How to take a picture of a black hole | Katie Bouman

3,361,362 views ・ 2017-04-28

TED


Please double-click on the English subtitles below to play the video.

00:13
In the movie "Interstellar,"
0
13436
1860
00:15
we get an up-close look at a supermassive black hole.
1
15320
3327
00:18
Set against a backdrop of bright gas,
2
18671
2143
00:20
the black hole's massive gravitational pull
3
20838
2118
00:22
bends light into a ring.
4
22980
1435
00:24
However, this isn't a real photograph,
5
24439
2109
00:26
but a computer graphic rendering --
6
26572
1786
00:28
an artistic interpretation of what a black hole might look like.
7
28382
3390
00:32
A hundred years ago,
8
32401
1166
00:33
Albert Einstein first published his theory of general relativity.
9
33591
3601
00:37
In the years since then,
10
37216
1439
00:38
scientists have provided a lot of evidence in support of it.
11
38679
2973
00:41
But one thing predicted from this theory, black holes,
12
41676
3084
00:44
still have not been directly observed.
13
44784
2350
00:47
Although we have some idea as to what a black hole might look like,
14
47158
3206
00:50
we've never actually taken a picture of one before.
15
50388
2779
00:53
However, you might be surprised to know that that may soon change.
16
53191
4279
00:57
We may be seeing our first picture of a black hole in the next couple years.
17
57494
4164
01:01
Getting this first picture will come down to an international team of scientists,
18
61682
3958
01:05
an Earth-sized telescope
19
65664
1567
01:07
and an algorithm that puts together the final picture.
20
67255
2832
01:10
Although I won't be able to show you a real picture of a black hole today,
21
70111
3528
01:13
I'd like to give you a brief glimpse into the effort involved
22
73663
2911
01:16
in getting that first picture.
23
76598
1613
01:19
My name is Katie Bouman,
24
79477
1437
01:20
and I'm a PhD student at MIT.
25
80938
2566
01:23
I do research in a computer science lab
26
83528
2027
01:25
that works on making computers see through images and video.
27
85579
3298
01:28
But although I'm not an astronomer,
28
88901
2162
01:31
today I'd like to show you
29
91087
1285
01:32
how I've been able to contribute to this exciting project.
30
92396
2903
01:35
If you go out past the bright city lights tonight,
31
95323
2831
01:38
you may just be lucky enough to see a stunning view
32
98178
2436
01:40
of the Milky Way Galaxy.
33
100638
1493
01:42
And if you could zoom past millions of stars,
34
102155
2462
01:44
26,000 light-years toward the heart of the spiraling Milky Way,
35
104641
3755
01:48
we'd eventually reach a cluster of stars right at the center.
36
108420
3521
01:51
Peering past all the galactic dust with infrared telescopes,
37
111965
3206
01:55
astronomers have watched these stars for over 16 years.
38
115195
3867
01:59
But it's what they don't see that is the most spectacular.
39
119086
3589
02:02
These stars seem to orbit an invisible object.
40
122699
3066
02:05
By tracking the paths of these stars,
41
125789
2323
02:08
astronomers have concluded
42
128136
1294
02:09
that the only thing small and heavy enough to cause this motion
43
129454
3129
02:12
is a supermassive black hole --
44
132607
1968
02:14
an object so dense that it sucks up anything that ventures too close --
45
134599
4178
02:18
even light.
46
138801
1494
02:20
But what happens if we were to zoom in even further?
47
140319
3061
02:23
Is it possible to see something that, by definition, is impossible to see?
48
143404
4733
02:28
Well, it turns out that if we were to zoom in at radio wavelengths,
49
148719
3244
02:31
we'd expect to see a ring of light
50
151987
1682
02:33
caused by the gravitational lensing of hot plasma
51
153693
2411
02:36
zipping around the black hole.
