You are on page 1of 84

Copyright

by

Matthias Reuss

2005
Optical trapping and manipulation of nanoparticles in a
Gaussian standing wave

by

Matthias Reuss

Thesis

Presented to the Faculty of the Graduate School of

The University of Texas at Austin

in Partial Fulfillment

of the Requirements

for the Degree of

Master of Arts

The University of Texas at Austin

December 2005
Optical trapping and manipulation of nanoparticles in a
Gaussian standing wave

Approved by
Supervising Committee:
For Bianca
Acknowledgments

I would like to express particular appreciation to Prof. Dr. Ernst-Ludwig Florin for
giving me this great opportunity to do research. His tireless efforts, guidance and
support added considerably to my research experience.
A very special thanks goes out to Dr. Saša Jonáš, who was always willing to
contribute with fruitful discussions, suggestions and comments on the spot.
Many thanks also to Chieze Ibeneche for proof-reading and to the rest of the
biophysics group, including her, for their support and the great time I had here.
This holds for many more people at CNLD, UT and in Austin; to mention them all
would go beyond the scope of what I can do here.
I also acknowledge the great work of the people in the machine shop, espe-
cially Jack Clifford, and the help of the CNLD administratives, Olga Vera and Elena
Simmons.

v
Further thanks go to Prof. Dr. Manfred Böhm, Prof. Dr. Hansheinrich
Langhoff and Prof. Dr. Manfred Fink for making the exchange program possible,
and to Prof. Dr. Wolfgang Frey for agreeing to be the co-reader for this thesis.
Last but not least, I would also like to thank my parents and my girlfriend
Bianca for their support and understanding while I was away.

Matthias Reuss

The University of Texas at Austin


December 2005

vi
Optical trapping and manipulation of nanoparticles in a
Gaussian standing wave

Matthias Reuss, M.A.


The University of Texas at Austin, 2005

Supervisor: Ernst-Ludwig Florin

This thesis demonstrates optical trapping and manipulation of unprecedentedly


small dielectric nanoparticles. 40 nm-polystyrene beads were stably trapped in a
Gaussian standing wave, particles can be moved and sorted according to their prop-
erties. Furthermore, the ratio of axial to lateral trap confinement can be tuned, so
that trapped particles are either free to move independently in the plane of obser-
vation or are tightly confined, allowing the study of dynamic interactions of several
trapped objects.

vii
Contents

Acknowledgments v

Abstract vii

List of Tables x

List of Figures xi

Chapter 1 Introduction 1
1.1 Understanding radiation forces – a brief history . . . . . . . . . . . . 1
1.2 Motivation and aim of this thesis . . . . . . . . . . . . . . . . . . . . 2

Chapter 2 Theory of optical trapping 6


2.1 Forces on particles in an electromagnetic field . . . . . . . . . . . . . 6
2.1.1 Origin of description of radiation forces . . . . . . . . . . . . 6
2.1.2 Limiting cases . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Using light for trapping – the single beam trap . . . . . . . . . . . . 10
2.3 Optical trapping in a Gaussian standing wave . . . . . . . . . . . . . 12
2.4 Other forces exerted on a trapped particle . . . . . . . . . . . . . . . 15

Chapter 3 Implementation of the standing wave trap 18


3.1 General assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

viii
3.2 Observation of trapped objects . . . . . . . . . . . . . . . . . . . . . 23
3.2.1 Wide-field fluorescence and single particle tracking . . . . . . 24
3.2.2 Confocal laser-scanning fluorescence microscopy . . . . . . . . 27
3.3 Mechanical stability of the setup . . . . . . . . . . . . . . . . . . . . 28

Chapter 4 Experiments 31
4.1 Independent trapping and observation . . . . . . . . . . . . . . . . . 31
4.2 Stiffness of individual antinodes . . . . . . . . . . . . . . . . . . . . . 35
4.3 Selective storage of objects in antinodes . . . . . . . . . . . . . . . . 45
4.3.1 Transport of a particle between antinodes . . . . . . . . . . . 45
4.3.2 Sorting of particles according to size and refractive index . . 49
4.4 Trapping with low numerical aperture . . . . . . . . . . . . . . . . . 52
4.5 Optical trapping of 40-nanometer polystyrene particles . . . . . . . . 55

Chapter 5 Summary and outlook 59

Appendix A Formal description of intensity and forces in a Gaussian


standing wave 62

Appendix B Correction of low-pass filtered position data 64

Bibliography 67

Vita 72

ix
List of Tables

4.1 Standing wave trapping with low and high numerical aperture . . . . 53

x
List of Figures

2.1 Intensity distribution in a focused beam and a standing wave . . . . 13


2.2 Intensity gradient along the optical axis for a focused beam and a
standing wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3.1 Setup for a standing wave optical trap . . . . . . . . . . . . . . . . . 19


3.2 Photograph of the setup . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 Sample chamber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.4 Effect of defocusing on a point-like object . . . . . . . . . . . . . . . 25
3.5 Particle tracking calibration . . . . . . . . . . . . . . . . . . . . . . . 26
3.6 Schematic drawing of a confocal microscope . . . . . . . . . . . . . . 27
3.7 Mechanical stability . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4.1 Focusing into different antinodes . . . . . . . . . . . . . . . . . . . . 32


4.2 Imaging particle distributions with confocal laser scanning microscopy 34
4.3 Positions of a trapped particle over time and histogram of displacements 36
4.4 Normalized standing wave trap stiffness for NA=0.7 . . . . . . . . . 37
4.5 Normalized standing wave trap stiffness for NA=1.3 . . . . . . . . . 40
4.6 Simulated and experimental point-spread-function . . . . . . . . . . 42
4.7 Trapping on rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.8 Energy landscape for different beam waist positions . . . . . . . . . 46

xi
4.9 Transport between antinodes . . . . . . . . . . . . . . . . . . . . . . 48
4.10 Hopping towards the beam waist . . . . . . . . . . . . . . . . . . . . 49
4.11 Energy landscape for particles of different sizes . . . . . . . . . . . . 49
4.12 Particle sorting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.13 Histograms for low and high numerical aperture . . . . . . . . . . . . 52
4.14 Two particles trapped using different numerical apertures . . . . . . 54
4.15 Trapping and manipulation of a 40-nanometer-bead . . . . . . . . . 56
4.16 Assembly of 40-nanometer-beads . . . . . . . . . . . . . . . . . . . . 57

xii
Chapter 1

Introduction

1.1 Understanding radiation forces – a brief history

Studying the force exerted by electromagnetic radiation on matter is an important


part of modern physics. However, the history of this field can be traced back to
1619 when the German astronomer Johannes Kepler suggested in his treatise De
Cometis, that solar light pressure is responsible for comet tails pointing away from
the sun. During this time, it was thought that astronomy was the only field where
light pressure is a factor, since the forces that could be generated in the laboratory
using light were very small at that time and were therefore overshadowed by other
forces, such as forces stemming from Brownian motion. This explains why it took
several centuries for any major advance. It wasn’t until Scottish theorist James Clerk
Maxwell presented his theoretical work A Dynamical Theory of the Electromagnetic
Field, that the foundation for explaining light pressure was fully laid. Published
in 1865, it opened the door for physicists such as P. N. Lebedev who managed to
overcome the experimental difficulties and precisely measured the force exerted by
light on a flat metallic sheet.
However, these forces were still too small for anyone to make practical use

1
of them. This changed with the invention of the laser by Theodore H. Maiman in
1960 at Hughes Research Laboratories in Malibu, California [22]. The coherence
and high spectral purity of laser light enables focusing into a spot comparable in
size to the wavelength of light, making it possible to deliver large amounts of energy
into a small volume. This way, light can exert forces on dielectric objects located in
the focal spot that can greatly exceed gravitational or thermal forces.
In the 1970s Arthur Ashkin observed small spheres in water being pushed
against a glass wall by radiation pressure exerted by a laser beam [1]. He replaced
the glass wall by using the force from another laser beam pointing in the opposite
direction and was able to hold small particles in place using nothing but light.
This is considered the beginning of optical trapping. A few years later Ashkin had
developed the single beam trap or optical tweezer, a device capable of trapping living
biological objects such as cells using only one laser beam [4, 3]. It is from this point
that we mark the beginning of the huge success of optical trapping [33, 30, 8, 31, 25],
a technique today indispensable for biologists and one that also led to a Nobel price
for Steven Chu, a student of Ashkin, for optical trapping of atoms [11].

1.2 Motivation and aim of this thesis

Biological cells have been thoroughly investigated under the microscope since Robert
Hooke discovered the “porous” structure of cork in 1665, which in fact was formed
by the walls of cork cells. Already, around the 1900’s the available microscopes had
reached a resolution close to the diffraction limit and many mysteries of the cell had
been unraveled. However, it soon became clear, that in order to fully understand
the biological principles of life it would require investigations on a molecular level,
since cell activity is ultimately taking place in the realm of individual molecules.
For example, cell movement is accomplished by restructuring the cytoskeleton that
gives the cell its shape and stiffness, that is, by changing the distribution of polymer

2
filaments inside the cell. This is one example of changes on a molecular level that
have a great impact on much bigger length scales. Furthermore, the digestion of
food, cell division and many other cellular functions can only be fully understood
when dealing with the particular molecules involved. To date, the problem of inves-
tigating individual molecules inside a cell has been successfully addressed by labeling
molecules with fluorescent dyes. As an example, the motion of a single molecule of
RNA inside a bacterium can be tracked and quantified [14]. Unfortunately, even the
simplest cells contain thousands of different types of proteins, carbohydrates, lipids
and other biological material that can interact with the target molecule in an over-
whelming multitude of ways. This illustrates the need to investigate biomolecules
outside the cell where much more control over external parameters can be exercised.
The natural way to examine the behavior of molecules is by isolating the
particular molecules of interest from cells and then study how they interact and
react with each other using chemical methods. In fact, biochemistry has revealed a
great deal about cell functionality. However, biochemical techniques typically pro-
vide a mere average of what is happening in a test tube based on the observation of
statistical ensembles of molecules. Still, a protein might be in different conforma-
tional states during its action, with some of these states occurring much less likely
than others. These rare states will not be observed in a measurement that only
detects averages. It can be seen that this method easily suppresses the recognition
of states that are not present most abundantly – states that might be essential to
the function of the protein, nonetheless. In contrast to analyzing averages of large
ensembles, another approach is to manipulate single molecules in an effort to under-
stand the way all of these smallest units of the cell work, one by one. Once a single
molecule is fully understood, a second molecule can be added in order to explore the
interaction dynamics between both of these molecules; this method can be contin-
ued until eventually one can begin to study more complex arrangements. For this,

3
it is evidently necessary to actively handle single molecules at room temperature.
The challenge is to keep the molecules inside a defined volume, where they can be
observed over a long time, while disturbing them as little as possible. Mechanical
approaches for immobilization have been successful, for example, molecules can be
attached to a much bigger structure, which in turn can be trapped optically [9, 39],
or to a surface [12, 16, 20, 28]. These techniques, although they provide important
results, have several drawbacks and do not work for every type of specimen. During
adsorption, the functionality of molecules can be altered significantly, for example,
proteins are likely to undergo conformational changes when fixed to a substrate
and not all binding sites might be accessible. In general, everything that might be
used to mechanically clamp molecules cannot be much smaller than the molecules
themselves and will therefore seriously interfere with the specimen. Moreover, the
presence of an additional body will contribute to or even hide the cooperation of
several molecules, making investigation of interaction dynamics impossible. Optical
tweezers have the ability to hold more than one object contactlessly, but they are far
from being able to trap single molecules, and again, looking at dynamic interaction
of more than one molecule is not practical if they are attached to beads that are
orders of magnitude larger. The need for a system that is capable of handling single
biological molecules in aqueous solution, their natural environment, is evident. Since
this non-mechanical assay can be carried out using optical forces, a method needs
to be developed to extend the working range of optical tweezers down to objects
that are only a few tens of nanometers in size.
The aim of this thesis is to show that optical confinement of such objects over
more than an hour is indeed possible using a standing wave optical trap (SWT). This
type of trap makes use of the interference structure generated by two counterprop-
agating laser beams and it can in principle be built from a single beam trap (SBT)
with only one additional component. The principle was developed at the Institute

4
of Scientific Instruments at Brno in the Czech Republic by a team led by Pavel
Zemanek [42, 19, 41, 17] and simulations done by them [41, 40] suggest that trap-
ping of tiny objects is possible in a SWT. In this thesis, a setup is presented that
enhances the principle of the standing wave trap to allow greater control over the
trap itself as well as trapped objects, and it is demonstrated that the observation
of trapped specimen can be improved at the same time. Furthermore, experimental
evidence is given that confirms trapping of unprecedentedly small particles.

