Sei sulla pagina 1di 17

Optical Trapping

Giovanni Pinuellas
8 November 2010

Abstract
In this experiment investigate the process of trapping silical beads
using an optical trap comprised of a 845nm wavelength (near-infrared)
laser beam with a maximum power of 200mW. We trap dielectric particles
in the focus of this laser beam and are able to detect small displacements
of the silica beads by using a photodiode and calibrating our equipment
to quantify size, position and forces at nm and pN scales. We also study
E. coli locomotion by investigating the relationship between cell rotation
and cell size. We also study active transport by motor proteins in onion
cells by investigating vesicle velocity.

1 Introduction
An optical trap, or optical tweezers, is a device used to apply piconewton-sized
forces on dielectrics and make very precise measurements. It is created using
radiation pressure from a laser beam. It has immense biophysical applications
and has been used for many ground-breaking experiments. For example, it has
been used for monitor the activities of motor proteins such as myosin, which
take part in the active transport of vesicles in onion cells. Optical tweezers have
also been applied to such manipulations as cell sorting and unzipping DNA. In
this experiment we use an atom trap to manipulate silica beads and measure
how precise our optical trap can be. We also investigate E. Coli cell rotation and
try to find a corrolation with cell size. Finally, we investigate active transport
in onion cells by manipulating vesicles carried by the cell’s motor protein.

2 Theory
The optical trap works by attracting a dielectric particle to the focus of a laser
beam. At the “waist” of the focus of the laser beam there is a gradient force
that tends to push a dielectric particle where the light intensity is the strongest.
This gradient force is produced from the momentum transfer as photons hit the
dielectric particle. The following diagram shows the forces that arise from a
gradient force due to varying laser intensity.

1
Figure 1: Ray diagram of optical trap.

In Figure 1 (a) the bead is slighly displaced to the left and up from the
center. As beams of light hit the bead, it refracts though the bead and bends
inwards, finally refracting again and leaving the bead. The reactionary force
is created by the momentum carried by the laser beam and is proportional to
its intensity. In this example we show two beams of light. Beam 2, because
it is closer to the center of the laser, has a higher intensity than beam 1 (and
is represented by a thicker line). Therefore its reactionary force, F2 , will be of
larger intensity than F1 . Thefore the total gradient force, Fnet , is mostly due to
the contribution of the beams of higher intensity, so its direction points to the
right and down.
In (b) the bead is displaced laterally upwards but is centered where the
intensity of the two beams are equal. Here the gradient force will push the bead
downward. In both (a) and (b) the bead will equilibriate to the center of the
trap, which is the center of the focus of the laser beam.

3 Apparatus
3.1 Laser Beam Path
The optical trap consists more of just a trapping laser. It starts out with a
Lumics LU0845 M200 diode laser that generates 845 nm wavelength light at a
maximum power of 200 mW. It first passes though collimating lens (that serves
to better align it to the objective lens of the microscope).
It passes through a beam expander, which expands the diameter of the
laser beam path to fit the aperture of the microscope objective. It is then
reflected by a dichroic mirror. In the dichroic mirror there are different layers
of coating with different refractive indexes, which if applied correctly, produce

2
phased relections caused by the interference of these layers. Therefore we can
have IR wavelength to reflect while light from the illuminator to pass through.
Since white light is made up of different wavelengths, blue light is used to
illuminate the beads, as our dichroic mirror can be made to only pass blue
wavelengths while rejecting others, thus being able to reflect IR light.
From the reflection of the dichroic mirror we pass through the objective
lens, which focuses our laser beam to a small waist in the vacinity of our speci-
men. It is here where we will do our optical trapping. It then expands again and
is recollimated by the condenser lens. The beam is then reflected by another
dichroic mirror and then is focus into the surface of a quadrant photodiode
(QPD).

3.2 Visible Light Path


The visible light path is simpler. It starts on top of the illiminator, where a
blue LED is collimated by a collimating lens, then it passes through the dichroic
mirror (it is not reflected). It is then focused on the specimen. Here the specimen
is held on the stage by an inverted microscope slide. There is a gap between
the bottom of the slide (the coverslip) and he objective lens. We fill this gap
with immersion oil that matches the refractive index of the glass, which vastly
improves the resolution of the lens. The objective lens magnifies the image,
which then passes through the dichroic mirror and is magnified again into a
CCD camera in order to record the image.
Figure 2 shows the basic paths for both the laser beam path and the visible
light path.

3.3 The Stage


In this experiment we will need to manipulate small horizontal movements of
the stage, including movement by joystick or programmable software. There
are two piezoelectric motors (used to reduce vibrations and noise) that drive
the stage in the X direction (left and right) and Y direction (towards and away
from us). Vertical movement can be achieved by moving the objective lens up
and down. It moves both the place of focus and the focus of the trap.

