1
00:00:01,246 --> 00:00:04,830
Computers used to be as big as a room.

2
00:00:04,854 --> 00:00:06,446
But now they fit in your pocket,

3
00:00:06,470 --> 00:00:07,641
on your wrist

4
00:00:07,665 --> 00:00:10,984
and can even be implanted
inside of your body.

5
00:00:11,008 --> 00:00:12,289
How cool is that?

6
00:00:12,809 --> 00:00:17,146
And this has been enabled
by the miniaturization of transistors,

7
00:00:17,170 --> 00:00:19,662
which are the tiny switches
in the circuits

8
00:00:19,686 --> 00:00:21,462
at the heart of our computers.

9
00:00:22,051 --> 00:00:25,223
And it's been achieved
through decades of development

10
00:00:25,247 --> 00:00:28,045
and breakthroughs
in science and engineering

11
00:00:28,069 --> 00:00:30,741
and of billions of dollars of investment.

12
00:00:31,352 --> 00:00:34,100
But it's given us
vast amounts of computing,

13
00:00:34,124 --> 00:00:35,929
huge amounts of memory

14
00:00:35,953 --> 00:00:40,895
and the digital revolution
that we all experience and enjoy today.

15
00:00:41,665 --> 00:00:44,433
But the bad news is,

16
00:00:44,457 --> 00:00:47,589
we're about to hit a digital roadblock,

17
00:00:47,613 --> 00:00:51,963
as the rate of miniaturization
of transistors is slowing down.

18
00:00:52,471 --> 00:00:55,345
And this is happening
at exactly the same time

19
00:00:55,369 --> 00:00:59,367
as our innovation in software
is continuing relentlessly

20
00:00:59,391 --> 00:01:03,151
with artificial intelligence and big data.

21
00:01:03,175 --> 00:01:08,215
And our devices regularly perform
facial recognition or augment our reality

22
00:01:08,239 --> 00:01:12,464
or even drive cars down
our treacherous, chaotic roads.

23
00:01:12,959 --> 00:01:14,166
It's amazing.

24
00:01:14,618 --> 00:01:19,285
But if we don't keep up
with the appetite of our software,

25
00:01:19,309 --> 00:01:23,096
we could reach a point
in the development of our technology

26
00:01:23,120 --> 00:01:27,330
where the things that we could do
with software could, in fact, be limited

27
00:01:27,354 --> 00:01:28,625
by our hardware.

28
00:01:29,075 --> 00:01:33,583
We've all experienced the frustration
of an old smartphone or tablet

29
00:01:33,607 --> 00:01:36,771
grinding slowly to a halt over time

30
00:01:36,795 --> 00:01:40,770
under the ever-increasing weight
of software updates and new features.

31
00:01:40,794 --> 00:01:44,177
And it worked just fine
when we bought it not so long ago.

32
00:01:44,201 --> 00:01:48,711
But the hungry software engineers
have eaten up all the hardware capacity

33
00:01:48,735 --> 00:01:50,041
over time.

34
00:01:51,883 --> 00:01:55,495
The semiconductor industry
is very well aware of this

35
00:01:55,519 --> 00:01:59,403
and is working on
all sorts of creative solutions,

36
00:01:59,427 --> 00:02:03,738
such as going beyond transistors
to quantum computing

37
00:02:03,762 --> 00:02:07,974
or even working with transistors
in alternative architectures

38
00:02:07,998 --> 00:02:09,601
such as neural networks

39
00:02:09,625 --> 00:02:12,638
to make more robust
and efficient circuits.

40
00:02:13,270 --> 00:02:16,609
But these approaches
will take quite some time,

41
00:02:16,633 --> 00:02:21,260
and we're really looking for a much more
immediate solution to this problem.

42
00:02:22,899 --> 00:02:27,681
The reason why the rate of miniaturization
of transistors is slowing down

43
00:02:27,705 --> 00:02:32,391
is due to the ever-increasing complexity
of the manufacturing process.

44
00:02:33,142 --> 00:02:36,392
The transistor used to be
a big, bulky device,

45
00:02:36,416 --> 00:02:39,725
until the invent of the integrated circuit

46
00:02:39,749 --> 00:02:42,440
based on pure crystalline silicon wafers.

47
00:02:42,946 --> 00:02:45,725
And after 50 years
of continuous development,

48
00:02:45,749 --> 00:02:49,122
we can now achieve
transistor features dimensions

49
00:02:49,146 --> 00:02:51,675
down to 10 nanometers.

50
00:02:52,361 --> 00:02:54,798
You can fit more than
a billion transistors

51
00:02:54,822 --> 00:02:57,785
in a single square millimeter of silicon.