5
Chapter 2

Theory of optical trapping

2.1 Forces on particles in an electromagnetic field

In order to understand how trapping of material objects in a focused beam can be


extended down to smaller specimen, it is necessary to understand the mechanism
behind the generation of optical forces. In this chapter, the origin of these forces will
be discussed, and equations will be given that allow for the quantitative computation
of the force that an electromagnetic field exerts on small dielectric specimen. On
this basis, the mechanism of optical tweezers can be understood, and it will become
clear how this principle can be improved.

2.1.1 Origin of description of radiation forces

It is obvious from Lorentz’ law that charged particles are affected by electromagnetic
radiation, but this also implies that a light field can alter the distribution of charges
in a dielectric medium. It is this response of matter to external fields, described
by the permittivity ε and the permeability µ, that is finally responsible for forces
exerted by electromagnetic radiation on matter.
Electromagnetic fields can be fully described by a set of four equations named

6
Maxwell’s equations after their discoverer:

~ = 0
∇B (2.1)
~ = ρ
∇D (2.2)
~
~ = − ∂B
∇×E (2.3)
∂t
~
~ = J~ + ∂ D .
∇×H (2.4)
∂t

~ and H
B ~ are the magnetic induction and field intensity, respectively, D
~ is the elec-
~ is the electric field intensity. ρ denotes the free charge
tric displacement, and E
density, and J~ the free current density. These equations, together with the principle
of conservation of momentum which holds for the joint system of dielectric and elec-
tromagnetic fields, yield the force acting on any volume V occupied by a dielectric
body:
~− 1 d
I Z
F~ = T̄ dA ~
SdV, (2.5)
A c2 dt V

~ is the Poynting vector S


where T̄ is the Maxwell stress tensor and S ~=E
~ × H.
~ The
~ enclosing V , in
first integral is the flux of momentum passing through a surface A
~ while the second term is the change in
other words, it denotes the force across A,
momentum of the field itself.
For the stationary fields dealt with here, it follows immediately that the
second term can be dropped, and that the force is given by the time average of
the integral over the dot product of the Maxwell tensor and the surface normal,
evaluated over the surface enclosing the volume:

I
F~ = h ~
T̄ dAi. (2.6)
A

Ref. [7] concludes that the form to use for the Maxwell tensor, in the case of
a homogeneous spherical particle immersed in a dielectric medium illuminated by a

7
stationary monochromatic electromagnetic beam, is the so-called Minkowski form:

1
T̄ij = εEi Ej + µHi Hj − (εEi Ei + µHi Hi ) δij . (2.7)
2

According to eqs. 2.6 and 2.7 it is possible to find the force acting on an
irradiated object if the components of the electric and magnetic fields are known
on any surface enclosing the object. Unfortunately, as material interacts with an
electromagnetic field, multipoles are induced in it, which give rise to secondary
radiation. This means that the field that the particle is subject to is altered by
the particle itself through scattering. The field that has to be considered does not
therefore only consist of the incident beam but also of scattered fields, which depend
strongly on the properties of the scatterer.
For a homogeneous spherical object, the components of the scattering field
can still be evaluated as done in [6], but to compute the field in the vicinity of
an arbitrarily shaped object requires some effort. Fortunately, in order to explain
optical trapping qualitatively, the scattered field, which is much weaker than the
incident beam for most cases, can be discounted. Moreover, the expressions for
optical forces get significantly simpler if the size of the object is either very large or
very small.

2.1.2 Limiting cases

Large spherical homogeneous objects

If the size of the object is much larger than the wavelength of the incident radiation,
the forces acting on the object can be readily calculated using a simple ray optics
approach. Incident light is subject to refraction at the interface of the sphere and
the surrounding medium and it follows from conservation of momentum that a force
is acting on the particle [2].

8
This gives a very simple qualitative explanation for optical trapping, but
since all the particles that are being dealt with here do not fulfill the condition of
being sufficiently large, the ray optics regime will not be discussed further.

Rayleigh regime

If the size of the object in question is much smaller than the wavelength of the
irradiating electromagnetic field, that is, in other words, when the field is uniform
across the object, then the particle is acting as a single point dipole, which can be
treated using quasi-electrostatics. In this case, the force acting on the object can be
decomposed into a scattering force due to absorption and re-radiation of light

!2
128π 5 a6 m2 − 1 nm σnm
F~scat = I0~z = I0~z, (2.8)
3λ4 m2 + 2 c c

and a gradient force


!
2πa3 m2 − 1 2πα
F~grad = 2
∇I0 = 2 ∇I0 , (2.9)
c m +2 nm c

stemming from the interaction of the induced dipole with the inhomogeneous ex-
ternal electric field [23]. I0 is the intensity of incident light, a is the radius of the
trapped sphere, nm is the index of refraction of the surrounding medium, c is the
speed of light in vacuum, m is the relative index of refraction between the particle
and the surrounding medium, λ is the wavelength of the incident beam, σ is the
scattering cross section, α is the polarizability of the object and ~z points in the
direction of beam propagation.
These simple equations are sufficient to understand the mechanisms of optical
trapping.

9
2.2 Using light for trapping – the single beam trap

As discussed in the previous section, light is in principle able to exert a force on


objects. A closer look at eqn. 2.9 reveals that the gradient force F~grad depends on
the intensity gradient of the irradiating beam, where the sign is given by the term
m2 −1
m2 +2
. Therefore, the gradient force will point towards regions of higher intensity if
the relative index of refraction m of the particle to be trapped is higher than one,
and it will point away from high intensity if m < 1. In the case m = 1, no gradient
force will be present on the object. In practice, for the overwhelming majority of
objects of interest, the relative index of refraction is greater than one.
It is evident that the gradient force can be utilized for optical trapping if an
appropriate distribution of light in the vicinity of the object to be trapped can be
generated, that is one with strong field non-uniformities. The simplest way is to use
a parallel laser beam, which naturally has a Gaussian-like distribution in the lateral
direction, with the intensity being high in the center and dropping off towards the
periphery. This would force an object with m > 1 towards the center of the beam.
However, the particle would still be free to move along the optical axis, since there
is no noticeable intensity gradient in this direction. Focusing of the laser beam with
a lens overcomes this problem, since a region is created, where the intensity is high
in the focus, falling off in all directions. Therefore, if m > 1, the gradient force
always points towards one point, trying to restore a minimal energy configuration
where the particle is located in the center of the trap. The extent r of this region
from its intensity maximum to the first minimum is given solely by the numerical
aperture (NA) of the lens [26] and the wavelength of the trapping beam λ where

λ
rlateral = 0.61 (2.10)
NA
λnm
raxial = 2 . (2.11)
NA2

10
It is typically on the order of 400 nm in the lateral direction and 1200 nm axially for
the most tightly focused beams that can be produced. This means that the intensity
varies from close to zero to its maximum over a distance of several hundred nanome-
ters in the lateral direction. Provided that the maximal intensity is sufficiently high
for the given object size a and relative index of refraction m, this intensity gradient
can be steep enough to ensure stable trapping. Unfortunately, at the same time, the
gradient in the axial direction is significantly smaller due to the greater extension
of the focal volume in this direction, rendering the axial gradient force smaller by a
factor of about three.
The other major force to be considered, the scattering force F~scat = σnm
c I0 ~
z
(eqn. 2.8) depends on the intensity of the incoming beam and is always pointing in
the direction of light propagation, therefore trying to push the particle away from
the focal plane and out of the trap. In order to achieve stable trapping, the gradient
force has to be high enough to overcome the contribution of the scattering force,
preventing loss of the object along the optical axis. This is only achievable if a
high-quality microscope objective lens with a high numerical aperture is used for
focusing the trapping beam. Only such a lens is capable of generating adequate
intensity gradients to prevail over the scattering force. This is the principle of the
single beam trap.
Since, according to eq. 2.9, the gradient force drops with the volume, smaller
objects can only be trapped if the intensity gradient is increased, especially along
the optical axis, the direction in which the scattering force acts. The easiest way
to achieve this is by increasing the maximum intensity, however, there are limits
– especially for biological specimen, which suffer from radiation damage. For this
reason, it is a problem to stably confine small objects axially without using very high
laser powers. In order to trap these particles, it is necessary to have the intensity
variations along the optical axis happen over a much smaller distance, that way, the

11
gradient will be higher for a given power. Unfortunately, eq. 2.11 states that the
length scale over which the intensity varies axially in a single beam trap is limited
by the properties of the objective lens, namely its numerical aperture that can not
be made infinitely high1 . It follows, that in order to enhance the intensity gradient,
a different approach has to be considered. A scheme for achieving higher gradients
without the need for more power is discussed in the following section.

2.3 Optical trapping in a Gaussian standing wave

The setup suggested in [42] is only a slight modification of the single beam prin-
ciple. A surface that is reflectively coated for the trapping wavelength is placed into
the beam path close to the beam waist of a Gaussian beam focused by a high-NA
objective lens. Superposition of the incident and reflected wave creates a stand-
ing wave in which the parasitic scattering force is strongly suppressed since two
counterpropagating beams are present. Moreover, axial variations in intensity occur
over length scales of λ/4, consequently increasing the axial intensity gradient signif-
icantly. Both give rise to an improved trap, where the net axial force on particles is
strongly enhanced for the given laser power, relative index of refraction and particle
size.
The plot on the left in fig. 2.1, displays the intensity distribution in a single
beam trap generated by focusing a beam. The elongation of the focal spot along the
optical axis (vertical) is clearly visible. The right plot shows the intensity distribu-
tion for the same beam after a reflective surface has been inserted at the location
of the beam waist. It is obvious that the intensity changes over a much smaller
distance axially.
Fig. 2.2 shows a plot of the intensity gradient along the optical axis for
1
The alternative, using a trapping laser with a shorter wavelength, also has constraints. See
section 3.1.

