Approaching single molecule detection by laser-induced fluorescence of flowing dye solutions in a capillary


Material Information

Approaching single molecule detection by laser-induced fluorescence of flowing dye solutions in a capillary
Physical Description:
viii, 165 leaves : ill. ; 29 cm.
Lehotay, Steven John, 1965-
Publication Date:


bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1992.
Includes bibliographical references (leaves 159-164).
Statement of Responsibility:
by Steven John Lehotay.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001753105
notis - AJG6068
oclc - 26576147
System ID:

Full Text







UNIERmSITYO n a 1W L'!'~ U'' "

This dissertation is dedicated to my father, Andrew L. Lehotay. Though he

died, he still lives in me.


I think Albert Einstein once said something like, "If I have seen farther, it is

because I have been standing on the shoulders of giants." I may not be able to see

very far, but it is more of a problem with "visual acuity" than elevation. The "giants"

in my case are Chris Stevenson, Ramee Indralingam, and Tye Barber whose research

efforts have had direct bearing on this dissertation. I would like to thank them for

their work, otherwise, this thesis would not have been possible.

Most of all, I would like to thank Dr. James D. Winefordner for his guidance,

insight, and knowledge. I am grateful for the opportunity to have been one of his

students and will forever be amazed and inspired by his intelligence, diligence, and

personality. He brought this project out of the clutches of despair with the simple

placement of a black piece of construction paper between the metal vapor filter and

the monochromator.

I also sincerely thank Dr. Benjamin W. Smith and Dr. Giuseppe A. Petrucci

for their help, patience, and friendship through the months of research leading to this

dissertation. I cannot count the numerous times I turned to Ben for advice, and

Giuseppe spent much time with me aligning the Ti:sapphire laser. He also modeled

the focusing aspects in a capillary presented in the dissertation.

I am also grateful for the work of several others in the group: Mike Wensing

wrote the program that calculates the Voigt profiles for the metal vapor cells; Nancy

Petrucci took the Raman spectra of the solvents; Yuan-Hsiang Lee helped collect

some of the data; and Wellington Masamba and Dennis Hueber helped with the

diode array software and use of the HR1000. Furthermore, I should thank the entire

JDW research group for their input, companionship, and spirit. They have all helped

make life in graduate school as stimulating, rewarding, and fun as it has been.

I very much appreciate the monetary support granted me by the state of Florida

and Texaco during my graduate school years.

Finally, I would like to thank my wife, Joann, for being loving, supportive, and

understanding during the stressful times, and all other times, leading to this

dissertation. I am a lucky soul to have her with me.


ACKNOWLEDGEMENTS ...................................... iii

ABSTRACT ................................................ vii

CHAPTER 1 INTRODUCTION AND THEORY ..................... 1

Introduction ............................................. 1
Applications of Single Molecule Detection ................. 2
Choice of Analytical Technique for Single Molecule
Detection .................................... 6
Theory of Single Molecule Detection .......................... 8
Definitions ........................................ 9
Statistics of Data in Single Molecule Detection ............. 14
Theory of Laser-Induced Fluorescence ......................... 16
Sources of Noise in LIF and Means of Noise Reduction ........... 25
Laser Scatter ....................................... 25
Raman Scatter ..................................... 30
Background Fluorescence ............................. 30
History of Single Molecule Detection .......................... 32

PARAMETERS ......................................... 41

The Metal Vapor Filter .................................... 41
Theory of the Metal Vapor Filter ....................... 41
Choice of Metal for the Filter .......................... 43
Calculation of Spectral Linewidths and Absorbances for Rb .... 45
The Laser .............................................. 55
Criteria of the Laser fo Single Molecule Detection ........... 56
The Ti:Sapphire Laser .............................. 58
The Ti:Sapphire Laser/Rb Metal Vapor Filter Combination ... 62
The Sample ............................................. 65
Choice of Analyte .................................. 65
Choice of Solvent ................................... 75

Sample Containment ............
Optical Considerations Regarding the
Focusing the Laser ........
Collection of the Fluorescence
Detection ....................
Choice of the Detector .....
Photon Counting ..........
Control of the Sample Flow .......
Experimental .................

Capillary .

. . .

. . .

. . .
.. l.. .. .


Studies of the Metal Vapor Filter ............
Absorption Properties ................
Transmittance Properties ..............
Additional Spectral Filtering ................
Spectral Filters .....................
Polarization ........................
R results ................................
Limits of Detection ..................
Noises of the System .................
Discussion of Limits of Detection .............