52
156128
1829
02:37
In other words,
53
157981
1160
02:39
the black hole casts a shadow on this backdrop of bright material,
54
159165
3171
02:42
carving out a sphere of darkness.
55
162360
1842
02:44
This bright ring reveals the black hole's event horizon,
56
164226
3339
02:47
where the gravitational pull becomes so great
57
167589
2400
02:50
that not even light can escape.
58
170013
1626
02:51
Einstein's equations predict the size and shape of this ring,
59
171663
2859
02:54
so taking a picture of it wouldn't only be really cool,
60
174546
3208
02:57
it would also help to verify that these equations hold
61
177778
2618
03:00
in the extreme conditions around the black hole.
62
180420
2466
03:02
However, this black hole is so far away from us,
63
182910
2558
03:05
that from Earth, this ring appears incredibly small --
64
185492
3098
03:08
the same size to us as an orange on the surface of the moon.
65
188614
3590
03:12
That makes taking a picture of it extremely difficult.
66
192758
2824
03:16
Why is that?
67
196645
1302
03:18
Well, it all comes down to a simple equation.
68
198512
3188
03:21
Due to a phenomenon called diffraction,
69
201724
2416
03:24
there are fundamental limits
70
204164
1355
03:25
to the smallest objects that we can possibly see.
71
205543
2670
03:28
This governing equation says that in order to see smaller and smaller,
72
208789
3672
03:32
we need to make our telescope bigger and bigger.
73
212485
2587
03:35
But even with the most powerful optical telescopes here on Earth,
74
215096
3069
03:38
we can't even get close to the resolution necessary
75
218189
2419
03:40
to image on the surface of the moon.
76
220632
2198
03:42
In fact, here I show one of the highest resolution images ever taken
77
222854
3617
03:46
of the moon from Earth.
78
226495
1397
03:47
It contains roughly 13,000 pixels,
79
227916
2557
03:50
and yet each pixel would contain over 1.5 million oranges.
80
230497
4050
03:55
So how big of a telescope do we need
81
235396
1972
03:57
in order to see an orange on the surface of the moon
82
237392
2765
04:00
and, by extension, our black hole?
83
240181
2214
04:02
Well, it turns out that by crunching the numbers,
84
242419
2340
04:04
you can easily calculate that we would need a telescope
85
244783
2610
04:07
the size of the entire Earth.
86
247417
1393
04:08
(Laughter)
87
248834
1024
04:09
If we could build this Earth-sized telescope,
88
249882
2119
04:12
we could just start to make out that distinctive ring of light
89
252025
2925
04:14
indicative of the black hole's event horizon.
90
254974
2183
04:17
Although this picture wouldn't contain all the detail we see
91
257181
2918
04:20
in computer graphic renderings,
92
260123
1506
04:21
it would allow us to safely get our first glimpse
93
261653
2299
04:23
of the immediate environment around a black hole.
94
263976
2487
04:26
However, as you can imagine,
95
266487
1613
04:28
building a single-dish telescope the size of the Earth is impossible.
96
268124
3624
04:31
But in the famous words of Mick Jagger,
97
271772
1887
04:33
"You can't always get what you want,
98
273683
1791
04:35
but if you try sometimes, you just might find
99
275498
2187
04:37
you get what you need."
100
277709
1215
04:38
And by connecting telescopes from around the world,
101
278948
2464
04:41
an international collaboration called the Event Horizon Telescope
102
281436
3538
04:44
is creating a computational telescope the size of the Earth,
103
284998
3109
04:48
capable of resolving structure
104
288131
1537
04:49
on the scale of a black hole's event horizon.
105
289692
2199
04:51
This network of telescopes is scheduled to take its very first picture
106
291915
3387
04:55
of a black hole next year.
107
295326
1815
04:57
Each telescope in the worldwide network works together.
108
297165
3338
05:00
Linked through the precise timing of atomic clocks,
109
300527
2712
05:03
teams of researchers at each of the sites freeze light
110
303263
2657
05:05
by collecting thousands of terabytes of data.