12
Figure 2.1: Intensity distribution for a focused beam as in a single beam trap (left, wave-
length in medium: λ = 800 nm, beam waist radius: w0 = 350 nm) and after a reflective
surface has been inserted at the location of the beam waist (right, same beam parame-
ters, phase shift during reflection: 3π/2), forming a standing wave trap. Both plots were
computed using a fifth-order corrected Gaussian beam description [5].

the same configurations as in fig. 2.1. Again, a single beam trap (dotted line)
is compared to the same beam focused onto a reflective surface, which generated a
standing wave (solid line). Obviously, the smaller structures along the optical axis in
a standing wave translate into an intensity gradient that is considerably higher than
what can be achieved by simply focusing a parallel beam. Equilibrium positions
for a particle in such a light field occur at positions where the intensity gradient is
zero with a negative slope. The single beam trap provides only one such position,
whereas the standing wave exhibits several with steeper slopes. This translates into
several traps with better confinement since already a small displacement brings the
object into high gradient and therefore high force regions. In practice, intensity

13
Figure 2.2: Intensity gradient along the optical axis for a focused beam (dotted line) and
a standing wave (solid line), calculated using a fifth-order corrected Gaussian beam [5]. The
beam waist is located at z = 0. The beam is incident from the right, the reflective surface
is located at the beam waist position, therefore, no standing wave exists for z < 0. Same
parameters as in fig. 2.1.

gradients of an empty trap do not directly translate into forces, because the trapped
object itself changes the intensity distribution through scattering, but it remains
true that the forces on a particle in a standing wave are orders of magnitude higher.
In short, whereas simple focusing has the drawback that lateral forces are
always higher than axial forces, in a standing wave this ratio can be improved and
even reversed. This is still the case even when a low-NA lens is used for focusing
the trapping beam in a SWT. Since the length scale over which the intensity varies
axially is only given by the wavelength of the trapping beam, axial forces will still
be high. The lateral forces will be smaller, due to the bigger beam waist, but forces
in all directions are still sufficient to trap small objects, because the scattering force
is negligible.
From now on the equilibrium positions of a particle in a standing wave will be

14
referred to as antinodes of the standing wave or simply traps. The term “standing
wave trap” is reserved for the overall structure consisting of several antinodes. The
antinode that is closest to the surface will be named first antinode or first trap;
the next one will be named second antinode (trap) and so on. “Higher” antinodes
(traps) are formed farther away from the surface. Furthermore, the direction along
the optical axis will be denoted as z, while the lateral directions (perpendicular to
the optical axis) are x and y, where y is parallel to the polarization of the trapping
beam.
Trapping of 100 nm-radius polystyrene beads in a single beam trap is a stan-
dard procedure carried out daily in many laboratories around the world. A. Jonáš
showed that, in theory, the maximum trapping force on these beads in a single beam
trap is comparable to the force that a standing wave trap can exert on polystyrene
beads with a radius of only a few nanometers under otherwise the same conditions
[18]. The refractive index of most proteins is comparable to that of polystyrene
[10, 24], hence it can be expected that standing wave optical trapping of micro-
tubules, which have a diameter of approximately 25 nm, poses no problem and that
trapping of proteins, which have sizes in the order of a few tens of nanometers, is
within reach.

2.4 Other forces exerted on a trapped particle

Additional to gradient and scattering forces, there are also thermal and viscous
forces acting on a dissolved particle in an optical trap. The equation of motion of
such a particle can be written as

p
mẍ(t) + γ ẋ(t) + κx(t) = ξ(t) 2kB T γ. (2.12)

15
using a simplified approximation of a particle in a harmonic potential [27] with
an external force. The terms on the left hand side represent the inertial, viscous
and elastic forces respectively, while the term on the right hand side represents the
thermal force. x(t) denotes the particle’s trajectory whereas m is its mass. γ is the
coefficient of friction and κ the stiffness of the trapping potential which is assumed
to be harmonic. In air or vacuum, that is for small γ, and for T = 0, the oscillator
would exhibit a well-defined resonant frequency. If the particle is submerged in a
fluid, which is always the case for biological specimen, the coefficient of friction can
be approximated by Stokes’ drag γ = 6πrη, where r is the particle’s radius and η
is the viscosity of the surrounding fluid. In this case, viscous damping cannot be
neglected anymore, in fact, the result is an overdamped oscillator which is governed
by drag and optical forces and where inertial forces play only a minor role and can be
dropped from the equation. Additionally, since experiments will always take place
at a finite temperature T , a trapped particle will be subject to impacts from the
surrounding water molecules which move due to their thermal energy. According to
the Einstein-Ornstein-Uhlenbeck theory of Brownian motion, this can be accounted
for by the term on the right hand side of eq. 2.12. It represents Brownian motion
through a random Gaussian process with mean zero;

hξ(t)i = 0 (2.13)

hξ(t)ξ(t0 )i = δ(t − t0 ) (2.14)

for all times t, t0 [27].


The impacts of surrounding particles give rise to thermal fluctuations, result-
ing in a mean square displacement h∆x2 i of the particle from the potential minimum
of the trap, which can be calculated using the equipartition theorem

1 1
κh∆x2 i = kB T. (2.15)
2 2

16
This means that the stiffness of an optical trap can be measured easily by observing
the degree of thermal fluctuations.
Boltzmann’s equation states that the relative probability p, to encounter a
system in a certain state with energy E, is a function of E itself. In the case of an
optical trap, the probability to find a trapped particle with a certain displacement
from its equilibrium position depends on the potential Ed of the particle at its
displaced position (x, y, z) relative to the potential at the equilibrium position E:

Ed (x, y, z) − E
 
p(x, y, z) = C · exp − , (2.16)
kB T

where kB is Boltzmann’s constant, T is the temperature of the surrounding fluid.


Therefore, by measuring the probability for all displacements to occur, one can solve
eq. 2.16 for Ed and get the optical potential for all displacements that are accessible
by thermal fluctuations.

17
Chapter 3

Implementation of the standing


wave trap

The theoretical principles of optical trapping have been examined in the previous
chapter, and it has been stated that a standing wave trap can in principle be used
to confine smaller objects with less laser power than in a single beam trap. The
following chapter will deal with the practical realization of such an enhanced optical
trap, carried out by the author.

3.1 General assembly

Figure 3.1 illustrates the schematic assembly of the setup for standing wave optical
trapping while fig. 3.2 shows an actual photograph. The setup is built around a Zeiss
Axiovert 100 inverted microscope with a confocal laser scanning unit LSM 410 (Carl
Zeiss, Germany), allowing for several methods of imaging, the two most important
being wide-field fluorescence and confocal laser scanning. The trapping laser is
an ultra-stable 1064 nm-Nd:YAG from CrystaLaser (Reno, USA), with a maximum
output power of 1 W and a 1/e2 -beam diameter of 0.45 mm. The wavelength was

18
Fiber coupling
Dichroic mirror
Laser
Fiber

Beam expander

Trapping lens

PI-Foc
Sample chamber

Sample holder

Imaging lens

Reflective slide

Confocal microscope

Widefield-fluorescence illumination
Removable mirror

Dichroic mirror

Reflective slide

Hot mirror

Eyepiece / Camera

Figure 3.1: Setup for a standing wave optical trap. The removable mirror is used to switch
between confocal and widefield mode. The red and green beams are mutually exclusive, the
valid beam depending on whether the removable mirror is taken out or not.

19
chosen because water and biological specimen both exhibit low absorption in this
region of the spectrum, thus heating effects can be largely minimized. The beam is
expanded five times in order to match the numerical aperture (0.12) of the optical
fiber it is coupled into. This polarization-maintaining single-mode fiber (Schäfter &
Kirchhoff, Germany) delivers the beam to the central part of the setup, the trapping
assembly mounted on top of the microscope. The parallel beam, now expanded to
a diameter of 6.0 ± 0.1 mm, is coupled into the microscope from the condenser side
via a dichroic mirror that also allows transmission microscopy to be carried out.
Focusing of the trapping beam is achieved by a Zeiss Plan-Neofluar oil immersion
objective lens (100x/NA 0.7-1.3) with variable numerical aperture to allow for tuning
of the lateral potential of the trap. The back aperture of the objective (diameter
4.3 mm) is overfilled by 140% in order to make full use of the available numerical
aperture. The trapping lens is supported by a PI-Foc piezo actuator (PI, Germany)
with feedback-controlled position stabilization, this way, it can be shifted axially to
place the beam waist at the desired position inside the sample chamber with high
precision. Fiber collimator, dichroic mirror and PI-Foc, together with the trapping
lens, are mounted on a common aluminum rack, resting on the microscope sample
platform. This mount can be moved laterally using a xy-stage and tilted using fine
adjustment screws, permitting alignment of the trapping lens with the optical axis
of the microscope.
The objective lens used for observation is a Zeiss Plan-Neofluar 100x/NA
1.3 Ph3 oil immersion. Two reflective slides and a hot mirror placed at different
positions in the optical path further attenuate the remaining infrared radiation
passing through the sample chamber to an amount that cannot be detected in the
eyepieces or on the camera. The first of these reflective slides is tilted against the
optical axis to prevent any backreflections of the trapping beam into the sample
chamber.

20
Figure 3.2: Photograph of the setup for the standing wave trap. Visible are parts of
the microscope (A), the fiber (B) delivering the trapping beam, which is coupled into the
trapping lens (D) by a dichroic mirror (C). The three latter parts are mounted on the top
aluminum frame, which can be positioned by the xy-stage visible on the right. The black
box behind the trapping lens is the PI-Foc, (E) is the sample mount.

21
Trapping lens
Immersion oil
Spacer beads Trapping beam
Vacuum grease Top coverslip

Sample Holder Bottom coverslip


w/ reflective coating
Nail polish for IR
Solution with specimen

Immersion oil Imaging lens

Figure 3.3: Schematic assembly of the sample chamber.

The sample itself is mounted on an aluminum holder that can be tilted using
fine adjustment screws to render the sample chamber perpendicular to the optical
axis of the microscope. A 24 × 32 mm coverslip, reflectively coated for IR with a
multilayer TiO2 /SiO2 coating1 , but transparent for visible light, is glued into the
aluminum frame with nail polish and forms the bottom of the sample chamber,
which is covered by a 18 × 18 mm coverslip and sealed with vacuum grease (see
fig.3.3). Latex beads with a diameter of 5 µm (Duke Scientific, USA) are distributed
inside the sample chamber and act as spacers to keep the top and bottom coverslips
separate in a well defined manner. The specimen is inside the sample chamber in
a solution whose composition may vary, depending on the specimen itself and the
application.
1
A complex layer structure ensures that the reflectivity is high for a large range of angles of
incidence, and that is high only for the wavelength of the trapping beam and not fractions or
multiples of it.

22
3.2 Observation of trapped objects

As mentioned above, a wide range of microscopical techniques can be utilized for


the observation of trapped particles. The two that are most widely used here, are
discussed in the following paragraphs. Both of them require specimen labeled with
dyes that can be excited by a certain wavelength in the visible spectrum and emit
light at a slightly higher wavelength through fluorescence. This is not a serious
constraint, since fluorescence microscopy is widely used in biology, where labeled
specimen such as latex beads2 are fairly easy to obtain.
In a typical optical trap configuration imaging and trapping beams share
a common objective lens. This is feasible for a single beam trap, since there is
only one stable trapping position which is guaranteed to be close to the focal plane
of the imaging lens at any time when imaging and trapping lens are the same.
With the recent evolution of trapping methods, more complex optical potential
landscapes can be generated, using time sharing of the trapping beam between
more than one SBT [38, 37] or holographic techniques [13, 21]. In these cases, energy
landscapes can extend spatially beyond the depth of view of the imaging objective
and more elaborate observation methods have to be used, because refocusing would
perturb the trap if the same objective lens is used for imaging and trapping. This
consequently leads to separation of imaging from trapping, pioneered by Visscher
et al. [35, 36], who realized that the integration of optical traps into a confocal
microscope allows imaging of extended objects in the trap without having to move
the trapping lens. Later, Vossen et al. combined time shared traps generated using
acousto-optic deflectors with confocal microscopy [38] .
Especially in a standing wave there are several equilibrium positions for which
stable trapping is achieved. These traps are all located on the optical axis, and
2
Latex beads can serve as a model system for proteins, since the refractive index is nearly the
same for both materials [10].

23
stable trapping is possible over tens of antinodes, which, for a 1064 nm-trapping
beam, extend over several microns. Here, refocusing into higher antinodes would
inevitably lead to a heavily disturbed trap if the optical paths for imaging and
trapping are not separated. With observation that is independent from trapping,
all compartments of a standing wave trap can be analyzed without disturbing the
trap itself.

3.2.1 Wide-field fluorescence and single particle tracking

When only one object is present in the trap, wide-field fluorescence is the imag-
ing tool of choice. Excitation source is a standard fluorescence lamp, and emitted
fluorescence is observed either via a monochrome camera (Basler AG, Germany) in
connection with a PC running LabView (National Instruments, Texas), or by simply
looking directly through the microscope eyepieces.
In combination with single particle tracking methods, the spatial position of a
specimen can be determined with nanometer precision in all three dimensions with a
temporal resolution limited only by the image acquisition system. A point-like object
in the focal plane observed through a microscope appears as an Airy-diffraction
pattern with the central disk having a diameter3 of approximately 1.22λ/NA, where
λ is the observation wavelength and NA is the numerical aperture of the objective
lens used. As the object goes out of focus, this diameter grows, until, at a certain
amount of defocusing, it changes into a complex structure of rings that is determined
by the characteristics of the objective lens. It has been found [32], that the radius
of the outer rings grows linearly with displacement from the focal plane. Thus, the
ring size can be utilized to determine the z-position of an object with respect to the
focal plane. The center of the rings yields the lateral (x, y) position of the object.