Conclusions .............................
Future W ork ............................
Elimination of Scatter ................
Other Future Improvements ...........

REFERENCE LIST............................

BIOGRAPHICAL SKETCH .....................



Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



Steven John Lehotay

May 1992

Chairperson: James D. Winefordner
Major Department: Chemistry

The ability to detect single molecules in a solution has long been the ultimate

goal in ultratrace chemical analysis. In the strictest sense, single molecule detection

is defined as the efficient detection of a monomeric chemical species with near-100%

statistical certainty. Of current analytical techniques, laser-induced fluorescence has

the best chance of achieving single molecule detection due to its high sampling

efficiency, nondestructive probing, and high sensitivity. The limitations of laser-

induced fluorescence stem from noises associated with specular scatter from the

laser, Rayleigh and Raman scattering from the solvent, and background fluorescence

from sample contaminants. This dissertation concerns a new approach to eliminate

or greatly reduce these sources of noise for the purpose of single molecule detection.

In this approach, laser specular scatter is completely absorbed by a metal vapor filter,

which is simply a glass cell containing a metal element under reduced pressure.

When heated, the metal enters the vapor state and specifically absorbs the laser

scatter provided the laser spectral linewidth is narrower than the absorption band of

the metal vapor. Raman scatter from the solvent is reduced by containing the

sample in the small volume of a narrow capillary, and virtual elimination of

background fluorescence is accomplished by using near-infrared excitation with a

titanium:sapphire laser which specifically induces the fluorescence of a polymethine

dye. Currently, the Ti:sapphire laser is the only source with the power, tunability,

and narrow spectral linewidth that can be used in conjunction with a metal vapor

filter for single molecule detection. Rubidium was the element chosen for the metal

vapor filter due to its low melting point and ground state absorption transitions at

780.02 and 794.76 nm. In this dissertation, the theory of single molecule detection

and a review of previous approaches are presented, and the development and results

of this new approach are discussed. The lowest detection limit attained was 800

molecules of the dye, IR 140, in methanol flowing through a 140 pL probe volume.

Although single molecule detection was not achieved, the theoretically possible

detection limit of a single molecule could be attained through future experimental




The singular purpose of this research is to detect a single molecule in a

chemical solution. If one defines single molecule detection (SMD) as being able to

determine individual molecules in a sample containing billions of undetected

molecules of that species, the goal of this research has already been accomplished

in many instances by others. However, in order to fulfill the requirements for

practical application of SMD, the definition must be expanded to require the

individual detection of nearly every molecule of the analyte contained in the sample

solution. SMD remains elusive by this definition, and if SMD is to be truly realized,

this definition must be met.

Obviously, the achievement of SMD is no simple task. It has taken many

years of experimentation and instrumental refinements for other researchers to reach

near-SMD detection limits. In this project, it is hoped that with a thorough

understanding of the theory and careful design and implementation of experimental

components, the approach described in this dissertation will be able to achieve SMD

in its initial attempts.

The purpose of this dissertation is to 1) present the theoretical criteria for

SMD; 2) review previous research on the subject; 3) discuss the development of a

novel approach to SMD; 4) show the results of experiments designed to attain SMD;

and 5) discuss these results and future research possibilities.

Applications of Single Molecule Detection

For the analytical chemist, being able to detect individual atoms and

molecules in a sample is a worthwhile goal in itself, and indeed, it constitutes the

goal of this project. However, analytical chemists are not usually involved in the type

of research that could be labeled as "purely academic." As scientists, analytical

chemists often take pride in the practical nature of their work. Therefore, other than

the fundamental, inherently worthwhile aspect of pursuing SMD, there are several

important potential applications of this research.

DNA sequencing. In consideration of the highly publicized project to decode

the human DNA sequence, Jett et atL at Los Alamos National Laboratory have

proposed a strategy for rapid DNA sequencing relying on the detection of individual

nucleotides. This strategy involves specifically tagging the four different DNA (or

RNA) nucleotides with four different highly fluorescent dye molecules, then detecting

each of these molecules as they pass through a volume probed by a tightly focused

argon ion laser. If successful, the scheme is projected to sequence DNA 1000 times

faster than current techniques.' Although it remains to be seen if this application can

be accomplished as proposed, research concerning the detection aspects of the


project have come very close to SMD.2" This research at Los Alamos will be

discussed in more detail later in this chapter.

