How do Electron Microscopes Work? 🔬🛠🔬 Taking Pictures of Atoms

0:00 Hi, this is Jan Jíša,

0:01 I'm the physicist from Thermo Fisher who helped with this video.

0:05 I studied Condensed Matter Physics at Masaryk University in Brno,

0:09 Czech republic.

0:10 Now I specialize in charged particle optics

0:12 and physics of Transmission Electron Microscope (TEM's).

0:15 The small city of Brno in central Europe

0:17 is often considered the city of electron microscopes.

0:20 There are 3 commercial companies developing and producing these machines

0:25 and several state-funded research institutions in related areas.

0:30 As a consequence,

0:31 more than 30% of the world electron microscope production comes from Brno.

0:35 This means several hundred to a few thousand microscopes each year,

0:38 just from our company alone.

0:40 Let's take a quick look at the history

0:42 and then move to the interesting technical stuff.

0:45 The first TEM was built in 1931 by Ernst Ruska and Max Knoll.

0:50 It had a magnification of only 17x,

0:52 but proved the working princples for further development.

0:55 The first commercial one was built by Siemens in 1939 and operated at 60 000 V.

1:00 It had a maximum magnification of 30 000x and a resolution of 10 nm.

1:05 The first Scanning Electron Microscope (SEM) prototype was

1:07 built in 1937 by Manfred von Ardenne,

1:10 but had a relatively low resolution of> 50 nm

1:12 due to poor electronics stability at the time.

1:15 It took a few decades of stability improvements

1:17 that finally lead to the first commercial SEM.

1:20 Introduced in 1965 by the Cambridge Instruments,

1:22 it had a resolution of around 50 nm.

1:25 Then, both the TEM's and SEM's gradually improved in resolution,

1:28 contrast and analytical capabilities.

1:30 Let's quickly look at the video now:

1:32 The colors in an EM image are usually artificial.

1:35 However, in case of elemental mapping (EDXS or EELS methods),

1:38 they do carry actual meaning.

1:40 They correspond to individual elements and their concentration

1:42 at a given point of the specimen

1:45 SEM images have a nice 3D feel to them,

1:47 which makes them easier to understand for our brains.

1:50 This is due to the way the signal is

1:52 created and processed- it maps the surface layer(s).

1:55 TEM images can be thought of as thin 2D

1:57 slices averaged over the thickness of the specimen.

2:00 A "shadow image", if you will, similar to the x-ray machine.

2:05 However, there are several techniques to create 3D

2:07 reconstructions from a series of 2D images.

2:10 One of the common techniques is EM Tomography,

2:12 where the specimen is gradually tilted.

2:15 Another is the Single Particle Analysis where

2:17 a huge amount of identical macromolecules (proteins,

2:19 virus particles) are prepared in a solution and flash-frozen.

2:22 Their random orientation in large numbers

2:25 then provides enough information to reconstruct a 3D model of the macromolecule.

2:30 Yet another method is the Microcrystal Electron Diffraction.

2:32 Here, the examined molecules are left to

2:35 crystallize,

2:35 forming crystals often too small for x-ray diffraction methods (10-100 nm).

2:40 The specimen is then rotated and a series of

2:42 diffraction patterns (instead of direct images) is collected.

2:45 The diffraction patterns are then back-projected in a 3D reciprocal space,

2:50 inverse Fourier transformed back into "normal" space,

2:52 and used for 3D model reconstruction.

2:55 It was quite surprising to me that a very

2:57 large number of substances like to form crystals

3:00 even if only tiny ones.

3:01 This includes materials typically thought of as amorphous in bulk volume.

3:05 Now back to the video: The beam currents are very low, typically 0.001-10 nA.

3:10 Since the electrons are moving extremely fast, in certain optical conditions,

3:15 you can get only 1 electron in the entire column at a time, on average.

3:20 In some sense, this is similar to the electron double-slit experiment,

3:25 except here, our specimen plays the role of a very complicated "slit."