3
The distance between the two first minima.

24
Figure 3.4: Left column: a 200 nm-bead (red) is located in the focal plane (dotted line) of
the imaging lens (bottom in leftmost column). It is imaged by the objective as an Airy-disk
(top). As the particle goes out of focus, rings are formed that grow with the displacement
from the focal plane (central and right columns). Amount of defocus was 0 , 1.3 µm and
2.4 µm from left to right.

Fig. 3.5 depicts this dependency. The left graph shows ring radius versus
bead displacement from the focal plane for a 200 nm4) -orange fluorescent bead fixed
between lens and focal plane, observed with epi-fluorescence. The ring radius and
the lateral position of the rings were found by cropping the central part of the
image taken by the camera (everything except the outmost ring) and fitting a ring
with Gaussian profile of the form f (r) = exp −k(r − r0 )2 [32]. This was done


20 times for each bead displacement for 20 different positions covering a total of
1.5 µm, and the average ring radius was plotted. A line fit yields a slope of 55.2 ±
0.9 nm/camerapixel for this particular lens (a Zeiss PlanNeoFluar 100x/1.3 Ph3 oil
immersion), meaning that if a defocused bead moves 55.2 nm with respect to the
lens along the z-axis, the radius of the outmost ring grows or shrinks by one pixel.
4
Numbers in this context denote particle diameter. All beads used are fluorescently labeled
carboxylated nanospheres form Molecular Probes, USA.

25
Figure 3.5: Interrelation between particle displacement from the focal plane and ring
radius on the camera (left). Standard deviation of the measured radius for 20 frames with
the same displacement are shown on the right for all three directions x (red), y (green) and
z (blue).

The error of the line fit has a negligible effect, since only small position fluctuations
of no more than 100 nm are of interest in practice (see section 4.2). In such a
case, other sources of error dominate as described below. Lateral calibration is
64.5 nm/camerapixel, from the size of one pixel on the CCD-chip (6.45 µm) and the
magnification of the objective used (100x). Error bars are due to inaccuracies in
lens position, which was set and determined using the motorized z-positioning of
the confocal microscope. Refractive index mismatch has been corrected for5 .
The graph on the right hand side depicts the standard deviation for each of
the 20 measurements for a particular distance between bead and focal plane. Dif-
ferences between measurements for the same position are mostly caused by noise
associated with photon statistics and mechanical noise. For a large range of dis-
placements between 800 and 1800 nm, the standard deviation is well below 4 nm.
This means that, for these displacements of the bead from the focal plane of imag-
ing, the position of a bead can be determined with a spatial resolution better than
5
Induced by refraction at the glass-water interface on the bottom of the sample chamber, this
causes the focal plane to move only by 0.86 µm when the objective is moved by 1 µm. See [26] for
details.

26
4 nm for all directions relative to the focal plane. The temporal resolution for most
applications is determined by the integration time needed to obtain a reasonable
image. Typically this is less than 82 ms.

3.2.2 Confocal laser-scanning fluorescence microscopy

In a standing wave there is naturally more than one position of stable trapping and
so several compartments can be employed to store numerous objects. For such ex-
tended, laser-scanning confocal microscopy provides the possibility of investigating
the three-dimensional organization of a multitude of particles with high spatial pre-
cision. Its principle relies on a mechanism that blocks out-of-focus light that would
otherwise lead to blurred images.

Light emitted Sample


above focal
plane Focal plane

Objective lens
Fluorescence
excitation
Dichroic mirror

Lens

Fluorescence
emission from Pinhole
focal plane
Photomultiplier

Figure 3.6: Schematic drawing of a confocal microscope.

Fig. 3.6 depicts the principle of confocal microscopy schematically. A laser


beam, denoted by solid lines, is used to excite the fluorescent dye that must be

27
present in the specimen. It is focused into the sample by the very same objective
lens that is used for imaging, where it excites the dye molecules in the focal spot.
Unfortunately, the laser intensity outside the focal spot is also sufficient for excita-
tion. Therefore, light is also emitted from above and below the focal plane (dashed
rays in fig 3.6) that would lead to images containing information from both the plane
of interest as well as information from regions above and below, in short, the final
image would be blurred. In order to avoid this, a pinhole is introduced, which acts
as a spatial filter, preventing out-of-focus light from reaching the detection system.
Only light originating from the focal plane (dotted lines in fig. 3.6) can be detected.
This depth-discrimination allows the imaging of selective planes in a specimen that
have a thickness that is only determined by the resolution of the objective lens in
the z-direction.

3.3 Mechanical stability of the setup

As it has been discussed above, the position of a single object can be determined
with a resolution of only a few nanometers. However, when position data has to be
recorded over a long time, mechanical drift of the trapping setup and the microscope,
which has not been designed to yield spatial resolution in this range, has to be
considered. Fortunately, the time scales over which noticeable drift takes place are
very long compared to the maximum temporal length of data recording. Therefore,
drift appears in position data recordings mostly as a linear component, on which
the desired data is superimposed, and can be removed easily by subtracting a linear
fit. Typical amounts of drift are displayed in fig. 3.7. The left hand column shows
the recorded x, y, and z-position of a 200 nm orange fluorescent bead attached to
a coverslip. It depicts fluctuations on small timescales, due to photon counting
statistics, and larger timescales that stem from drift in the optical system (objective
lenses, camera, etc.) and the sample chamber. The time range of the graphs is

28
longer than an actual measurement typically is, namely a maximum of 25 seconds.
By taking any 25-second-long subrange of a recording with the linear component
subtracted, and analyzing the fluctuations over these 25 seconds, it can be shown,
that they are of the same order as the fluctuations in fig. 3.5. Therefore, if data is
acquired over times between a few and up to 25 s, drift has a negligible effect .

Sample chamber and microscope drift Drift of the top platform

Figure 3.7: Measured position in all three directions of a 200 nm-fluorescent bead attached
to the top of the sample chamber (left) and a bead in a single beam trap (right); both were
acquired at the same time using the single particle tracking technique described in 3.2.1.
Integration time was 81.9 ms. Drift of the top platform was corrected for drift (see text).

The right hand column of fig. 3.7 shows the position of a bead trapped in a
single beam trap over time. It was recorded at the same time as the left hand side,
which was then subtracted, rendering fluctuations on long timescales merely due to
movement of the trapping lens. The fact that the high-frequency fluctuations are
bigger than on the left graph can be attributed to movement of the bead in the
trap and additional noise stemming from the subtraction of the noisy data from
the left. It can be seen that the trapping lens remains at its position within 10 nm.

29
Movements on this scale do not have a noticeable impact on the standing wave.

30
Chapter 4

Experiments

Many features of the standing wave trap as well as the experimental setup have
been discussed in the previous chapters. Now, that the theoretical and practical
foundations have been laid, several experiments will be presented that show the
power of standing wave optical trapping. Among these are transport and sorting of
trapped objects, the ability to unravel the interaction dynamics of several particles,
and trapping of unprecedentedly small specimens.

4.1 Independent trapping and observation

As discussed in 3.2, independent trapping and imaging allows for analyzing of dif-
ferent compartments of the standing wave, since it permits focusing into individual
traps while the focus of the trapping beam can be held on the reflective surface.
This ensures optimal and constant performance of the standing wave trap at all
times.
To demonstrate this, two orange fluorescent beads, numbered “2” and “3”,
were simultaneously trapped in different antinodes of a standing wave. Fig. 4.1 A
shows bead number three in focus in the lowest trap in the top right corner; the

31
A 2

1
3

B 2

1
3

C
2
1

Figure 4.1: Focusing into different antinodes without perturbing the standing wave. In
three rows, three different positions of the imaging lens are shown (A, B, C), each row
displays the actual picture on the camera (left) and a schematic drawing of the respective
situation (right). Numbers on the left denote affiliation of the rings to the beads as numbered
on the right and in the text. The dotted line on the right denotes the focal plane of
observation; the envelope of the trapping beam is indicated to show that the beam waist of
the trapping beam is placed on the reflective surface. 200 nm-orange fluorescent beads were
used, integration time: 81.9 ms, trapping power: 30 mW. The trapped beads are separated
by 1.6 microns or four antinodes. See text for further details.

32
dotted line denotes the focal plane. A reference bead (“1”) is glued to the reflective
surface at the bottom of the sample chamber and is visible in the lower left corner.
The first trap is formed approximately 160 nm above the surface, so both, the bead
in the lowest trap (3) and the reference bead (1) are in focus at the same time1 .
Since the other bead in the higher trap (bead number two) is located far away from
the focal plane, it appears as blurred rings around bead three in the lower trap. By
moving the imaging lens, the focus can then be shifted to the higher trap which
contains the second bead (fig. 4.1 B). The lower trapped bead and the reference
bead (beads 1 and 3) are now out of focus and show rings. Moving the focus even
higher causes also the bead in the higher trap to leave the focal plane, and all three
beads now exhibit rings (fig. 4.1 C). The outer ring radius of the lower trapped
bead, compared to the outer ring radius of the bead in the higher trap reveals
their separation of 1.6 microns. Note that the rings of bead number two have a
smaller diameter than those belonging to bead three; this shows that number two is
trapped in a higher trap closer to the focal plane. Both, the reference bead and the
lower trapped bead, have rings with almost the same outer radius because they are
still separated by only 160 nm in z-direction. If the trapping lens would have been
moved by 3.3 µm as it has happened here to the imaging lens, the configuration
of the standing wave would have changed considerably and the bead in the first
antinode would have been lost for the low trapping power used here. All this would
not be accomplishable with the imaging and trapping path sharing one objective
lens.
In combination with laser scanning confocal microscopy, three-dimensional
distributions of particles in different antinodes can be analyzed. For that, a horizon-
tal slice is being imaged, then the focus of the imaging lens is moved, and another
cut section is taken at a different position on the optical axis. Using a PC, these
1
Focal depth is about 400 nm.

33
Figure 4.2: A distribution of multiple particles in the standing wave trap, imaged using
confocal laser scanning and software deconvolution with “Huygens” from Scientific Volume
Imaging, Netherlands. The reflective surface is located on the bottom of the picture. The
volume that was scanned was 8 × 8 × 7.5 µm (x-width × y-width × height) and consisted
of 150 slices separated by 50 nm, each one 128 × 128 pixels. Trapping power was 130 mW.
White lines indicate the 400 nm-spacing between antinodes.

34
images can be combined for a volume rendering of the trapped particle distribution
that can be rotated or cut along arbitrary planes. Fig. 4.2 shows an xz-cut of a
standing wave trap containing several fluorescent 200 nm-beads after deconvolution.
Their vertical spacing corresponds to the expected distance between antinodes, that
is, half the trapping wavelength in the medium, or 400 nm. The lateral displacement
of the topmost bead is most likely due to shadowing effects caused by the two beads
in the trap below and trapping on rings (see page 42). The lowest bead jumped
between two adjacent antinodes during acquisition and is therefore blurred across
two traps.