111
305944
2962
05:08
This data is then processed in a lab right here in Massachusetts.
112
308930
5017
05:13
So how does this even work?
113
313971
1794
05:15
Remember if we want to see the black hole in the center of our galaxy,
114
315789
3403
05:19
we need to build this impossibly large Earth-sized telescope?
115
319216
2982
05:22
For just a second, let's pretend we could build
116
322222
2232
05:24
a telescope the size of the Earth.
117
324478
1842
05:26
This would be a little bit like turning the Earth
118
326344
2455
05:28
into a giant spinning disco ball.
119
328823
1747
05:30
Each individual mirror would collect light
120
330594
2200
05:32
that we could then combine together to make a picture.
121
332818
2597
05:35
However, now let's say we remove most of those mirrors
122
335439
2661
05:38
so only a few remained.
123
338124
1972
05:40
We could still try to combine this information together,
124
340120
2877
05:43
but now there are a lot of holes.
125
343021
1993
05:45
These remaining mirrors represent the locations where we have telescopes.
126
345038
4373
05:49
This is an incredibly small number of measurements to make a picture from.
127
349435
4079
05:53
But although we only collect light at a few telescope locations,
128
353538
3838
05:57
as the Earth rotates, we get to see other new measurements.
129
357400
3423
06:00
In other words, as the disco ball spins, those mirrors change locations
130
360847
3819
06:04
and we get to observe different parts of the image.
131
364690
2899
06:07
The imaging algorithms we develop fill in the missing gaps of the disco ball
132
367613
4018
06:11
in order to reconstruct the underlying black hole image.
133
371655
3033
06:14
If we had telescopes located everywhere on the globe --
134
374712
2636
06:17
in other words, the entire disco ball --
135
377372
1941
06:19
this would be trivial.
136
379337
1284
06:20
However, we only see a few samples, and for that reason,
137
380645
3322
06:23
there are an infinite number of possible images
138
383991
2388
06:26
that are perfectly consistent with our telescope measurements.
139
386403
2964
06:29
However, not all images are created equal.
140
389391
3016
06:32
Some of those images look more like what we think of as images than others.
141
392849
4458
06:37
And so, my role in helping to take the first image of a black hole
142
397331
3222
06:40
is to design algorithms that find the most reasonable image
143
400577
2932
06:43
that also fits the telescope measurements.
144
403533
2222
06:46
Just as a forensic sketch artist uses limited descriptions
145
406727
3942
06:50
to piece together a picture using their knowledge of face structure,
146
410693
3514
06:54
the imaging algorithms I develop use our limited telescope data
147
414231
3315
06:57
to guide us to a picture that also looks like stuff in our universe.
148
417570
4322
07:01
Using these algorithms, we're able to piece together pictures
149
421916
3651
07:05
from this sparse, noisy data.
150
425591
2180
07:07
So here I show a sample reconstruction done using simulated data,
151
427795
4529
07:12
when we pretend to point our telescopes
152
432348
1933
07:14
to the black hole in the center of our galaxy.
153
434305
2585
07:16
Although this is just a simulation, reconstruction such as this give us hope
154
436914
4455
07:21
that we'll soon be able to reliably take the first image of a black hole
155
441393
3453
07:24
and from it, determine the size of its ring.
156
444870
2595
07:28
Although I'd love to go on about all the details of this algorithm,
157
448118
3199
07:31
luckily for you, I don't have the time.
158
451341
2174
07:33
But I'd still like to give you a brief idea
159
453539
2001
07:35
of how we define what our universe looks like,
160
455564
2302
07:37
and how we use this to reconstruct and verify our results.
161
457890
4466
07:42
Since there are an infinite number of possible images
162
462380
2496
07:44
that perfectly explain our telescope measurements,
163
464900
2365
07:47
we have to choose between them in some way.