52
00:02:58,273 --> 00:03:00,295
And to put this into perspective:

53
00:03:00,319 --> 00:03:04,145
a human hair is 100 microns across.

54
00:03:04,169 --> 00:03:06,688
A red blood cell,
which is essentially invisible,

55
00:03:06,712 --> 00:03:08,311
is eight microns across,

56
00:03:08,335 --> 00:03:11,735
and you can place 12 across
the width of a human hair.

57
00:03:12,467 --> 00:03:15,567
But a transistor, in comparison,
is much smaller,

58
00:03:15,591 --> 00:03:19,439
at a tiny fraction of a micron across.

59
00:03:19,463 --> 00:03:23,009
You could place more than 260 transistors

60
00:03:23,033 --> 00:03:25,011
across a single red blood cell

61
00:03:25,035 --> 00:03:29,499
or more than 3,000 across
the width of a human hair.

62
00:03:29,523 --> 00:03:33,847
It really is incredible nanotechnology
in your pocket right now.

63
00:03:35,204 --> 00:03:37,392
And besides the obvious benefit

64
00:03:37,416 --> 00:03:41,250
of being able to place more,
smaller transistors on a chip,

65
00:03:41,984 --> 00:03:45,476
smaller transistors are faster switches,

66
00:03:46,166 --> 00:03:50,567
and smaller transistors are also
more efficient switches.

67
00:03:50,591 --> 00:03:53,068
So this combination has given us

68
00:03:53,092 --> 00:03:57,391
lower cost, higher performance
and higher efficiency electronics

69
00:03:57,415 --> 00:03:59,478
that we all enjoy today.

70
00:04:02,415 --> 00:04:05,179
To manufacture these integrated circuits,

71
00:04:05,203 --> 00:04:08,411
the transistors are built up
layer by layer,

72
00:04:08,435 --> 00:04:10,788
on a pure crystalline silicon wafer.

73
00:04:11,332 --> 00:04:13,560
And in an oversimplified sense,

74
00:04:13,584 --> 00:04:17,865
every tiny feature
of the circuit is projected

75
00:04:17,889 --> 00:04:20,221
onto the surface of the silicon wafer

76
00:04:20,245 --> 00:04:23,924
and recorded in a light-sensitive material

77
00:04:23,948 --> 00:04:26,887
and then etched through
the light-sensitive material

78
00:04:26,911 --> 00:04:29,932
to leave the pattern
in the underlying layers.

79
00:04:30,612 --> 00:04:34,696
And this process has been
dramatically improved over the years

80
00:04:34,720 --> 00:04:37,493
to give the electronics
performance we have today.

81
00:04:38,279 --> 00:04:41,721
But as the transistor features
get smaller and smaller,

82
00:04:41,745 --> 00:04:44,782
we're really approaching
the physical limitations

83
00:04:44,806 --> 00:04:46,689
of this manufacturing technique.

84
00:04:48,515 --> 00:04:51,620
The latest systems
for doing this patterning

85
00:04:51,644 --> 00:04:53,947
have become so complex

86
00:04:53,971 --> 00:04:58,701
that they reportedly cost
more than 100 million dollars each.

87
00:04:58,725 --> 00:05:03,012
And semiconductor factories
contain dozens of these machines.

88
00:05:03,036 --> 00:05:07,462
So people are seriously questioning:
Is this approach long-term viable?

89
00:05:08,441 --> 00:05:12,121
But we believe we can do
this chip manufacturing

90
00:05:12,145 --> 00:05:16,168
in a totally different
and much more cost-effective way

91
00:05:16,966 --> 00:05:20,939
using molecular engineering
and mimicking nature

92
00:05:20,963 --> 00:05:24,576
down at the nanoscale dimensions
of our transistors.

93
00:05:25,267 --> 00:05:29,928
As I said, the conventional manufacturing
takes every tiny feature of the circuit

94
00:05:29,952 --> 00:05:32,076
and projects it onto the silicon.

95
00:05:32,818 --> 00:05:35,562
But if you look at the structure
of an integrated circuit,

96
00:05:35,586 --> 00:05:37,560
the transistor arrays,

97
00:05:37,584 --> 00:05:41,213
many of the features are repeated
millions of times.

98
00:05:41,237 --> 00:05:43,845
It's a highly periodic structure.

99
00:05:44,331 --> 00:05:47,399
So we want to take advantage
of this periodicity

100
00:05:47,423 --> 00:05:50,120
in our alternative
manufacturing technique.