4.2 Stiffness of individual antinodes

It has already been discussed that an object in a standing wave trap has several equi-
librium positions2 , the antinodes of the standing wave, that could be used to store
simultaneously different objects and keep them separate from each other. However,
this would only be possible if all antinodes to be used are potentially capable of
stably confining one or more objects. In order to verify this, the antinodes of the
standing wave have to be characterized for their ability to stably trap specimen of
a given size and index of refraction. One way to achieve this is to utilize the trap
stiffness as a measurement for the trapping efficiency. The stiffness of a trap at a
particular location and along a certain direction is defined as the derivative of the
force at that position with respect to that particular direction. In other words, it
describes how much the force changes for a small displacement of the trapped object
from this position. It requires some effort to measure this for all positions, but in
practice the optical energy landscape of a trap close to its minimum can well be
approximated by a harmonic potential. Since any stably trapped object will stay
2
Objects of specific dimensions actually cannot be trapped due to the periodic nature of the
standing wave. However, that only holds for sizes that are much larger than what is of interest
here. For a detailed discussion see [18, 41].

35
close to the energy minimum for most of the time, the trap stiffness can be expressed
by a single value, the curvature of this potential.
To probe the stiffness around the equilibrium position, one can make use of
the fact that every particle trapped in solution at a finite temperature is subject to
Brownian motion, and, according to eq. 2.15, the mean square displacement caused
by thermal fluctuations is directly linked to the stiffness of the trap3 .

Figure 4.3: Left: z-position of a trapped 200 nm-bead over time. Shown are 1000 frames,
integration time was 81.9 ms. Trapping power was 70 mW with a numerical aperture of 1.3.
Right: Histogram of the position data from the left graph (bars) and the Gaussian fit with
a width of 10.6 ± 3.0 nm.

Fig. 4.3, left shows the first 1000 frames of 3200 position recordings of a
trapped 200 nm-bead, acquired using the technique described in section 3.2.1. On
the right hand side is the histogram of the position data that fits well to a Gaussian,
showing that the potential is indeed harmonic for the displacements probed (see
eq. 2.16). The width of the fitted curve corresponds to the amount of fluctuations,
however before the stiffness can be calculated, some corrections have to be applied.
The characteristic timescale τ0 = β/κ of the overdamped movement of a particle in a
harmonic potential depends on the trap stiffness κ and the drag coefficient β = 6πηa,
where a is the radius of the bead and η is the viscosity of the surrounding medium.
For a typical trap, τ0 is on the order of a few milliseconds, whereas data acquisition
was done with an integration time of 81.9 ms. This implies that the bead crosses
3
Other methods for obtaining the stiffness are discussed in [37].

36
the trap several times during the acquisition of a single image and the resulting
position data shown above is therefore heavily low-pass filtered. However, the filter
is known and can be corrected for. The original trajectory cannot be fully recovered,
but its statistical properties, such as the amount of fluctuations, can be computed
(see appendix B for details). For the case shown above the original fluctuations
were 53.5 ± 8.3 nm and the trap stiffness in this direction is 1.4 ± 0.4 · 10−6 N/m,
according to eq. 2.15.
This procedure can be applied to all antinodes of a standing wave trap while
external parameters are varied, the two most important ones being the power of
the trapping laser and the numerical aperture of the trapping lens. Since stiffness
scales linearly with trapping power, the former can be accounted for by normalizing
stiffness to laser power.

Figure 4.4: Normalized stiffnesses for the first four antinodes with NA=0.7 for the x
(red), y (green) and z-direction (blue). Markers are experiments, solid lines are simulations.
Integration time: 81.9 ms, beam waist radius used for simulations: 800 nm. Slide reflectivity
for 1064 [nm]: 98%.

Fig. 4.4 shows the normalized stiffness in the first four antinodes of a standing

37
wave trap for a numerical aperture of the trapping lens of 0.7, corresponding to a
theoretical beam waist radius of 600 nm. The lateral stiffnesses (red and green) are
of the same order and are approximately the same as the axial stiffness of a single
beam trap with NA = 1.3 (ca. 1 − 2 · 10−5 N/Wm). However, single beam trapping
with the numerical aperture of 0.7 used here is unachievable, since the axial gradient
would be too low to prevent particles from escaping along the optical axis in the
direction of the scattering force. The axial stiffness of a NA=0.7 standing wave trap
is one order of magnitude higher than the axial stiffness of a single beam trap with
NA=1.3, thus, the bigger intensity gradient and the cancellation of the scattering
force is sufficient to allow trapping for a moderately focused beam. The slow drop
of the axial stiffness for higher traps implies that trapping is possible for antinodes
above the fourth position, and in fact as many as 20 traps have been observed in a
standing wave trap with NA=0.7.
Trap stiffness measurements were compared to simulations done by Alexandr
Jonáš. In principle, the field distribution in a Gaussian standing wave can be de-
scribed by the interference of two counter propagating focused beams ([5]). From the
field distribution, one can calculate the force for a given object and finally the stiff-
ness (see A for details). For the 200 nm-beads used here, the Rayleigh-approximation
already shows noticeable deviations and the forces have to be calculated using the
more elaborate Lorentz-Mie theory. Even with this rigorous formulation, full agree-
ment between experiments and simulations in terms of absolute numbers cannot
be expected, since several parameters needed for simulation are hard to obtain in
practice. For example, the intensity of a focused beam in the focal plane is diffi-
cult to measure accurately. In addition, any external noise will add to the particle
fluctuations and artificially deflate the measured stiffness. Furthermore, even for
the much simpler case of a single beam trap, good agreement between theory and
measured trapping stiffness has only recently been obtained [29].

38
Compatibility between measurements and simulations is typically better than
one order of magnitude, however, to make ratios comparable, data obtained by
simulations have been normalized to the experimental value for the axial stiffness
of the first antinode4 in fig. 4.3, so that the z-stiffness of the first antinode is
the same for theory and experiment. It can be seen that the ratio of stiffness
between the first and higher traps can be measured as predicted for all directions.
The characteristic drop of the z-stiffness is present in the experimental data, as
well as the flat behavior for both lateral stiffnesses. However, the simulated ratio
between axial and lateral stiffness for an individual trap is not exactly reproduced.
This can be explained qualitatively. It was assumed for the simulations, that the
trapping beam has Gaussian form, so wavefronts close to the beam waist are parallel.
Therefore, when the beam waist of a Gaussian beam is placed exactly on the surface,
all incident waves5 hit the surface perpendicularly. In practice, the beam shape
deviates from a simple Gaussian, because the refractive index missmatch between
the glass coverslip and the solution inside the sample chamber causes peripheral
rays to be focused closer to the objective than central rays [26], inducing spherical
aberrations. This means that some waves do not hit the surface perpendicularly,
therefore the reflectivity for these waves will be less than assumed, and an additional
phase shift is introduced. Since the axial modulation in a standing wave is created
by interference, it is naturally very sensitive to phase shifts. In particular, the effects
described above cause the axial modulation to be less than assumed, lowering the
measured axial stiffness. Lateral stiffness depends much less on the interference of
incident and reflected beams and so effects caused by a deviation of the beam from
Gaussian form are minor here. In conclusion, axial stiffness is lowered much greater
than lateral stiffness by beam imperfections. Therefore, the measured ratio between
axial and lateral stiffness is less than in simulations.
4
The normalization factor is the same for stiffnesses in all three directions.
5
The beam can be decomposed into a spectrum of plane waves.

39
Error bars on the measured stiffness are due to several contributions. Due
to photobleaching, only a certain number of frames can be recorded with one bead,
here 300. This affects the accuracy of the position histogram and hence its Gaussian
fit. Also, traps themselves differ in stiffness depending on where they are formed on
the reflective slide. Similar to what has been described above, a slight deviation of
the surface properties can introduce unaccounted phase shifts that have a noticeable
effect on the generated interference structure. This holds also for the data presented
in fig. 4.4.

Figure 4.5: Normalized stiffnesses for the first four antinodes with NA=1.3 for the x
(red), y (green) and z-direction (blue). Markers are experiments, solid lines are simulations.
Integration time: 81.9 ms, beam waist radius used for simulations: 400 nm (1/e2 ). Slide
reflectivity for 1064 [nm]: 98%. The dotted line denotes the single beam trap axial stiffness
for the same numerical aperture.

Fig. 4.5 shows the stiffness in the first four antinodes of a standing wave,
formed using an objective with a numerical aperture of 1.3. In contrast to NA=0.7,
several traps with almost the same stiffness have been traded for one antinode which
exhibits very high stiffness and a quick drop in stiffness for higher traps. The axial
stiffness for the first trap is two orders of magnitude higher than the axial stiffness
of a single beam trap (dotted line in fig. 4.5) for the same numerical aperture. Also,

40
lateral stiffness has increased significantly. This shows that the weakest part of the
single beam system, the low axial to lateral stiffness ratio, has been successfully
eliminated as this ratio has been turned around from κaxial /κlateral ≈ 0.25 for a
single beam trap to four in the standing wave case. The improvement in lateral
stiffness can be explained as follows: The dimensions of the potential wells in a
standing wave trap are given by λ/2 along the axis and the numerical aperture lat-
erally. Therefore, an increase in axial stiffness, that is an increase in axial curvature
of the potential, will cause a deeper potential well. The lateral dimensions of the
potential well are fixed, therefore, an increase in depth of the well is accompanied
by higher lateral curvature.
Even higher traps show at least the same axial stiffness as a single beam trap,
in fact, high-NA standing wave trapping is possible up to ten antinodes depending
on the power of the trapping beam.
Again, experiments were compared to normalized simulations. For the first
antinode, the ratio of axial to lateral stiffness can be measured as predicted, as well
as the drop of lateral stiffness for higher antinodes. Also, for this numerical aper-
ture, lateral stiffnesses differ in the x and y-direction, a fact that can be explained
with the polarization of the trapping beam. For the direction perpendicular to the
polarization direction, x in this case, the stiffness is expected to be slightly higher
[29] and this difference can be observed in concurrence with simulations. However,
the measured drop in axial stiffness towards higher antinodes is much more severe
than predicted. This is a reproducible deviation from theory, which can be explained
by comparing the real intensity distribution to the distribution as it is assumed in
the model.
Fig. 4.6, left shows the intensity distribution of a focused beam as described
by a Gaussian formalism with higher order corrections in [5]. Shown on the right
is the point-spread function (PSF) of the same objective lens used for trapping,

41
Figure 4.6: rz-plane of the point-spread function of the trapping lens for NA = 1.3, as
assumed in the model (left) and measured using confocal laser scanning (right). Beam waist
radius for simulation: 400 nm (1/e2 ), corresponding to NA=1.3

obtained by imaging a 200 nm fluorescent bead. Ideally, they should be almost the
same, as both the image of a small object and the intensity distribution of a focused
parallel beam, are given by the point-spread function6 . It can be seen that the mea-
sured PSF differs significantly from what is used in stiffness simulations. The actual
standing wave distribution formed when a reflective surface is placed near the beam
waist is very hard to predict, since also phase information, which is not accessible,
has to be taken into account. Hence, when the bottom part of the distribution on
the right is reflected back by the surface over the top half, both interfere and form a
6
Imaging an object with a size of 200 nm does not exactly yield the PSF, since it does not
completely fulfill the requirement of being much smaller than the diffraction-limited resolution of
the objective lens. The shown distribution is therefore a convolution of the real PSF and a sphere
with a diameter of 200 nm, also influenced by confocal imaging. However, the fact that the PSF
differs significantly from a Gaussian distribution holds, nevertheless. Additionally, the PSF was
measured in the visible range, but an infrared beam is used for trapping. Since the objective lens
used was optimized for the visible spectrum, the point spread function for IR most likely differs
even more from Gaussian shape.