3.4 The Quadrant Photodiode (QPD)


We use the QPD to directly collect data on the position of the trapped particle
relative to the center of the trap. The QPD consists of 4 photodiodes in a
quadrant formation, which produce a certain output voltage dependent on the
intensity of light incident upon it. The circuitry then outputs a voltage Vx
and Vy which are proportional to the actual X and Y position of the incident
beam (as we will determine in the position calibrations). When there is nothing
trapped (or when we have a bead perfectly trapped in the center), the laser
beam is tightly focused on the center of the QPD, with Vx and Vy signals of

3
Figure 2: Optical trap block diagram, including light ray paths.

zero. When a trapped bead moves slightly away from the center of the trap,
the laser spot moves on the QPD, causing Vx and Vy to vary accordingly.

4 Position and Stiffness Calibration of the QPD


We use silica beads 1 micron in diameter to calibrate the QPD due to their
symmetry and standardized characteristics, such as refractive index, size and
shape. We do two kinds of calibrations: one for position and one for stiffness.
In the position calibration we determine the sensitivity of our QPD. Since
the QPD records raw Vx and Vy data, we need a quantitity that will help us
convert volts to micrometers. Since the power of the laser will also affect the
light intensity incident on the QPD, the voltage responses also change. Thus
we also look at the relationship between power level (proportion to maximum
power) at four levels and sensitivity. This relationship is linear.
Forces, on the other hand, cannot be measured directly in an optical trap.
We can instead calculate it by using Hooke’s Law, F = −kx. We can measure
the trap stiffness, k, and its displacement from the center of the trap to calculate
forces. Therefore we also do a stiffness calibration of the QPD.

4
Figure 3: Silica beads 1 micron in diameter. In one trap we were able to capture
about 15 beads.

4.1 Stuck Bead Scanning Method


We use the Stuck Bead Scanning Method for position (sensitivity) calibration.
Here the goal is to program the movement of the microscope’s stage so as to
move a stuck bead through the center of the trap. We scan using the QPD to
measure the responses in Vx and Vx.
The bead is successfully stuck to the bottom of the coverslip by means of
diluting our bead stock with a 1 M NaCl solution. Exposing the beads to this
NaCl solution causes them to stick to the glass by hydrophobic interactions.
The QPD measures the Vx and Vy responses, and if we plot voltage versus
time, the slope of the linear region of our plot can be converted to sensitivity.
Figures 4 and 5 show our raw data for the voltage read from the QPD, in the
x-direction and the y-direction, respectively, at the indicated power proportion.
The linear portions of the middle of the graphs are what interests us. We
perform a least-squares regression on the middle linear portion of these graphs
individually and calculate the slopes of these linear fits as well as the error on
this slope. An example for a stuck bead measurement in the x-direction at
power proportion level 0.249 is shown in Figure 6:
The slope, s, of the Vx vs time here is −0.0012201 ± 8.1883 × 10−6 in units
V /ms. This particular linear fit has χ2 = 0.84. We convert V /ms to V /µm
by exploiting the fact that we take 150 steps (to the left) per second. Previous
stage calibration performed by Dr. Eric Ayars in September 2009 reported that
movement of the stage to the left (with apparent bead motion to the right) found
the average step size to be 23.79 nm/step (with no reported error). Therefore
it is trivial to determine sensitivity, ρ, with its new error.
Figures 7 and 8 show all linear regressions for Vx and Vy vs time for our
stuck bead calibration.
Tables 1 and 2 chart all slopes and sensitivities for the x-direction and the
y-direction using the stuck bead scanning method:

5
Figure 4: QPD Voltages, x-direction.

Figure 5: QPD Voltages, y-direction.

6
Figure 6: Detailed example of the linear regression of the central region of the
Vx vs time plot.

Figure 7: Linear regressions for the x-direction.

7
Figure 8: Linear regressions for the y-direction.

x-direction s σs χ2 ρx σρx
Power level 0.249 0.0012 8.19 × 10−6 0.84 0.3419 0.0023
Power level 0.501 0.0028 1.45 × 10−5 0.89 0.7487 0.0041
Power level 0.767 0.0042 2.03 × 10−5 0.89 1.1734 0.0057
Power level 1.000 0.0060 2.02 × 10−5 0.91 1.6593 0.0057

Table 1: Stuck Bead Calibration in the x direction.