3:30 Image contrast is created by the specimen

3:32 atoms "bending" the path of the electrons.

3:35 In terms of wave optics,

3:36 the specimen introduces a position-dependent phase shift of the electron wave.

3:40 Almost no electrons get absorbed and only a few lose any energy.

3:45 In regular TEM imaging,

3:46 the electrons losing energy (through inelastic

3:48 scattering) contribute only to noise.

3:50 However,

3:50 the same electrons carry additional interesting information that can be accessed

3:55 using the energy filter module.

3:57 This module consists of a magnetic prism

4:00 which sorts (spatially separates) the electrons based on their energy.

4:05 With it,

4:05 you can investigate either the energy loss spectrum of each point of a specimen

4:10 in the scanning mode (EELS method), or create the whole image at once

4:15 whose contrast corresponds only to a selected

4:17 range of energy losses (EFTEM method).

4:20 These very sensitive modes can map the creation

4:22 and movement of various solid state excitations

4:25 and quasiparticles, like phonons, plasmons, polaritons, magnons etc.

4:30 This is how sound, heat and electrical current travel through solids,

4:33 as described by quantum mechanics.

4:35 Back to the video:

4:36 The total magnification can be varied between 25x

4:38 and 200 000 000x in some modes.

4:40 However, meaningful magnifications due to resolution limitations

4:42 usually end around 2 000 000x.

4:45 In optical microscopes,

4:46 aberrations can be overcome by shaping the geometry of the lens' surface

4:50 and selecting a transparent material with a

4:52 suitable index of refraction for the lens.

4:55 In electron microscopy,

4:56 we can't use material lenses- those would stop the electron beam.

5:00 Instead, we must use electric and magnetic fields to control,

5:03 guide and focus the beam.

5:05 These fields and their gradients are governed by

5:07 nature and described by the Maxwell equations,

5:10 so we can't force them to take any shape we'd like.

5:15 And this- together with mechani imperfections-

5:17 is the reason optical aberrations arise.

5:20 Some of the aberrations can be corrected using fairly simple elements.

5:25 These are, typically, electric or magnetic dipoles or quadrupoles.

5:27 Other aberrations are more difficult to deal with.

5:30 Let's look at the video:

5:31 The diffraction limit is a physical limitation for imaging using a wave.

5:35 And it applies to any wave,

5:36 whether it is made of EM fields (light), probability density (particles),

5:40 pressure oscillations (sound) or macroscopic waves of water in a bathtub.

5:45 There are some techniques to overcome the

5:47 diffraction resolution limit even in light microscopy.

5:50 However, they often require a specialized setup,

5:52 multiple image acquisitions, or introduce other inconveniences.

5:55 In electron microscopy,

5:56 the diffraction actually isn't the main limiting factor for image resolution.

6:00 Instead, some of the optical aberrations are.

6:02 Specifically, the spherical and chromatic ones.

6:05 The 50 pm resolution is typically for TEMs

6:07 with a complex optical module called a Corrector.

6:10 This module, not shown in the video,

6:12 corrects the objective spherical, chromatic and higher-order aberrations.

6:15 It consists of several electric or magnetic multipoles

6:20 that gradually shape the beam wavefront to compensate for various aberrations.

6:25 The resolution of a TEM with a Corrector is roughly 2-3x better than without it.

6:30 What always fascinated me about electron microscopes is

6:32 how different areas of physics and engineering

6:35 need to come together and be combined in a single machine,

6:38 in order for it to work.

6:40 There's electricity and magnetism, optics,

6:42 quantum and wave mechanics, special relativity, thermodynamics,

6:45 solid state physics, mechanics, electronics, acoustics,

6:47 vibrational dynamics, precision machining and more.

6:50 All that wrapped up nicely in different flavors and layers of mathematics.

6:55 Fascinating, as Mr.

6:56 Spock would surely put it :o)

7:00 Now to the video: In terms of solid state physics,

7:02 the external electric field lowers the potential barrier

7:05 that electrons have to overcome to be emitted from the material.