42
highly complicated structure. Nevertheless, the measured intensity distribution can
explain the discrepancy between measurement and simulations qualitatively.
The first antinode is formed 160 nm from the beam waist. In this region, the
measured distribution can still be described well by the Gaussian model used for
simulations as can be seen by comparing the two plots in 4.6. Hence, the measured
ratio of axial to lateral stiffness as shown in Fig. 4.5 agrees with theory within
error bars. Higher traps are formed farther away from the beam waist, where the
intensity distribution shows a strong deviation from Gaussian shape. Especially
when the distance from the beam waist exceeds approx. 1 µm, tails become visible
that by no means are included in the model. The deviation from a Gaussian increases
with distance from the beam waist, thus, the stiffness for higher traps drops faster
than predicted. It has to be mentioned that the formalism used to calculate the
stiffness from the field distribution is in principle correct, the deviations explained
here arise because the field-behavior deviates from the assumed distribution at high
numerical apertures. If an expression could be found that sufficiently resembled the
actual field situation, the stiffness could be computed accurately.
The point-spread function is rotationally symmetric around the optical axis
(vertical in fig. 4.6), so the tails addressed above actually form cones7 . Since a
significant amount of intensity gets redistributed from the optical axis to cones, the
question arises whether the gradient there is sufficient for stable trapping. When
a bead is trapped in an antinode higher than the second, shutting off the trapping
beam causes it to diffuse, and if the beam is turned on again at a time when the
bead has an off-axis position, it can indeed be trapped on a cone.
Fig. 4.7 shows the position histogram for that case. The cones are broken
into discrete rings by the axial modulation of the standing wave, so a bead trapped
7
These cones are caused by diffraction at the aperture of the lens and are similar to the rings
used for particle tracking (see section 3.2.1). The rings appear when the focal plane of observation
(which is parallel to the xy-plane) is not located close to the object, but cuts through a cone
perpendicularly to the cone’s axis of symmetry (the optical axis).

43
Figure 4.7: Position histogram of a 200 nm-bead moving on a ring in the xy-plane.
Trapping power: 90 mW, integration time: 10 ms. The particle is trapped at the z-position
of the third antinode.

on a cone moves along a ring in the xy-plane. Usually, the ring intensity is not
completely symmetric, and the bead prefers to stay in one or sometimes two sections
of the ring for most of the time. The diameter of the ring shown corresponds to the
diameter the cone is expected to have at a distance from the beam waist where the
third antinode is observed. To study trapping on rings for a long time, the trapping
power has to be relatively high, since the on-ring intensity is still smaller than the
light intensity along the optical axis. After about 40 seconds the bead falls back
into the deeper potential well on the axis, causing the central peak in Fig. 4.7.
It can be shown that for low NA, there is no noticeable deviation of the
measured PSF from Gaussian shape. No cones are formed, and no trapping on
rings has been observed for a weakly focused beam.

44
4.3 Selective storage of objects in antinodes

As described in the previous section, the stiffness of higher traps for NA=1.3 is
smaller than expected, but it is still sufficient to trap objects stably in several
compartments, the antinodes of the standing wave. In order to make full use of
these compartments as storage space, a method is required to selectively deposit
different objects in different antinodes.

4.3.1 Transport of a particle between antinodes

In a high-NA standing wave trap, when the beam waist is placed on the reflective
surface, the deepest potential well is generated close to the reflective slide. In
general, this situation is desirable, since it guarantees the greatest overlap between
incident and reflected wave, and therefore the biggest intensity-modulation along
the optical axis. If the beam waist is lifted from the reflective surface, the stiffness
and the trap depth in all antinodes decreases, however, it remains true that the
deepest trap is in the vicinity of the beam waist. Also, the further an antinode is
away from the beam waist, the shallower the trap is that is formed by this antinode,
as before. Thus, movement of the beam waist can induce particle jumps to other
antinodes.
Fig. 4.8 shows the simulated potential energy of an object in a standing
wave and explains the mechanism of transport. In the top left graph, a particle
is trapped in the energy minimum corresponding to the first antinode. The beam
waist, denoted by a black arrow, is located on the reflective surface at z = 0. For a
numerical aperture of NA=1.3, the first trap is the deepest one, with higher traps
being more and more shallow. If the beam waist is moved away from the surface,
the position of the deepest potential well follows the beam waist, as discussed above.
At some point, the potential barrier between the first and the second antinode gets
low enough for the particle to cross it, moving into the next antinode (top right

45
Figure 4.8: Simulated energy landscapes along the optical axis and behavior of a trapped
particle (black dot) when the beam waist is lifted off of the surface. The reflective surface
is located on z = 0, the black arrow denotes the position of the beam waist. The distance
between beam waist and reflective surface is 0 µm (top left), 0.8 µm (top right), 1.2 µm
(bottom left) and 0 µm (bottom left). Beam waist radius: 400 nm. In the bottom left graph,
the power or the trapping beam was increased. See A for how the plots were obtained.

plot). This is induced by thermal motion if trapping power is chosen low enough,
but not so low that the particle is lost completely. By shifting the beam waist even
further, the particle can be dragged along into higher traps (lower left).
As discussed in section 4.2, the time τ0 it takes for a particle to fully probe its
trap is on the order of milliseconds, so changes of the optical potential take a time
τ0 reflect on the particle. If the beam waist is moved fast enough, the particle will
not be able to follow due to viscous drag acting on it in immersion medium. This
way, the initial optical configuration with the beam waist being on the surface can
be restored, while the particle has effectively been transported to a higher antinode.

46
Possibly, trapping power has to be increased, so that the particle is stable in its new
position.
Fig. 4.9 shows the four stages of particle transport depicted schematically
in Fig. 4.8, as they appear on the camera. The particle on the left is a reference
particle attached to the surface, the particle on the right in subfigure A is confined
in the first trap 160 nm above the surface. Both show large rings in subfigure A,
because the imaging lens is focused into a plane far above the surface. As the
trapping beam waist is moved upwards (subfigures B, C, D), the trapped particle
follows and moves closer into focus, thus, its ring diameter decreases. No change
is noticeable for the reference particle as is keeps its distance from the focal plane
(compare this to Fig. 4.1 where the particles are fixed in space and the focal plane
is being moved, causing also the rings of the reference bead to change).
For a high-NA standing wave trap with the beam waist on the surface, a
particle in a higher trap corresponds to a high-energy state, since the particle is
trapped in a local, but not a global energy minimum. It will remain trapped only if
the local potential barriers are higher than several kB T . Fig. 4.10 left depicts the
energy landscape for the situation of a particle that is stably trapped in a higher
antinode with the beam waist on the surface. If the potential barriers are lowered
by decreasing the intensity of the trapping beam, the particle will leave the current
trap. As discussed above, the potential wells get deeper closer to the beam waist.
Furthermore, due to imperfect reflectivity of the surface, the intensity modulation in
the standing wave is not complete, i.e. the intensity will not completely reach zero
at the nodes of the standing wave. For this reason, the potential wells corresponding
to an individual trap are asymmetric with the lower barrier oriented towards the
beam waist. Thus, if a local potential well is not deep enough to confine a particle,
it will leave this antinode by crossing the potential barrier closer to the beam waist,
heading for a deeper potential well (fig. 4.10, right). This way, particles can be

47
Figure 4.9: A 200 nm-bead is moved into higher antinodes. A, left: The trapped bead is
in the first antinode. A reference bead is attached to the bottom of the sample chamber left
of it. A, right: Schematic drawing of the situation on the left. The beads (red) are close to
the reflective surface (blue) and far away from the focal plane of observation (dashed line),
thus they appear as large rings on the left. The beam waist is on the surface, indicated by
the envelope of the trapping beam. B, C, D, left and right: The beam waist is lifted off
from the surface, and the trapped particle follows in discrete steps. It moves closer to the
focal plane, and the ring diameter decreases in the pictures on the left side. Trapping power
was constant at 180 mW for pictures A through D.

48
Figure 4.10: Trapped particles preferably hop towards the beam waist when the trapping
power is decreased.

transported downwards from a higher antinode without having to move the beam
waist.

4.3.2 Sorting of particles according to size and refractive index

In order to make full use of the fact that several antinodes are formed within one
standing wave trap, not only transportation, but also the ability for selective trans-
portation is required, so that different materials can be stored in separate antinodes.
Fortunately, the mechanisms for transportation described in the previous chapter
have a distinctive effect on objects with different properties. In the following, only
size-effects will be discussed, but the same holds for different refractive indices.

Figure 4.11: Energy landscapes for a small particle (small black dot and red curve), and
a particle twice as big (large black dot and green curve). In the left plot, the beam waist is
0.8 µm above the surface, on the right side it is on the reflective surface at z = 0.

49
The impact of particle size on trapping-stability is illustrated in fig. 4.11.
Since the gradient force on an object decreases with its volume, potential barriers for
a given trapping power are lower for smaller particles. In the left plot, two particles
with different sizes are trapped in the second antinode, with the beam waist being
0.8 µm above the surface which is at z = 0. When the beam waist is shifted towards
the reflective slide, the deepest trap is always formed close to the position of the
beam waist, as discussed in the previous section. For low trapping power, both the
big and the small object, follow the deepest trap along with the beam waist. At a
certain power however, only the small particle can be dragged through the standing
wave with the beam waist. The potential barrier of the trap that the bigger particle
is in, is always high enough to prevent its escape, no matter what the location of
the beam waist is (fig. 4.11, right). By moving the beam waist and readjusting
power, so that the potenial well is deep enough to always confine the bigger particle
while the small particle can follow the beam waist, differently sized objects can be
separated by several microns. Of course, the same holds for particles with different
refractive indices.
Fig. 4.12 shows confocal scans along the optical axis for a 100 nm orange
fluorescing (red in the graphs) and a 200 nm yellow-green fluorescent bead (green
curve). Maximum intensity corresponds to the particle’s position. In the top graph,
the beam waist is close to the location of the two beads, several microns away from
the reflective surface. Both particles are trapped in the antinode formed there.
With a trapping power of ca. 100 mW in the focal plane, the beam waist was moved
2 µm closer to the surface, and another scan (center) revealed that only the smaller
orange particle had followed. Shifting the beam waist by another 1 µm resulted in
the configuration at the bottom, where the red curve has been displaced by more
than two microns compared to the topmost plot, indicating that the smaller orange
particle was separated from the bigger one by five antinodes.

50
Figure 4.12: Sorting of trapped particles for size and refractive index. Given z-positions
are relative, z = 0 does not correspond to the reflective surface. More details see text.

51
This process is fully reversible, by moving the beam waist back a little more
than two microns, separated particles can be mixed and mixed particles can easily
be separated again.

4.4 Trapping with low numerical aperture

Building the standing wave optical trap was motivated by the desire to investigate
the interaction dynamics of several nanoparticles in solution. In order to arrive
at this goal, the need for a system is evident that is capable of confining two or
more immersed objects within the field of view of a microscope, and especially in
the focal plane of observation, for a sufficiently long time. Furthermore, it has to
allow for the controlled aggregation of separated particles. The standing wave is

Figure 4.13: Position histograms for a 200 nm-particle trapped in the first antinode with
NA=0.7 (top) and NA=1.3 (bottom). Trapping power in the focal plane was 70 mW in both
cases.

ideal for that purpose, since the lateral potential can be tuned over a large range,

52
while the axial stiffness is high at all times, keeping particles from escaping from
the focal plane. It can be inferred from figs. 4.4 and 4.5, that the lateral stiffness
shows a strong dependence on the numerical aperture of the lens used for focusing
the trapping beam. For high NA, the ratio of axial to lateral stiffness for the lowest
trap is typically around 3.5. When the numerical aperture is decreased to 0.7, the
lateral stiffness of the first antinode drops much faster than the axial stiffness, the
ratio rises up to over 30. This is illustrated in fig. 4.13 that depicts xz-position
histograms of a trapped 200 nm bead for NA = 0.7 (top) and NA = 1.3 (bottom).
Table 4.1 summarizes the fluctuations h∆xi and h∆zi as well as the stiffness ratios
κz /κx = (h∆xi/h∆zi)2 . It can be seen that, while going from NA=0.7 to NA=1.3,
the mean fluctuations of a particle along the optical axis merely drop by a factor of
two, but the mean fluctuations along x decrease six fold.

NA h∆xi/nm h∆zi/nm κz /κx


0.7 60.1 ± 2.8 10.0 ± 0.5 36
1.3 9.7 ± 0.5 5.0 ± 0.2 4

Table 4.1: Fluctuations and stiffness ratios for the numerical apertures shown in
fig. 4.13.