The identical method was used for the linear regressions in the y-direction.
But here, in order to convert V /ms to V /µm we use the fact that we took 150
steps upwards per second. According to the previous stage calibration, this
corresponds to an average step size of 34.23 nm/step. We then calculate the
corresponding sensitivites in the y-direction at the indicated power and report
then in Table 2.
We now test for the linearity of our sensitivities, as seen in Figures 9 and 10.
Indeed, sensitivity has a linear relationship with power. As the power is
increased, we lose a little sensitivity in our measurements. We can detect smaller

y-direction s σs χ2 ρy σρy
Power level 0.249 0.0017 1.53 × 10−5 0.81 0.3347 0.0030
Power level 0.501 0.0035 2.99 × 10−5 0.89 0.6858 0.0058
Power level 0.767 0.0056 4.14 × 10−5 0.92 1.0833 0.0086
Power level 1.000 0.0077 5.27 × 10−5 0.92 1.4996 0.0103

Table 2: Stuck Bead Calibration in the y direction.

8
Figure 9: Linearity of X Sensitivity.

Figure 10: Linearity of the Y Sensitivity.

9
changes in displacement with a smaller power setting.

4.2 Power Spectrum Method


We use the Power Spectrum Method to calibrate for both sensitivity and stiff-
ness. Here we can set our QPD to measure Power Spectrum Density (PSD)
data at four power levels. The PSD of a trapped bead had a Lorentzian profile
described by:
ρ2 k b T
Sυυ =
π 2 β (fo2 + f 2 )
in units V 2 /Hz, where β = 3πηd is the drag, with η being the fluid viscosity
of water, d the dimater of the bead and ρ the sensitivity of the trap. This
Lorentzian can be recast as log Sυυ (f ) = log α −log(fo2 −f 2 ). We then minimize
the least squares distance of every point from the predicted curve to give the
parameters α and fo (the rolloff frequency). Yielding α, we can extract ρ, the
sensitivity, from it. Furthermore, with fo and the relation k = 2πfo β, we can
obtain trap stiffness.
The computer program originally collected 4/3 seconds of voltage data recorded
by the QPD ten times, then computing the average power spectral density func-
ton over those ten trials, which is supposed to reduce the variance. This would
have been all fine and dandy, but it is useless for us when we only receive the
average data and not the original ten raw points, making error analysis difficult.
What I did instead was take the average data points given by the computer and
average each ten points successively again, yielding another average point as
well as its sigma. This allowed calculating χ2 a little easier after careful editing
of the several MATLAB scripts.
The log Sυυ (f ) curves and their fits are shown below for each direction and
power level. We conducted this measurement for four different power levels and
using two beads.
For our first bead we plot log Sννx and log Sννy in Figures 11 and 12, along
with the nonlinear regression contants here:
Bead 1 αx αy fox foy χ2P SDx χ2P SDy
Power level 0.253 0.0042 0.0096 188.53 191.24 0.57 0.66
Power level 0.504 0.0097 0.0280 191.48 284.15 0.60 0.95
Power level 0.778 0.0209 0.0662 264.36 403.50 0.50 0.71
Power level 1.000 0.0328 0.0950 255.75 407.24 0.60 0.62
For our second bead we also plot log Sννx and log Sννy in Figures 13 and 14,
as well as the constants below.
Bead 2 αx αy fox foy χ2P SDx χ2P SDy
Power level 0.256 0.0048 0.0061 128.58 130.65 0.46 0.58
Power level 0.508 0.0164 0.0220 217.70 235.31 0.70 0.63
Power level 0.766 0.0352 0.0473 301.51 321.63 0.65 0.56
Power level 0.997 0.0592 0.0818 409.34 442.76 0.47 0.51

10
Figure 11: PSD data, x direction (bead 1)

Figure 12: PSD data, y direction (bead 1)

11
Figure 13: PSD data, x direction (bead 2), first power level

Because there was a lot of noise in the beginning of the plot (most likely due
to vibrations of the table), we had to obtain the α and fo constants seperately
by cutting the beginning frequencies. Figures 13 and 14 are only for the first
power level, but they are representative of all other PSD plots.
This yields the following plots for sensitivity and stiffness (k in pN/µm):

Bead 1 ρx ρy kx ky
Power level 0.253 0.2907 0.4384 9.9362 10.0791
Power level 0.504 0.4417 0.7505 10.0917 14.9758
Power level 0.778 0.6484 1.1540 13.9328 21.2633
Power level 1.000 0.8123 1.3824 13.4790 21.4631

Bead 2 ρx ρy kx ky
Power level 0.258 0.3115 0.3510 6.7764 6.8857
Power level 0.508 0.5739 0.6656 11.4733 12.4017
Power level 0.766 0.8421 0.9752 15.8907 16.9511
Power level 0.997 1.0913 1.2830 21.5735 23.3348

4.3 Equipartition Method


We can use the Equipartition Theory to also calculate stiffness. It uses the
relation between particles’ energies and temperature. We can relate the thermal
energy of a system to its instantaneous displacement of small particles with the
formula 12 kb T = 12 kx < (x− < x >)2 >. Here < x > denotes the average over x,
the position of the particle along the x direction. Since < (x− < x >)2 > is the

12
Figure 14: PSD data, y direction (bead 2), first power level

variance of x, stiffness can be calculated from the variance of the bead position
at each power level. We use the previous stuck bead calibration of sensitivity to
convert QPD equipartition data to displacements, and we use the Equipartition
theory to solve for the stiffnesses kx and ky . Here we used the same Bead 2 as
in the PSD calibration, as well as another bead.

