Y. Van Dommelen, et

al., "Towards manufacturing of advanced logic devices by double-patterning," in
SPIE Advanced Lithography, 2011, pp. 79730F-79730F-18.

[37]

E. Han, T. R. Younkin, M. Chandhok, A. M. Myers, T. A. Tronic, F. Gstrein, et

al., "Material readiness for generation 2 directed self-assembly (DSA) < 24nm
pitch," 2015, pp. 94250O-94250O-6.

[38]

T. Skotnicki, J. A. Hutchby, T.-J. King, H.-S. Wong, and F. Boeuf, "The end of
CMOS scaling: toward the introduction of new materials and structural changes to
improve MOSFET performance," Circuits and Devices Magazine, IEEE, vol. 21,
pp. 16-26, 2005.

[39]

S.-M. Hur, C. J. Garca-Cervera, E. J. Kramer, and G. H. Fredrickson, "SCFT

Simulations of Thin Film Blends of Block Copolymer and Homopolymer
Laterally Confined in a Square Well," Macromolecules, vol. 42, pp. 5861-5872,
2009/08/11 2009.

[40]

P. Mansky, Y. Liu, E. Huang, T. P. Russell, and C. Hawker, "Controlling

Polymer-Surface Interactions with Random Copolymer Brushes," Science, vol.
275, pp. 1458-1460, March 7, 1997 1997.

[41]

B. S. Haran, A. Kumar, L. Adam, J. Chang, V. Basker, S. Kanakasabapathy, et al.,

"22 nm technology compatible fully functional 0.1 um 6T-SRAM cell," in

82

Electron Devices Meeting, 2008. IEDM 2008. IEEE International, 2008, pp. 1-4.
[42]

H. Yi, X.-Y. Bao, C. Bencher, H. Dai, Y. Chen, and H.-S. P. Wong, "Design
Space for One-hole Pattern using Block Copolymer Directed Self-Assembly,"
presented at the 56th International Conference on Electron, Ion, and Photon Beam
Technology and Nanofabrication (EIPBN),, 2012.

[43]

H. Yi, X. Y. Bao, J. Zhang, C. Bencher, L. W. Chang, X. Chen, et al., "Flexible

Control of Block Copolymer Directed SelfAssembly using Small, Topographical
Templates: Potential Lithography Solution for Integrated Circuit Contact Hole
Patterning," Adv. Mater., vol. 24, pp. 3107-3114, 2012.

[44]

H. Yi and H.-S. P. Wong, "Block Copolymer Directed Self-Assembly Two-Hole

Pattern inside Peanut-Shaped Templates," presented at the 57th International
Conference on Electron, Ion, and Photon Beam Technology and Nanofabrication
(EIPBN), 2013.

[45]

Y. Du, D. Guo, M. D. Wong, H. Yi, H.-S. P. Wong, H. Zhang, et al., "Block

copolymer directed self-assembly (DSA) aware contact layer optimization for 10
nm 1D standard cell library," in Computer-Aided Design (ICCAD), 2013
IEEE/ACM International Conference on, 2013, pp. 186-193.

[46]

G. Fredrickson, The equilibrium theory of inhomogeneous polymers (international

series of monographs on physics): Oxford University Press, USA, 2013.

[47]

M. W. Matsen, "Thin films of block copolymer," The Journal of chemical physics,

vol. 106, pp. 7781-7791, 1997.

[48]

H. Takahashi, N. Laachi, K. T. Delaney, S.-M. Hur, C. J. Weinheimer, D.

Shykind, et al., "Defectivity in Laterally Confined Lamella-Forming Diblock
Copolymers: Thermodynamic and Kinetic Aspects," Macromolecules, vol. 45, pp.
6253-6265, 2012/08/14 2012.

[49]

D. C. Hector and H. F. Glenn, "Numerical Solution of Polymer Self-Consistent

Field Theory," Multiscale Modeling & Simulation, vol. 2, pp. 452-474, 2004.

[50]

H. Yi, X.-Y. Bao, R. Tiberio, and H.-S. P. Wong, "Design strategy of small
topographical guiding templates for sub-15nm integrated circuits contact hole
patterns using block copolymer directed self assembly," in SPIE Advanced
Lithography, 2013, pp. 868010-868010-9.

83

[51]

H. Yi, X.-Y. Bao, R. Tiberio, and H. S. P. Wong, "A General Design Strategy for
Block Copolymer Directed Self-Assembly Patterning of Integrated Circuits
Contact Holes using an Alphabet Approach," Nano Letters, 2014.