In other words, sheet-like potentials are formed in the antinodes of a low-NA


trap, that allow extended trajectories in the xy-plane for one or more particles, while
axial confinement is high, preventing all objects from escaping from the focal plane
of observation. Furthermore, in a standing wave trap, the numerical aperture that
can be used for trapping is not limited from below, unlike for a single beam trap.
Since no scattering forces are present, the lateral stiffness can in principle approach
zero when a plane wave is used to form the trap8 . When two or more small particles
are present in such a low-NA trap, they are free to move laterally over length scales
that exceed their size by far, rendering them unaware of each other. The amount
8
That can be achieved by focusing the beam into the back focal plane of the trapping lens, or
simply by underfilling the back aperture if the desired NA is slightly above zero.

53
of lateral fluctuations can then be tuned continuously by adjusting the numerical
aperture to higher values, thus going from independent movement of particles to
forced interaction, as objects are confined into a increasingly smaller volume for
higher NA.

Figure 4.14: Two particles in a low-NA trap (left and center) and the same particles
trapped with high numerical aperture (right). The images are 2.8 × 2.8 µm.

Fig. 4.14, left and center show two consecutive frames of 200 nm-beads in
a low-NA trap created by underfilling the back aperture of the trapping lens. The
dimensions of the images are 2.8 × 2.8 µm, the particles in the left picture are
separated by ca. 800 nm. This shows that two objects in a low-NA trap can be
separated horizontally by a distance that is a multiple of their size, while they are
confined to the focal plane of the imaging objective. Low lateral stiffness implies
that the potential well has a low curvature laterally. However, since its dimensions
are large, it is still deep enough to stably confine the particles9 . The right picture
shows the same particles after the numerical aperture has been increased to 1.3.
The objects are now separated by only twice their radii (200 nm) and cannot be
resolved individually. This could easily be solved by labeling both of them with
different fluorescent dyes that emit light at different wavelengths, and, using dichroic
mirrors to separate the light paths, they can effortlessly be observed independently.
Simultaneous observation of two fluorescent colors at the same time is currently not
9
Stable trapping is determined mostly by the depth of the potential well (see eq. 4.1). A
potential well with low curvature can still be deep if its dimensions are large.

54
possible on the setup, but it poses no principal difficulties. Nevertheless, fig. 4.14
proves that it is possible to investigate the interaction dynamics of more than one
object by tuning lateral trap stiffness.

4.5 Optical trapping of 40-nanometer polystyrene par-


ticles

It has been mentioned several times, that trapping of biomolecules in a single beam
trap is impossible. As force decreases with the volume, smaller particles can only be
trapped if the intensity gradient is enhanced. Due to physical limitations, this can
only be accomplished in a single beam trap when the intensity itself is increased,
however serious problems evolve for biological material in high-intensity fields. Fur-
thermore, even for very high intensities, particles below 50 nm, are pushed out of
the trap by the scattering force along the optical axis within seconds. Even metallic
particles of that size can hardly be confined by a single beam trap10 despite their
high index of refraction. In contrast, the scattering force is strongly suppressed in
the standing wave trap, and the axial gradient can be much higher than the lateral
gradient, which itself is already steeper than any gradient that can be achieved with
a single beam trap. It follows, that for given power and particle properties, the
escape time from the standing wave trap is much higher. In other words, particles
can be trapped with considerably less power in a standing wave.
Fig. 4.15 displays an overlay of five consecutive frames of a trapped polystyrene
particle with a diameter of 40 nm, as it is moved from A to B. The bigger reference
particle C is attached to the bottom of the sample chamber. The power in the focal
plane during movement was 200 mW to keep the particle in the trap despite viscous
10
Trapping of 36 nm-polystyrene particles is reported in [34] at laser powers on the order of
hundreds of milliwatts in the focal plane. However, the particles stayed in the trap for only a few
seconds.

55
Figure 4.15: A 40 nm-bead trapped and moved from A to B over several microns within
330 ms (overlay of five frames). The bigger object C is stuck on the surface. Numerical
aperture was 1.3.

forces, but for as little as 90 mW trapping power in the focal plane particles could
still be trapped for at least one hour11 . This shows, in contrast to what has been
achieved before [34], that 40 nm-polystyrene particles can be trapped stably at low
power.
The lateral trap stiffness for a 40 nm bead at 90 mW trapping power in
the focal plane was found to be κx = 6 ± 2 · 10−7 N/m along the x- and κx =
4 ± 1 · 10−7 N/m along the y-direction. Axial position detection was not possible
11
Two-photon fluorescence induced by the trapping beam caused photobleaching and prohibited
observation over longer times.

56
due to very dim rings that are not visible on the camera. However, according to fig.
4.2, axial stiffness should be about three to four times higher.
The escape time τesc of an overdamped particle immersed in an aqueous bath
with the temperature T from a potential well with depth U is given by

U
 
τesc = τ0 · exp , (4.1)
kB T

where τ0 = β/κ is again the characteristic timescale of particle movement in the


potential well [34]. Inserting the y-stiffness measured above for 90 mW, together
with an escape time of 1800 s and solving for U , gives a rough estimate of 14kB T for
the depth of the potential well in the above case. Extrapolation to smaller particles
shows that escape times in the order of seconds can be reached for 20 nm-particles,
when the power is increased fourfold. This means that 20 nm-polystyrene beads
could be trapped with a power of 400 mW. This is still moderate compared to what
a single beam trap would require (by far more than 1 W, when the data from [34]
is extrapolated).

Figure 4.16: Assembly of 40nm-particles in a standing wave trap with NA=1.3.

57
It has been mentioned on page 54 that, for more than one particle in the trap,
individual trajectories can be observed by labeling objects with different colors. In
practice, more than three colors cannot be discriminated, but it is still feasible to
obtain the number of trapped particles by looking at the fluorescence intensity at
the trap position, assuming all particles contribute equally. With this method, dy-
namic assembly of 40 nm-polystyrene particles has been observed. Fig. 4.16 shows
that indeed the intensity increases in discrete steps, when a trap is placed in a con-
centrated solution of beads. One particle is in the trap initially, and after 12 s, more
particles diffuse into the trap and leave again, changing the total intensity observed.
Up to five particles have been found in this 150 mW-standing wave. Additionally,
these discrete fluorescence intensity changes can be used to verify that single parti-
cles and not clusters are trapped, which would be confined more easily due to their
bigger volume. The intensity observed in fig. 4.16 changes in steps of 15.2 ± 1.9
units. Clusters of beads found on the bottom of the sample chamber exhibit discrete
intensity steps of 15.7 ± 1.4 units. Since it can safely be presumed that the number
of beads in these clusters increases in steps of one bead only, it follows that the
intensity steps in fig. 4.16 correspond to a single bead.

58
Chapter 5

Summary and outlook

An optical trap that makes use of the enhanced trapping features of two focused
counterpropagating beams formed by a reflective surface has been built and charac-
terized. In this standing wave optical trap the parasitic scattering force is strongly
suppressed, while it exhibits intensity gradients that are up to two orders of magni-
tude higher than what can be achieved in a single beam trap, rendering it capable of
exerting considerably higher forces on small objects. Additionally, multiple trapping
positions are generated.
The ratio of axial to lateral stiffness for 200 nm-polystyrene beads in a stand-
ing wave trap was measured to be as predicted from the generalized Loretz-Mie
theory for low numerical apertures. Up to 20 antinodes can be employed for con-
finement in a low-NA standing wave, with a normalized axial stiffness that was
approximately 35 times higher than what could be achieved in a high-NA single
beam trap. A weakly focused single beam does not permit trapping at all. For high
NA, the measured and calculated stiffness ratio differs significantly. However, this
could be explained qualitatively with the deviation of the trapping beam from the
Gaussian shape assumed in calculations. Axial stiffness in a high-NA standing wave
trap is about 150 times higher than in a single beam trap for the same power and

59
numerical aperture.
A method for transporting particles between standing wave antinodes has
been found, that is selective to the particle’s properties and can thus be utilized to
store objects differing in size or refractive index in different antinodes. The ability
to sort and mix particles allows bringing specimen together selectively to start and
stop specific interactions systematically.
Polystyrene beads with a diameter of 40 nm have been trapped with an un-
paralleled high escape time of more than an hour and with unprecedentedly small
laser power. Extrapolation of the measured stiffness and escape time have suggested
that trapping of 20 nm-polystyrene particles is within reach. This would allow trap-
ping of cellular protein complexes, such as microtubule, that exhibit similar size and
refractive indices.
By using a low NA-lens for trapping, sheet-like potentials can be formed,
that allow long-time observation of the interaction dynamics of several particles. In
combination with the ability to trap very small objects, the observation of intriguing
events should be possible real-time. An example is fusion of lipid vesicles, that
has never been observed directly free in solution. Vesicles are compartments that
cells use when material has to be concealed from the cytosol, or transported to a
different location. They are enclosed by a lipid bilayer, and two vesicles can fuse.
Such an event could be investigated in a low-NA standing wave trap, that could
keep interacting vesicles in the focal plane for a considerable time, while the amount
of interaction could be tuned. As fusion takes place, the individual trajectories of
the vesicles, that can be determined if the vesicles are fluorescently labeled with
different colors, should become increasingly correlated.
Low-NA trapping of small particles could also potentially be a method to
study assembly of virus capsids from subunits, that have a size slightly below 20 nm
and a refractive index similar to that of polystyrene.

60
Furthermore, a multitude of other assays is imaginable not only in the field
of biological physics. A low-NA standing wave could also serve to investigate mech-
anisms in colloidal systems. One application of colloidal systems is material science,
where they act as model systems that behave much like bulk matter from many
points of view, but have time- and length scales that are much more easily acces-
sible in experiments. In a low-NA trap a two-dimensional colloidal system can be
created, and phase transitions can be observed as the lateral potential is squeezed,
to mention only one approach.

61
Appendix A

Formal description of intensity


and forces in a Gaussian
standing wave

In order to calculate the force on a particle in a Gaussian standing wave, the intensity
distribution has to be known. To first approximation, scattering of light from the
particle itself can be neglected, and the intensity distribution at the position (z, r)
~
can be described by a superposition of the E-fields of the incident and reflected
Gaussian beams.
For a tightly focused beam, when the beam waist radius is of the same order
as the wavelength or smaller, the fundamental description of a Gaussian beam does
not accurately satisfy Maxwell’s equations [41, 5]. This is already the case for a beam
that is focused using an objective with a rather moderate numerical aperture of 0.7.
Better agreement is obtained using a description that expands the electromagnetic
fields to the fifth order in the beam size parameter s = 1/kω0 [5].
The intensity distribution shown in fig. 4.6 is that of a fifth-order cor-
rected Gaussian beam. The energy landscapes plotted in figs. 4.8, 4.10 and 4.11

62
were simply determined by taking the vector sum of the electromagnetic fields
with components Exi,r , Eyi,r , Ezi,r , Hxi,r , Hyi,r and Hzi,r for two counterpropagat-
ing fifth-order corrected Gaussian beams at (x, y, z), obtaining the intensity from
~ sum (x, y, z) × H
I(x, y, z) = const · |E ~ sum (x, y, z)|, and plotting negative on-axis in-

tensity. This corresponds to the Rayleigh-approximation of the potential energy W


of an object in the standing wave, since from F~ ∝ ∇I and the relation between en-
ergy and force, it follows that W ∝ −I. Despite the fact that other contributions to
the total intensity (scattering, diffraction, beam aberrations) are not considered and
the Rayleigh-approximation is used for particles that are not considerably smaller
than the wavelength, this description is sufficient to explain transport, hopping and
sorting of trapped objects qualitatively.
In order to arrive at a quantitative description of stiffness ratios as given
in 4.2, the more elaborate model described in [18] has to be used. It relies on the
general equations

I
F~ = h ~
T̄ dAi (A.1)
1
T̄ij = εEi Ej + µHi Hj − (εEi Ei + µHi Hi ) δij , (A.2)
2

which state that the force on a particle in a light field can be computed if the
total external electromagnetic fields are known on a closed surface surrounding the
object. As explained in 2.1.1, this requires to include scattering from the particle
itself. Barton and co-workers described the interaction of an arbitrary beam with a
spherical particle [6], and went on to derive theoretical expressions of the radiation
force on a spherical object relying on eqs. A.1 and A.2 [7], which in turn were used
to simulate the trap stiffness in 4.2.