Bead 2 kx ky Bead 3 kx ky
Power level 0.253 5.564 6.019 Power level 0.258 5.029 3.867
Power level 0.504 12.224 15.743 Power level 0.508 14.957 12.352
Power level 0.778 16.734 19.327 Power level 0.766 26.640 25.773
Power level 1.000 23.238 25.754 Power level 0.997 28.847 29.411

5 Investigating Flagellar Locomotion in E. coli


5.1 Methods
In this part we investigate flagellar locomotion in Escherichia coli (E. Coli ).
Specifically, we try to study the corrolation between E. coli cell length and
frequency. Movement of E. Coli is caused by a set of helical flagellar filaments.
Movement is actually not random but modulated by chemical, temperature or
light gradients. Its first type of moevement is called a “run” which propels the
E. Coli in a straight path. But changes in its environment can stimulate a
“tumble” that causes an erratic motion in a new direction at random. This
change is caused by sensing that it is headed into unfavorable environmental
conditions, and thus prefers to run to a better place by tumbling more frequently

13
Figure 15: A single E. Coli cell about 10.5 µm in length.

if the environment gets worse, or less frequent if it is better. This movement is


governed by a rotary motor protein.
With our optical trap technique we can use our position calibration to mea-
sure QPD responses of trapped E. Coli. Looking at the QPD responses we can
detect cell rotation by measuring the frequency. We then try to investigate if
there is any relation between cell size (actually the length of the E. Coli ) and
cell rotation.

5.2 Results
We painstakingly followed E. Coli swimming around the slide and trapped them
all at a power setting of 50%. We recorded QPD data to look at the frequency of
cell rotation and took pictures of each cell to determine their size. Cell size was
determined by a previous screen pixel size calibration, where 1µm=14.8±0.2
pixels.
We found that typically a power setting of at least 2.5% was needed to
capture E. Coli, although bigger E. Coli were able to escape and move to
another plane.
Below is a representative plot of QPD responses for the E. Coli pictured in
figure 15.

14
The plot y axis is converted voltage to displacement data using the average
of our position calibration. The red data was taken from Vy data, while the
blue data was taken from Vx data. We can see in the red plot a flat portion
which is actually a tumble of the cell. The phase difference demonstrates the
cell’s rotation. We take Fourier transforms of this data to calculate cell body
and flagellar rotation frequency. We took data for more than 30 cells but chose
the ten best ones for our investigation.

E. Coli Cell length (µm) Body Rotation (Hz) Flagellar Rotation (Hz)
1 6.76 3.5 58
2 2.92 3.4 58
3 4.79 3.6 59
4 5.00 2.3 58
5 5.67 4.5 58
6 7.63 3.8 58
7 10.47 3.6 58
8 6.13 5.1 59
9 4.80 3.3 58
10 16.08 2.4 59

There seemed to be no corrolation of cell body rotation with cell size, but
flagellar rotation was very consistant.

6 Investigating Internal Transport in Onion Cells


In this part of the experiment we measure the motion (internal transport) of
vesicles inside an onion cell. Vesicles are used to transport food, waste and
information around the cell. Vesicles themselves are transported by myosin,
a motor protein, along tracks called actin filaments. We try to measure the
velocity of the active transport by setting our laser at the lowest possible settings

15
so as to not trap vesicles. This still produces a response in the QPD, and with
the slopes of the plots we can calculate their velocities.
The next figure shows the path of vesicles along the invisible actin filaments.

Vesicles are easily trapped but not below a power of about 25%. Free vesicles
can be successfully trapped and moved around the cytoplasm. Once a vesicle
was trapped for a while and released, it took some time to regain its path.
Interestingly, while one vesicle was trapped we continued to trap about 2 more
vesicles while the rest routed around the trap. The average velocity of vesicles
was measured to be 2.4µm/s. This was calculated by the movement as in the
above picture’s QPD Vx slopes.

7 Conclusion
This lab was exceedingly difficult, not in the way the data was taken, but it’s
analysis. More than anything what was primarily learned from this lab was how
to program Matlab.

16
8 References
[1] Arthur Ashkin, "Acceleration and trapping of particles by radiation pres-
sure", Physical Rev. Let. 24(4), 156-159 (1970).
[2] J. W. Shaevitz, “A practical guide to optical trapping”
[3] Arthur Ashkin, Optical trapping and manipulation of neutral particles using
lasers, Proc. Natl. Acad. Sci. 94,4853-4860 (1997).

17

Potrebbero piacerti anche