[52]

C. Bencher, H. Dai, and Y. Chen, "Gridded design rule scaling: taking the CPU
toward the 16nm node," in SPIE Advanced Lithography, 2009, pp.
72740G-72740G-10.

[53]

M. C. Smayling, C. Bencher, H. D. Chen, H. Dai, and M. P. Duane, "Apf

pitch-halving for 22nm logic cells using gridded design rules," in SPIE Advanced
Lithography, 2008, pp. 69251E-69251E-8.

[54]

J. Belledent, M. Smayling, J. Pradelles, P. Pimenta-Barros, S. Barnola, L. Mage,

et al., "Sub-20nm hybrid lithography using optical, pitch-division, and e-beam,"
in SPIE Advanced Lithography, 2012, pp. 83230F-83230F-10.

[55]

M. C. Smayling, R. J. Socha, and M. V. Dusa, "22nm logic lithography in the

presence of local interconnect," 2010, pp. 764019-764019-8.

[56]

K. Lai, C.-c. Liu, J. Pitera, D. J. Dechene, A. Schepis, J. Abdallah, et al.,

"Computational Aspects of Optical Lithography Extension by Directed
Self-Assembly," in Proc. SPIE, 2013, p. 868304.

[57]

D. Morris, K. Vaidyanathan, N. Lafferty, K. Lai, L. Liebmann, and L. Pileggi,

"Design of embedded memory and logic based on pattern constructs," in VLSI
Technology (VLSIT), 2011 Symposium on, 2011, pp. 104-105.

[58]

S. Wuister, T. Druzhinina, D. Ambesi, B. Laenens, L. H. Yi, and J. Finders,

"Influence of litho patterning on DSA placement errors," presented at the SPIE
Advanced Lithography, 2014.

[59]

H. Yi, X.-Y. Bao, J. Zhang, R. Tiberio, J. Conway, L.-W. Chang, et al.,

"Contact-hole patterning for random logic circuits using block copolymer directed
self-assembly," in SPIE Advanced Lithography, 2012, pp. 83230W-83230W-6.

[60]

H. Yi, A. Latypov, and H. S. P. Wong, "Computational simulation of block

copolymer directed self-assembly in small topographical guiding templates," 2013,
pp. 86801L-86801L-7.

[61]

G. F. Lorusso, F. Van Roey, E. Hendrickx, G. Fenger, M. Lam, C. Zuniga, et al.,

"Flare in extreme ultraviolet lithography: metrology, out-of-band radiation, fractal

84

point-spread

function,

and

flare

map

calibration,"

Journal

of

Micro/Nanolithography, MEMS, and MOEMS, vol. 8, pp. 041505-041505-6,

2009.
[62]

A. Latypov, T. H. Coskun, G. Garner, M. Preil, G. Schmid, J. Xu, et al.,

"Simulations of spatial DSA morphology, DSA-aware assist features and block
copolymer-homopolymer blends," 2014, pp. 904908-904908-10.

[63]

L. Pang, Y. Liu, and D. Abrams, "Inverse lithography technology (ILT): a natural

solution

for

model-based

SRAF

at

45nm

and

32nm,"

2007,

pp.

660739-660739-10.
[64]

Z. Xiao, Y. Du, M. D. F. Wong, and H. Zhang, "DSA template mask

determination and cut redistribution for advanced 1D gridded design," 2013, pp.
888017-888017-8.

[65]

Z. Xiao, Y. Du, H. Tian, M. D. F. Wong, H. Yi, H.-S. P. Wong, et al., "Directed

Self-Assembly (DSA) Template Pattern Verification," presented at the
Proceedings of the The 51st Annual Design Automation Conference on Design
Automation Conference, San Francisco, CA, USA, 2014.

[66]

Y. Du, Z. Xiao, M. D. F. Wong, H. Yi, and H. S. P. Wong, "DSA-aware detailed

routing for via layer optimization," 2014, pp. 90492J-90492J-8.

[67]

K. Hur, Y. Francescato, V. Giannini, S. A. Maier, R. G. Hennig, and U. Wiesner,

"Three-Dimensionally Isotropic Negative Refractive Index Materials from Block
Copolymer Self-Assembled Chiral Gyroid Networks," Angewandte Chemie, vol.
123, pp. 12191-12195, 2011.

[68]

A. Tavakkoli, K. Gotrik, A. Hannon, A. Alexander-Katz, C. Ross, and K.

Berggren, "Templating three-dimensional self-assembled structures in bilayer
block copolymer films," Science, vol. 336, pp. 1294-1298, 2012.

85