63
Appendix B

Correction of low-pass filtered


position data

Since the position-signal of a trapped object is determined using a CCD-camera


which requires an integration time significantly longer than the characteristic time-
scale of particle motion τ0 , the result has to be corrected for low-pass filtering1 .
The measured signal O(t) will be a convolution of the original signal P (t)
and the response M (t) of the detector:

O(t) = P (t) ⊗ M (t), (B.1)

or, in frequency domain:


Ô(ν) = P̂ (ν) · M̂ (ν). (B.2)

Ô, P̂ , M̂ denotes the Fourier-transform of O, P , M , respectively.


In our case, M (t) is assumed to be a square window with the width being the
integration time D and a height of 1/D. Thus, the unfiltered signal is in principle
available when the frequency spectra of the filtered signal and the detector response
1
The following is from a personal communication with A. Jonáš, Dec. 1, 2005

64
function are known. This is not the case here as the video rate is too low to sample
particle motion sufficiently. However, the complete original signal is not required,
since we are only interested in obtaining the original mean square fluctuations hO2 i
from the measured mean square fluctuations hP 2 i. These are related to their power
spectral densities sO (ν) and sP (ν):

Z ∞ Z ∞
2
hO i = sO (ν)dν = |Ô(ν)|2 dν (B.3)
0 0
Z ∞ Z ∞
hP 2 i = sP (ν)dν = |P̂ (ν)|2 dν. (B.4)
0 0

The relation between sO (ν) and sP (ν) is sO (ν) = sP (ν) · |M̂ (ν))|2 , obtained from
eq. B.2. According to the Wiener-Kinchin theorem, yet another Fourier-transform
yields the relation between the autocorrelation functions rO (t), rP (t) and rM (t)

Z +∞
rO (t) = rP (t) ⊗ rM (t) = rP (t − τ )rM (τ )dτ. (B.5)
−∞

This is convenient, because

kB T
 
2
rP (t) = hP i exp − 2 |t| (B.6)
hP iβ

is known for the thermally driven overdamped harmonic oscillator a trapped particle
represents. rM (t) can be computed easily, it is the autocorrelation of a square win-
dow. The problem is solved, if an hP 2 i is found that fulfills the combined equations
B.5 and B.6. By computing the convolution integral in eq. B.5 one arrives at the
desired relation between the mean square fluctuations hO2 i and hP 2 i:

1
   
hO2 i = 2hP 2 i2 A 1 − AhP 2 i 1 − exp − , (B.7)
AhP 2 i

β
with A = kB T D . This equation has to be solved for hP 2 i numerically.
Additionally, for motion close to a surface as it happens for trapping in lower

65
antinodes, β = 6πηa does not hold due to hydrodynamic coupling between particle
and surface. To correct for this, the expressions for β given in [15] are used.

66
Bibliography

[1] A. Ashkin. Acceleration and trapping of particles by radiation pressure. Phys.


Rev. Lett., 24:156–159, 1970.

[2] A. Ashkin. Forces of a single-beam gradient laser trap on a dielectric sphere in


the ray optics regime. Biophys. J., 61:569–582, 1992.

[3] A. Ashkin and J. M. Dziedzic. Optical trapping and manipulation of viruses


and bacteria. Science, 235:1517–1520, 1987.

[4] A. Ashkin, J. M. Dziedzic, J. E. Bjorkholm, and S. Chu. Observation of a


single-beam gradient force trap for dielectric particles. Opt. Lett., 11(5):288,
1986.

[5] J. P. Barton and D. R. Alexander. Fifth-order corrected electromagnetic field


components for a fundamental Gaussian beam. J. Appl. Phys., 66:2800–2802,
1989.

[6] J. P. Barton, D. R. Alexander, and S. A. Schaub. Internal and near-surface


electromagnetic fields for a spherical particle irradiated by a focused laser beam.
J. Appl. Phys., 64:1632–1639, 1988.

[7] J. P. Barton, D. R. Alexander, and S. A. Schaub. Theoretical determination of


net radiation force and torque for a spherical particle illuminated by a focused
laser beam. J. Appl. Phys., 66:4594–4602, 1989.

67
[8] S. M. Block. Making light work with optical tweezers. Nature, 360:493–495,
1992.

[9] S. M. Block, L. S. B. Goldstein, and B. J. Schnapp. Bead movement by single


kinesine molecules studied with optical tweezers. Nature, 348:348–352, 1990.

[10] W. I. Bragg and A. B. Pippard. The form birefringence of macromolecules.


Acta Cryst., 6:865–867, 1953.

[11] S. Chu, J. E. Bjorkholm, A. Ashkin, and A. Cable. Experimental observation


of optically trapped atoms. Phys. Rev. Lett., 57:314–317, 1986.

[12] L. Cognet, G. S. Harms, G. A. Blab, P. H. M. Lommerse, and T. Schmidta. Si-


multaneous dual-color and dual-polarization imaging of single molecules. Appl.
Phys. Lett., 77(24):4352–4354, 2000.

[13] J. E. Curtis, B. A. Koss, and D. G. Grier. Dynamic holographic optical tweezers.


Opt. Com., 207:169–175, 2002.

[14] I. Golding and E. C. Cox. RNA dynamics in live Escherichia coli cells. PNAS,
101(31):11310–11315, 2004.

[15] J. Happel and H. Brenner. Low Reynolds Number Hydrodynamics. Springer,


1995.

[16] S. Hohng and T. Ha. Single-Molecule Quantum-Dot Fluorescence Resonance


Energy Transfer. Chem. Phys. Chem., 6:956–960, 2005.

[17] P. Jákl, M. Šerý, J. Ježek, A. Jonáš, M. Liška, and P. Zemánek. Behavior of


an optically trapped probe approaching a dielectric interface. J. Mod. Opt.,
50(10):1615–1625, 2003.

[18] A. Jonáš. Use of standing electromagnetic wave for manipulation of micron


and submicron-sized objects. PhD thesis, Brno University of Technology, 2001.

68
[19] A. Jonáš, P. Zemánek, and E.-L. Florin. Single-beam trapping in front of
reflective surfaces. Opt. Lett., 26(19):1466–1468, 2001.

[20] M. J. Lang, P. M. Fordyce, A. M. Engh, K. C. Neuman, and S. M. Block. Simul-


taneous, coincident optical trapping and single-molecule fluorescence. Nature
Methods, 1(2):1–7, 2004.

[21] J. Leach, G. Sinclair, P. Jordan, J. Courtial, M. J. Padgett, J. Cooper, and Z. J.


Laczik. 3D manipulation of particles into crystal structures using holographic
optical tweezers. Opt. Exp., 12(1):220–226, 2004.

[22] T. H. Maiman. Stimulated optical radiation in ruby. Nature, 187:493–494,


1960.

[23] K. C. Neuman and S. M. Block. Optical trapping. Rev. Sci. Inst., 75(9):2787–
2809, 2004.

[24] R. Oldenbourg, E. D. Salmon, and P. T. Tran. Birefringence of single and


bundled microtubules. Biophys. J., 74:645–654, 1998.

[25] R. Omori, T. Kobayashi, and A. Suzuki. Observation of a single-beam gradient-


force optical trap for dielectric particles in air. Opt. Lett.., 22(11):816–818,
1997.

[26] J. P. Pawley, editor. Handbook of biological confocal microscopy. Plenum Press,


1995.

[27] K. Berg-Sørensen and H. Flyvbjerg. Power spectrum analysis for optical tweez-
ers. Rev. Sci. Inst., 75(3):594–612, 2004.

[28] M. Rief, F. Oesterhelt, B. Heymann, and H. E. Gaub. Single molecule force


spectroscopy on polysaccharides by atomic force microscopy. Science, 275:1295–
1297, 1997.

69
[29] A. Rohrbach. Stiffness of optical traps: Quantitative agreement between ex-
periment and electromagnetic theory. Phys. Rev. Lett., 95:168102, 2005.

[30] S. Sato and H. Inaba. Optical trapping and manipulation of microscopic parti-
cles and biological cells by laser beams. Opt. Quantum. Electron., 28(1):1–16,
1996.

[31] K. Schütze and A. Clement-Sengewal. Catch and move – cut or fuse. Nature,
368:667–669, 1994.

[32] M. Speidel, A. Jonáš, and E.-L. Florin. Three-dimensional tracking of fluo-


rescent nanoparticles with subnanometer precision by use of off-focus imaging.
Opt. Lett., 28(2):69–71, 2003.

[33] K. Svoboda and S. M. Block. Biological applications of optical forces. Ann.


Rev. Biophys. Biomol. Struct., 23:247–285, 1994.

[34] K. Svoboda and S. M. Block. Optical trapping of metallic Rayleigh particles.


Opt. Lett., 19(13):930–932, 1994.

[35] K. Visscher and G. J. Brakenhoff. Single beam optical trapping integrated in


a confocal microscope for biological applications. Cyt., 12:486–491, 1991.

[36] K. Visscher, G. J. Brakenhoff, and J. J. Krol. Micromanipulation by ”multiple”


optical traps created by a single fast scanning trap integrated with the bilateral
confocal scanning microscope. Cyt., 14:105–114, 1993.

[37] K. Visscher, S. P. Gross, and S. M. Block. Construction of multiple-beam


optical traps with nanometer-resolution position sensing. IEEE J. Sel. Top.
Quantum E, 2(4):1066–1076, 1996.

[38] D. L. J. Vossen, A. van der Horst, M. Dogterom, and A. van Blaaderen. Op-
tical tweezers and confocal microscopy for simultaneous three-dimensional ma-

70
nipulation and imaging in concentrated colloidal dispersions. Rev. Sci. Inst.,
75(9):2960–2970, 2004.

[39] M. D. Wang, H. Yin, R. Landick, J. Gelles, and S. M. Block. Stretching DNA


with optical tweezers. Biophys. J., 72:1335–1346, 1997.

[40] P. Zemánek, A. Jonáš, P. Jákl, J. Ježek, M. Šserý, and M. Liška. Theoretical


comparison of optical traps created by standing wave and single beam. Opt.
Comm., 220:401–412, 2003.

[41] P. Zemánek, A. Jonáš, and M. Liška. Simplified description of optical forces on


a nanoparticle in the gaussian standing wave. J. Opt. Soc. Am. A, 19(5):1025–
1034, 2002.

[42] P. Zemánek, A. Jonáš, L. Šrámek, and M. Liška. Optical trapping of Rayleigh


particles using a Gaussian standing wave. Opt. Commun., 151:273–285, 1998.

71
Vita

Matthias Reuss was born on July 25, 1981 in Munich, Germany, to Bernhard and An-
gelika Reuss. He received his high-school degree (“Abitur”) from the Alexander-von-
Humboldt-Gymnasium Schweinfurt in 2000 and completed military service there-
after. In 2001 he enrolled in the physics program at the Julius-Maximillians-
Universität at Würzburg, where he passed the pre-diploma exams (“Vordiplom”)
in 2003 with “sehr gut” (grade A). After another year of studies in Würzburg, he
entered Graduate School at the University of Texas at Austin in fall 2004.

Permanent Address: An den Huehneraeckern 1


97456 Dittelbrunn
Germany

This thesis was typeset with LATEX 2ε 2 by the author.

2 A
LT EX 2ε is an extension of LATEX. LATEX is a collection of macros for TEX. TEX is a trademark
of the American Mathematical Society. The macros used in formatting this thesis were written by
Dinesh Das, Department of Computer Sciences, The University of Texas at Austin, and extended
by Bert Kay, James A. Bednar, and Ayman El-Khashab.

72

You might also like