101
00:05:50,144 --> 00:05:53,579
We want to use self-assembling materials

102
00:05:53,603 --> 00:05:56,580
to naturally form the periodic structures

103
00:05:56,604 --> 00:05:58,987
that we need for our transistors.

104
00:06:00,052 --> 00:06:02,194
We do this with the materials,

105
00:06:02,218 --> 00:06:05,655
then the materials do the hard work
of the fine patterning,

106
00:06:05,679 --> 00:06:10,538
rather than pushing the projection
technology to its limits and beyond.

107
00:06:11,909 --> 00:06:15,808
Self-assembly is seen in nature
in many different places,

108
00:06:15,832 --> 00:06:19,242
from lipid membranes to cell structures,

109
00:06:19,266 --> 00:06:22,321
so we do know it can be a robust solution.

110
00:06:22,345 --> 00:06:25,906
If it's good enough for nature,
it should be good enough for us.

111
00:06:26,549 --> 00:06:31,349
So we want to take this naturally
occurring, robust self-assembly

112
00:06:31,373 --> 00:06:35,338
and use it for the manufacturing
of our semiconductor technology.

113
00:06:36,929 --> 00:06:39,544
One type of self-assemble material --

114
00:06:40,388 --> 00:06:42,635
it's called a block co-polymer --

115
00:06:42,659 --> 00:06:47,442
consists of two polymer chains
just a few tens of nanometers in length.

116
00:06:47,466 --> 00:06:49,517
But these chains hate each other.

117
00:06:49,541 --> 00:06:51,025
They repel each other,

118
00:06:51,049 --> 00:06:54,946
very much like oil and water
or my teenage son and daughter.

119
00:06:54,970 --> 00:06:56,327
(Laughter)

120
00:06:56,351 --> 00:06:59,125
But we cruelly bond them together,

121
00:06:59,149 --> 00:07:01,844
creating an inbuilt
frustration in the system,

122
00:07:01,868 --> 00:07:04,074
as they try to separate from each other.

123
00:07:04,716 --> 00:07:08,001
And in the bulk material,
there are billions of these,

124
00:07:08,025 --> 00:07:11,326
and the similar components
try to stick together,

125
00:07:11,350 --> 00:07:14,159
and the opposing components
try to separate from each other

126
00:07:14,183 --> 00:07:15,338
at the same time.

127
00:07:15,362 --> 00:07:19,116
And this has a built-in frustration,
a tension in the system,

128
00:07:19,140 --> 00:07:23,449
so it moves around, it squirms
until a shape is formed.

129
00:07:24,209 --> 00:07:28,257
And the natural self-assembled shape
that is formed is nanoscale,

130
00:07:28,281 --> 00:07:32,008
it's regular, it's periodic,
and it's long range,

131
00:07:32,032 --> 00:07:35,890
which is exactly what we need
for our transistor arrays.

132
00:07:37,347 --> 00:07:39,878
So we can use molecular engineering

133
00:07:39,902 --> 00:07:42,966
to design different shapes
of different sizes

134
00:07:42,990 --> 00:07:45,053
and of different periodicities.

135
00:07:45,077 --> 00:07:47,808
So for example, if we take
a symmetrical molecule,

136
00:07:47,832 --> 00:07:50,907
where the two polymer chains
are similar length,

137
00:07:50,931 --> 00:07:53,602
the natural self-assembled
structure that is formed

138
00:07:53,626 --> 00:07:56,555
is a long, meandering line,

139
00:07:56,579 --> 00:07:58,389
very much like a fingerprint.

140
00:07:58,951 --> 00:08:01,273
And the width of the fingerprint lines

141
00:08:01,297 --> 00:08:03,307
and the distance between them

142
00:08:03,331 --> 00:08:07,242
is determined by the lengths
of our polymer chains

143
00:08:07,266 --> 00:08:10,560
but also the level of built-in
frustration in the system.

144
00:08:11,320 --> 00:08:13,878
And we can even create
more elaborate structures

145
00:08:15,487 --> 00:08:17,926
if we use unsymmetrical molecules,

146
00:08:18,839 --> 00:08:22,924
where one polymer chain
is significantly shorter than the other.

147
00:08:23,749 --> 00:08:26,459
And the self-assembled structure
that forms in this case

148
00:08:26,483 --> 00:08:30,283
is with the shorter chains
forming a tight ball in the middle,

149
00:08:30,307 --> 00:08:34,148
and it's surrounded by the longer,
opposing polymer chains,

150
00:08:34,172 --> 00:08:36,220
forming a natural cylinder.