Immunoassays. Like DNA sequencing, the immunological assay has become

a very prominent analytical technique in biotechnology.7 Immunoassay is a general

type of technique that relies on antigen:antibody interaction for specific analyses of

many types of biological species. Currently, the technique requires considerable time

for an organism to produce sufficient antibody for an assay. If these type of analyses

could be performed on a molecular scale, much time and trouble could be saved, and

potentially more information of biological systems could be obtained. SMD in an

immunological system would enable the study of fundamental biological processes

at the molecular level.

Flow cytometry. Flow cytometric methods are another large area of research

in the biological sciences.8 In fact, the DNA sequencing strategy mentioned above

relies on a sheath flow cuvette, the main component of a flow cytometer, to contain

the sample.1 Currently, the ability to rapidly detect and sort single cells through flow

cytometric methods.8 In being able to detect single molecules as opposed to single

cells, flow cytometry would possibly allow separation of individual molecules.

Obviously, this would enable some very interesting studies to be performed.

Tracer studies. It is common practice in the determination of environmental

flow patterns, fates of chemicals in the environment, and metabolic pathways in

organisms to add a tracer compound into the system and follow its path. Due to

dilution effects, decomposition, and other losses, the analysis at some removed time


and place from the injection point may require the detection of the tracer in very

small quantities, potentially single molecules. These kinds of studies could be carried

out in a system on a global scale (such as the study of ocean currents or wind

patterns) or on a microscale (such as cell transport).

Fundamental research. For many physicists and physical chemists, it would

be interesting to determine whether the characteristics of an individual molecule

match its bulk properties. In the case of fluorescence, questions could be answered

concerning fluorescence lifetimes, quantum yields, quenching, and environmental

effects. Experiments involving SMD in solution may uncover important

physicochemical effects on a submolecular scale.

Detection in capillary electrophoresis and microcolumn chromatography. In

the 1980s, Jorgenson's group made a major advance in separation science with the

development of capillary zone electrophoresis (CZE).9 CZE is capable of achieving

separation efficiences as large as 106 theoretical plates and allows for the separation

of several types of chemicals in aqueous solution, particularly biological species, that

were not easily separated previously.10 Similarly, it has been known for some time

that decreasing column diameter (or particle size) in liquid chromatography increases

separation efficiency." With the recently developed commercial ability to coat very

narrow capillaries with stationary phase material, microcolumn high-performance

liquid chromatography (HPLC) has become a viable method for separation analysis.

However, due to the small quantities involved in microcolumn HPLC and CZE,

detection is often the limiting factor in applications of these techniques. By their

nature, separation methods dilute the original analyte concentration, and for

complicated separations requiring a great deal of time, the analyte concentration at

the detector may be a small fraction of the original value. For many sample limited

applications, very small amounts of a particular analyte may be separated by

electrophoretic or chromatographic methods, but cannot be measured due to

detection problems. In essence, for separation techniques utilizing capillaries, SMD

may be required for some separations; however, very low detection limits are

required for all of these separations.

Due to the great potential of capillary separation techniques and their need

for sensitive detectors, a great deal of research has been undertaken concerning

detection of solutions flowing in capillaries. The most promising detection method

has been laser-induced fluorescence utilizing fluorescent probe molecules tagged to

the analyte.2 Based on the use of capillaries in both microcolumn chromatography

and capillary electrophoresis and the great applicability and needs of these methods,

a capillary was chosen as the sample container for this project. This point will

become more prominent later in this dissertation.

Unforeseen applications. In general, applicability of any technique depends

on whether the technique suits the specific needs of the application. In most cases,

the need for a type of analysis is the impetus for its development; this is only

partially true in this instance. As mentioned previously, the purpose of this project

is to achieve SMD. The experiment is designed to be useful for detection in capillary

separation techniques, and in the future could be used as such, but for the present,

SMD is an end in itself. If this project is successful, the announcement of SMD to

scientists in various disciplines would possibly cause the design of experiments to suit

the needs of the detector. It is difficult to predict what outcomes would arise by

accomplishing SMD, but the possibilities are very exciting.

Choice of Analytical Technique for Single Molecule Detection

Of all analytical methods currently in use, laser-induced fluorescence (LIF)

has the best chance of success for SMD in practice. Other techniques have been

considered, as discussed in the following paragraphs, but all were dismissed based on

theoretical or practical grounds.