164
467289
2605
07:49
We do this by ranking the images
165
469918
1838
07:51
based upon how likely they are to be the black hole image,
166
471780
2834
07:54
and then choosing the one that's most likely.
167
474638
2482
07:57
So what do I mean by this exactly?
168
477144
2195
07:59
Let's say we were trying to make a model
169
479862
1978
08:01
that told us how likely an image were to appear on Facebook.
170
481864
3183
08:05
We'd probably want the model to say
171
485071
1701
08:06
it's pretty unlikely that someone would post this noise image on the left,
172
486796
3557
08:10
and pretty likely that someone would post a selfie
173
490377
2419
08:12
like this one on the right.
174
492820
1334
08:14
The image in the middle is blurry,
175
494178
1639
08:15
so even though it's more likely we'd see it on Facebook
176
495841
2639
08:18
compared to the noise image,
177
498504
1360
08:19
it's probably less likely we'd see it compared to the selfie.
178
499888
2960
08:22
But when it comes to images from the black hole,
179
502872
2290
08:25
we're posed with a real conundrum: we've never seen a black hole before.
180
505186
3502
08:28
In that case, what is a likely black hole image,
181
508712
2291
08:31
and what should we assume about the structure of black holes?
182
511027
2938
08:33
We could try to use images from simulations we've done,
183
513989
2632
08:36
like the image of the black hole from "Interstellar,"
184
516645
2530
08:39
but if we did this, it could cause some serious problems.
185
519199
2938
08:42
What would happen if Einstein's theories didn't hold?
186
522161
3380
08:45
We'd still want to reconstruct an accurate picture of what was going on.
187
525565
3961
08:49
If we bake Einstein's equations too much into our algorithms,
188
529550
3371
08:52
we'll just end up seeing what we expect to see.
189
532945
2755
08:55
In other words, we want to leave the option open
190
535724
2276
08:58
for there being a giant elephant at the center of our galaxy.
191
538024
2923
09:00
(Laughter)
192
540971
1057
09:02
Different types of images have very distinct features.
193
542052
2989
09:05
We can easily tell the difference between black hole simulation images
194
545065
3548
09:08
and images we take every day here on Earth.
195
548637
2276
09:10
We need a way to tell our algorithms what images look like
196
550937
3104
09:14
without imposing one type of image's features too much.
197
554065
3249
09:17
One way we can try to get around this
198
557865
1893
09:19
is by imposing the features of different kinds of images
199
559782
3062
09:22
and seeing how the type of image we assume affects our reconstructions.
200
562868
4130
09:27
If all images' types produce a very similar-looking image,
201
567712
3491
09:31
then we can start to become more confident
202
571227
2057
09:33
that the image assumptions we're making are not biasing this picture that much.
203
573308
4173
09:37
This is a little bit like giving the same description
204
577505
2990
09:40
to three different sketch artists from all around the world.
205
580519
2996
09:43
If they all produce a very similar-looking face,
206
583539
2860
09:46
then we can start to become confident
207
586423
1793
09:48
that they're not imposing their own cultural biases on the drawings.
208
588240
3616
09:51
One way we can try to impose different image features
209
591880
3315
09:55
is by using pieces of existing images.
210
595219
2441
09:58
So we take a large collection of images,
211
598214
2160
10:00
and we break them down into their little image patches.
212
600398
2718
10:03
We then can treat each image patch a little bit like pieces of a puzzle.
213
603140
4285
10:07
And we use commonly seen puzzle pieces to piece together an image
214
607449
4278
10:11
that also fits our telescope measurements.
215
611751
2452
10:15
Different types of images have very distinctive sets of puzzle pieces.
216
615040
3743
10:18
So what happens when we take the same data
217
618807
2806
10:21
but we use different sets of puzzle pieces to reconstruct the image?
218
621637
4130
10:25
Let's first start with black hole image simulation puzzle pieces.
219
625791
4766
10:30
OK, this looks reasonable.