151
00:08:37,089 --> 00:08:39,164
And the size of this cylinder

152
00:08:39,188 --> 00:08:42,603
and the distance between
the cylinders, the periodicity,

153
00:08:42,627 --> 00:08:46,221
is again determined by how long
we make the polymer chains

154
00:08:46,245 --> 00:08:48,983
and the level of built-in frustration.

155
00:08:49,896 --> 00:08:53,774
So in other words, we're using
molecular engineering

156
00:08:53,798 --> 00:08:56,623
to self-assemble nanoscale structures

157
00:08:56,647 --> 00:09:01,557
that can be lines or cylinders
the size and periodicity of our design.

158
00:09:02,369 --> 00:09:05,666
We're using chemistry,
chemical engineering,

159
00:09:05,690 --> 00:09:10,479
to manufacture the nanoscale features
that we need for our transistors.

160
00:09:13,611 --> 00:09:17,660
But the ability
to self-assemble these structures

161
00:09:17,684 --> 00:09:20,121
only takes us half of the way,

162
00:09:20,145 --> 00:09:22,954
because we still need
to position these structures

163
00:09:22,978 --> 00:09:26,528
where we want the transistors
in the integrated circuit.

164
00:09:27,246 --> 00:09:29,984
But we can do this relatively easily

165
00:09:30,008 --> 00:09:36,985
using wide guide structures that pin down
the self-assembled structures,

166
00:09:37,009 --> 00:09:38,930
anchoring them in place

167
00:09:38,954 --> 00:09:41,801
and forcing the rest
of the self-assembled structures

168
00:09:41,825 --> 00:09:43,175
to lie parallel,

169
00:09:43,199 --> 00:09:45,599
aligned with our guide structure.

170
00:09:46,510 --> 00:09:51,149
For example, if we want to make
a fine, 40-nanometer line,

171
00:09:51,173 --> 00:09:55,311
which is very difficult to manufacture
with conventional projection technology,

172
00:09:56,274 --> 00:10:01,059
we can manufacture
a 120-nanometer guide structure

173
00:10:01,083 --> 00:10:03,587
with normal projection technology,

174
00:10:03,611 --> 00:10:10,202
and this structure will align three
of the 40-nanometer lines in between.

175
00:10:10,226 --> 00:10:14,995
So the materials are doing
the most difficult fine patterning.

176
00:10:15,790 --> 00:10:19,697
And we call this whole approach
"directed self-assembly."

177
00:10:21,586 --> 00:10:24,340
The challenge with directed self-assembly

178
00:10:24,364 --> 00:10:28,840
is that the whole system
needs to align almost perfectly,

179
00:10:28,864 --> 00:10:34,145
because any tiny defect in the structure
could cause a transistor failure.

180
00:10:34,169 --> 00:10:37,138
And because there are billions
of transistors in our circuit,

181
00:10:37,162 --> 00:10:40,390
we need an almost
molecularly perfect system.

182
00:10:40,977 --> 00:10:42,982
But we're going to extraordinary measures

183
00:10:43,006 --> 00:10:44,173
to achieve this,

184
00:10:44,197 --> 00:10:47,189
from the cleanliness of our chemistry

185
00:10:47,213 --> 00:10:49,539
to the careful processing
of these materials

186
00:10:49,563 --> 00:10:51,134
in the semiconductor factory

187
00:10:51,158 --> 00:10:55,730
to remove even the smallest
nanoscopic defects.

188
00:10:57,311 --> 00:11:02,501
So directed self-assembly
is an exciting new disruptive technology,

189
00:11:02,525 --> 00:11:05,094
but it is still in the development stage.

190
00:11:05,680 --> 00:11:09,541
But we're growing in confidence
that we could, in fact, introduce it

191
00:11:09,565 --> 00:11:11,252
to the semiconductor industry

192
00:11:11,276 --> 00:11:14,233
as a revolutionary new
manufacturing process

193
00:11:14,257 --> 00:11:16,324
in just the next few years.

194
00:11:17,014 --> 00:11:20,048
And if we can do this,
if we're successful,

195
00:11:20,072 --> 00:11:21,603
we'll be able to continue

196
00:11:21,627 --> 00:11:24,885
with the cost-effective
miniaturization of transistors,

197
00:11:24,909 --> 00:11:28,662
continue with the spectacular
expansion of computing

198
00:11:28,686 --> 00:11:30,568
and the digital revolution,

199
00:11:30,592 --> 00:11:34,137
and what's more, this could even
be the dawn of a new era

200
00:11:34,161 --> 00:11:36,392
of molecular manufacturing.

201
00:11:36,416 --> 00:11:37,947
How cool is that?

202
00:11:38,519 --> 00:11:39,677
Thank you.

203
00:11:39,701 --> 00:11:43,910
(Applause)