7:10 The cathode can also be heated, adding more kinetic energy to the electrons.

7:15 The electrons then either overcome the barrier, or quantum tunnel through it.

7:20 This depends on the specific design and

7:22 operating conditions of the electron source.

7:25 Additionally, the cathode is coated with a thin film of zirconium oxide,

7:30 which further lowers the potential barrier chemically.

7:35 The crystal orientation also plays a role in the potential barrier height.

7:40 Let's look at the video again:

7:41 The real acceleration cascade is wired a bit differently here.

7:45 In reality, the cathode is connected to a negative electric potential (e.g.

7:49 -200 000 V) and the rings gradually increase it to ground potential (0 V),

7:53 so the microscope is easier to insulate

7:55 and safe to touch.

7:56 The principle still holds,

7:57 as it is the potential difference that accelerates the electrons.

8:00 Interesting fact:

8:00 Up to 99.9% of emitted electrons (hundreds of microamps) hit the extractor

8:05 and only the remaining 0.1%(< 10 nA) make it

8:07 through along the axis and become the beam itself.

8:10 Another important reason for the high speeds of

8:12 the electrons is radiation damage of the specimen.

8:15 Counterintuitively, radiation damage in thin specimens usually

8:17 decreases with increasing electron speed.

8:20 This is due to the electrons having less time to interact with the specimen,

8:25 effectively appearing "smaller" for higher speeds.

8:30 In technical terms,

8:31 they have a smaller total scattering cross-section of interaction.

8:35 The typical vacuum pressures in the microscope are:

8:38 10^(-3) Pa around the detectors,

8:40 10^(-5) Pa around the specimen,

8:41 and 10^(-6)- 10^(-8) Pa around the electron source.

8:45 The normal atmospheric pressure is around 10^5 Pa, so in the microscope,

8:50 we create a pressure of up to a thousand billion times lower.

8:55 This is comparable to what's considered the "starting

8:57 point" for vacuum levels of outer space.

9:00 No single pump can reach such low pressures.

9:02 Therefore a cascade of several different types must be used.

9:05 These include scroll, turbomolecular, ion getter,

9:07 cryo and non-evaporative getter pumps.

9:10 Another reason for low pressures in the microscope is

9:12 to avoid the growth of thin contamination layers.

9:15 These can degrade the specimen or even the electron source.

9:20 Let's now take a look at the role of apertures in the microscope.

9:25 Some are indeed used to limit the beam size or its convergence angle.

9:30 But others may also be used to enhance the

9:32 image contrast (at the cost of total current)

9:35 or to create an image using only a certain diffraction peak,

9:38 which results in a different contrast.

9:40 The apertures' use depends on the optical plane where they're located.

9:45 Their typical sizes vary from a few tens of micrometers to several hundred.

9:50 There are both fixed and variable apertures.

9:52 The latter can be changed by the microscopist

9:55 based on the needs of their experiment or observation.

10:00 As for the holders,

10:01 there are even more types and allow for specimen cooling, heating,

10:05 applying mechanical strain or performing dynamic in situ experiments,

10:08 like observing chemical reactions.

10:10 Now let's talk a bit about the electron-specimen interaction.

10:15 Depending on the level of needed detail,

10:17 it can be described by different mathematical models.

10:20 Usually, this is either in terms of probabilities and

10:22 scattering cross-sections for electrons as a particles,

10:25 or in terms of phase and amplitude changes to the incoming electron wave.

10:30 Increasing the thickness of a specimen increases

10:32 the probability of multiple scattering events.

10:35 These obscure the information from the first scattering

10:37 event- such electrons then only contribute to noise.

10:40 Let's now look at the lenses and their aberrations.

10:45 The aberrations can be categorized in many ways:

10:47 axial or off-axial, energy dependent or independent,

10:50 geometrically allowed or parasitic, sorted by their order and symmetry.

10:55 The most common ones which are relatively "easy" to correct are:

11:00 shift, defocus, twofold astigmatism, coma and threefold astigmatism.