220
630581
1591
10:32
This looks like what we expect a black hole to look like.
221
632196
2694
10:34
But did we just get it
222
634914
1193
10:36
because we just fed it little pieces of black hole simulation images?
223
636131
3314
10:39
Let's try another set of puzzle pieces
224
639469
1880
10:41
from astronomical, non-black hole objects.
225
641373
2509
10:44
OK, we get a similar-looking image.
226
644914
2126
10:47
And then how about pieces from everyday images,
227
647064
2236
10:49
like the images you take with your own personal camera?
228
649324
2785
10:53
Great, we see the same image.
229
653312
2115
10:55
When we get the same image from all different sets of puzzle pieces,
230
655451
3366
10:58
then we can start to become more confident
231
658841
2046
11:00
that the image assumptions we're making
232
660911
1966
11:02
aren't biasing the final image we get too much.
233
662901
2921
11:05
Another thing we can do is take the same set of puzzle pieces,
234
665846
3253
11:09
such as the ones derived from everyday images,
235
669123
2489
11:11
and use them to reconstruct many different kinds of source images.
236
671636
3600
11:15
So in our simulations,
237
675260
1271
11:16
we pretend a black hole looks like astronomical non-black hole objects,
238
676555
3775
11:20
as well as everyday images like the elephant in the center of our galaxy.
239
680354
3849
11:24
When the results of our algorithms on the bottom look very similar
240
684227
3168
11:27
to the simulation's truth image on top,
241
687419
2096
11:29
then we can start to become more confident in our algorithms.
242
689539
3346
11:32
And I really want to emphasize here
243
692909
1867
11:34
that all of these pictures were created
244
694800
1934
11:36
by piecing together little pieces of everyday photographs,
245
696758
2936
11:39
like you'd take with your own personal camera.
246
699718
2215
11:41
So an image of a black hole we've never seen before
247
701957
3276
11:45
may eventually be created by piecing together pictures we see all the time
248
705257
3943
11:49
of people, buildings, trees, cats and dogs.
249
709224
2745
11:51
Imaging ideas like this will make it possible for us
250
711993
2645
11:54
to take our very first pictures of a black hole,
251
714662
2619
11:57
and hopefully, verify those famous theories
252
717305
2447
11:59
on which scientists rely on a daily basis.
253
719776
2421
12:02
But of course, getting imaging ideas like this working
254
722221
2608
12:04
would never have been possible without the amazing team of researchers
255
724853
3322
12:08
that I have the privilege to work with.
256
728199
1887
12:10
It still amazes me
257
730110
1163
12:11
that although I began this project with no background in astrophysics,
258
731297
3351
12:14
what we have achieved through this unique collaboration
259
734672
2619
12:17
could result in the very first images of a black hole.
260
737315
2759
12:20
But big projects like the Event Horizon Telescope
261
740098
2698
12:22
are successful due to all the interdisciplinary expertise
262
742820
2814
12:25
different people bring to the table.
263
745658
1790
12:27
We're a melting pot of astronomers,
264
747472
1706
12:29
physicists, mathematicians and engineers.
265
749202
2232
12:31
This is what will make it soon possible
266
751458
2554
12:34
to achieve something once thought impossible.
267
754036
2853
12:36
I'd like to encourage all of you to go out
268
756913
2256
12:39
and help push the boundaries of science,
269
759193
2096
12:41
even if it may at first seem as mysterious to you as a black hole.
270
761313
3901
12:45
Thank you.
271
765238
1174
12:46
(Applause)
272
766436
2397
About this website

This site will introduce you to YouTube videos that are useful for learning English. You will see English lessons taught by top-notch teachers from around the world. Double-click on the English subtitles displayed on each video page to play the video from there. The subtitles scroll in sync with the video playback. If you have any comments or requests, please contact us using this contact form.

https://forms.gle/WvT1wiN1qDtmnspy7