11:05 Then there are the spherical and chromatic

11:07 aberrations than require the Corrector to compensate.

11:10 This module also takes care of certain higher-order aberrations.

11:15 Other imperfections like distortions or field curvature

11:17 may not impact the resolution directly,

11:20 but influence magnification or focus in different parts of the image.

11:25 There's a whole zoo of various aberrations

11:27 and with better and better Correctors,

11:30 we're chipping away at them to get closer

11:32 and closer to the ultimate diffraction limit.

11:35 The lens' optical aberrations are not the

11:37 only limiting factors of practical image resolution,

11:39 though.

11:40 The indirect contributions include: Stability of the electronics,

11:43 brightness of the electron source,

11:45 external disturbances and quality of the specimen.

11:50 A bit of a sidenote: Although reducing optical aberrations is an active area

11:55 of research and development, it is by far not the only one.

12:00 Some of the other important ones include

12:02 reducing radiation damage to the specimen,

12:05 developing more sensitive and faster detectors, improving electron sources,

12:10 automating the microscopes, improving the specimen transfer,

12:15 or looking into time-resolved microscopy and multi-modal analysis.

12:20 Time-resolved microscopy allows for observing

12:22 extremely fast microscopic phenomena

12:25 and involves a pulsed electron beam instead of a continuous one.

12:30 Multi-modal analysis means collecting information from different detectors,

12:35 optical modes and energies and combining

12:37 them in a single information-rich object.

12:40 But let's now get back to the electron optics and look at the video again.

12:45 The typical lens currents are 1-10 A,

12:47 with the lens drawing power of a few hundred watts.

12:50 The corresponding magnetic field on the axis is around 1-2 T.

12:55 Although increasing the field strength in the

12:57 lens gap would reduce the aberrations,

13:00 we are limited by the amount of magnetic flux the

13:02 soft iron can hold and "deliver" to this gap.

13:05 Once the iron gets magnetically saturated,

13:06 the flux starts leaking out and spreading to undesired places.

13:10 Interesting mechanism:

13:10 The Lorentz force first makes the electrons spin (gyrate) along the field lines.

13:15 And it is this resulting rotational component of the electron velocity

13:20 that causes the Lorentz force to also focus the electrons towards the axis.

13:25 Without electron gyration, there would be no magnetic focusing of the beam.

13:30 In principle, electrostatic focusing and beam control could also be used.

13:35 But once the electrons reach their relativistic speeds,

13:38 this method becomes too inefficient.

13:40 The electrostatic force is independent of the electron speed.

13:43 And the faster the electron is,

13:45 the less time it spends in the lens' field.

13:47 The Lorentz force, on the other hand,

13:50 is proportional to the electron speed,

13:52 so the magnetic lenses keep their efficiency

13:55 and focusing power even for high electron speeds and energies.

14:00 Although it is not possible to construct a single magnetic diverging lens,

14:05 that's not a huge limiting factor.

14:07 By creating beam crossovers at appropriate optical planes

14:10 we can achieve the effect of spreading the

14:12 beam just like a diverging lens would.

14:15 There are quite a lot of analogies between light and electron optics (lenses,

14:19 mirrors phase plates),

14:20 and a lot of effects can be described by very similar mathematics.

14:25 It is the physical representation and construction

14:27 of individual optical elements that differs.

14:30 Let's look at the video now:

14:32 The phosphorescent screen is a polycrystalline scintillation layer

14:35 that also serves as a beam current measuring sensor.

14:40 In the older TEM models,

14:41 the microscopists looked directly at this screen through a glass window

14:45 instead of a camera, and saw its calming green glow.

14:50 Some of the high resolution cameras at the bottom also use a scintillation layer

14:55 that first converts the fast-moving electrons to light.

15:00 Other cameras use direct detection,

15:02 converting the primary electrons from the beam

15:05 to electrical signal in individual pixels directly.

15:10 Both detection methods have their pros and cons.

15:15 Direct electron detection offers ultimate sensitivity and low noise,

15:20 while indirect detection offers higher dynamic

15:22 range and more radiation hardness.

15:25 So, looking at the list of modules on the screen,

15:27 I haven't yet mentioned phase plates and monochromators.

15:30 Phase plates are carefully craft films placed into the beam path

15:35 that spatially modify the phase of the electron wave,

15:38 resulting in an improved image contrast.

15:40 They work by changing the interference condition

15:42 between the scattered and unscattered electrons.

15:45 Monochromators are modules placed just after the electron source.

15:50 They reduce the beam energy spread (at the cost of total current),

15:53 thus improving its temporal coherency.

15:55 Let's now talk about Thermo Fish itself, as the sponsored section is coming up.

16:00 Luckily, I can freely say whatever I want, as long as it's not confidential.

16:05 Which is a nice part of the company culture, at least in Brno.

16:10 Overall, I personally like working here.

16:12 Mostly for the friendly, smart and helpful bunch

16:15 of technical people,

16:15 with whom I interact and who usually have a good sense of humor, too.

16:20 I get to apply physics, maths and engineering to solve problems,

16:23 which I find enjoyable and satisfying.

16:25 To me, it is also important to have a meaningful work.

16:28 And building and improving scientific instruments

16:30 that help deepen our knowledge of the world

16:32 and push our civilization forward qualifies as such.

16:35 However, let's also not sugercoat it- the company

16:37 is a large corporation and that inevitably

16:40 comes with some baggage.

16:42 Organizational and process inefficiencies,

16:45 unnecessary meetings, weird TLAs (Three-Letter Acronyms) etc.

16:50 Still, I'd say it's a good place to work at, all things considered.

16:55 There's a sense of fairness among people and a good atmosphere.

17:00 Let's now dive back into the technical stuff.

17:05 Time is running short, so let's take a quick look at SEM's as well.

17:10 The SEM uses a lower accelerating voltage, typically between 100 V and 60 000 V.

17:15 This is because the electrons don't need to pass through the specimen.

17:20 They only need to interact with it,

17:22 point by point, to create enough signal to collect.

17:25 The primary electrons from the beam are typically absorbed by the specimen.

17:30 That's why SEM specimens often need to be coated with a thin conductive layer,

17:35 so that they don't electrically charge and deform the incoming beam.

17:40 SEM's are often also paired with a Focused Ion Beam (FIB) column in one machine.

17:45 The ion beam is controlled in a similar way to electrons

17:50 and is used for ultra-precise pattern milling and depositions.

17:55 Back to the video: Different signals come from different depths,

17:58 1-5 nm for Auger electrons,

18:00 5-50 nm for Secondary electrons, 100-300 nm for Back Scattered electrons,

18:05 and up to 1000-3000 nm for the X-ray and Cathodoluminescence signals.

18:10 The most common signals in an SEM are secondary and back scattered electrons,

18:14 and x-rays.

18:15 STEM and TEM modes have a similar resolution,

18:17 but due to a different principle of image formation,

18:20 they provide different contrasts.

18:21 The details of STEM images are more directly interpretable.

18:25 However, the STEM mode is also slightly more difficult to

18:27 set up and align than the TEM mode.

18:30 The detectors for this mode are not finely pixelated like cameras,

18:35 but usually have an annular shape,

18:37 either with one or several large detection segments.

18:40 They can also use either direct

18:42 or indirect electron detection (scintillation layer),

18:45 but unlike cameras,

18:46 they are typically radiation hard and with much quicker readout times.

18:50 When scanning,

18:50 the beam dwells on each specimen point on the order of micro- to milliseconds.

18:55 And that's about it for our quick

18:57 introduction and overview of Electron Microscopes.

19:00 I hope you enjoyed the video and learned something new.

19:05 For further questions or physics and engineering chat and discussions,

19:08 feel free to contact me at:

19:10 Jan Jíša- jan.jisa@thermofisher.com May the Lorentz Force be with you ;o)