All electrochemical methods in solution are incapable of rapid single molecule

detection due to the insufficient detection sensitivities to overcome the inherent

noises and response times associated with electrochemical measurements. Also,

molecular absorption spectrometry is eliminated on the fundamental basis that the

absorption cross section of a molecule is much smaller than the cross section of any

known light source beam. Similarly, Raman spectrometry also lacks sensitivity based

upon the small scattering cross-section." Several other types of spectroscopic

detection methods such as magnetic resonance, refractive index, thermal lensing,

photoacoustic spectroscopy, and all forms of thermally excited emission spectrometry

are also incapable of single molecule detection with current technology. Additionally,

radiochemical techniques possess insufficient sensitivity to detect a single molecule

on a practical basis.14


Mass spectrometry. On the other hand, many techniques utilizing mass

spectrometric detection"1 have sufficient signal-to-noise ratios to detect single

molecules and are routinely capable of detecting individual molecular ions striking

the electron multiplier. In this respect, mass spectrometry is capable of single

molecule detection, but as mentioned earlier, practical SMD must occur with near

100% sampling efficiency. Such is not the case with mass spectrometry because of

unavoidable sample losses during introduction into the ionizing chamber and during

transport and separation of ions by the mass spectrometer. Mass spectrometry would

be a viable, practical approach to SMD, with the added benefit of molecular

identification, if these sample losses can be eradicated. Unfortunately, these are very

difficult problems to solve and it is not expected that mass spectrometry will be

useful for true SMD in the near future.

Scanning tunneling microscopy. A technique capable of true SMD, but not

yet useful for practical application, is scanning tunneling microscopy (STM). STM

has been demonstrated to allow viewing the atomic structure of molecules such as

benzene," and this capability has been suggested as an approach to DNA

sequencing.1 Scientists have developed STM techniques to the point that they are

able to view and manipulate single atoms contained on a metal surface.'

Despite these remarkable advances, STM is impractical for SMD of large

samples and in solutions. The analytical procedure for treatment of a sample for

STM involves coating the solid with an electrically conductive layer (commonly gold).

This process is as much art as science and can be very time consuming and expensive.

Once the sample is coated, the analysis can also be time consuming in trying to view

single molecules on a large surface. Furthermore, data interpretation is often a

subjective process often criticized for the imposition of imaginative viewing by the


These problems are being addressed by current research, and if they can be

resolved, the potential for STM is enormous.19 Even so, STM has become a valuable

tool for the surface analysis of metals and many other solid materials, but it is not

currently practical as a general technique for SMD.20

Laser-induced fluorescence. As in mass spectrometry, lack of 100% sampling

efficiency plagues many atomic spectrochemical methods in flames and furnaces, such

as atomic fluorescence, laser-enhanced ionization, and resonance ionization.2122

However, the direct molecular fluorescence analysis in a solution is theoretically and

practically capable of detecting an individual highly fluorescent molecule. Laser-

induced fluorescence (LIF) is commonly the most sensitive fluorescence technique,

and for this and other reasons to be discussed in a later section in this chapter, LIF

is the chosen method in the attempt to attain SMD.

Theory of Single Molecule Detection

In recent years, since technology has been developed that is capable of

extremely low limits of detection, much discussion has appeared in the literature

concerning the requirements and statistics concerning detection of single species.21'

A useful summary of most of the concepts of SMD,28 as well as application of the

theory for the counting of atoms and molecules, has recently been given,' and

much of the following discussion is based on these writings.


Due to the many variables involved in any given analytical procedure, and the

many differences in those parameters when compared with other techniques,

acceptable terminology for SMD must be defined. Otherwise, the claims to SMD for

one type of analysis may not actually meet the requirements for another.

Single molecule. In general chemistry textbooks, a molecule, or compound,

is defined as, "a substance composed of more than one element, chemically

combined."3 By this definition, single molecule detection is accomplished by looking

at DNA under a microscope or touching a piece of plastic. This is unsuitable for the

ego of analytical chemists who perform ultratrace analysis, so for the purposes of this

dissertation, the word "molecule" has been modified to signify a monomeric

compound of reasonable size. For tagging purposes, "reasonable size" depends on

the application, but in general, a molecular weight of less than 1000 g/mol is

considered reasonable.

Single molecule detection. In classic papers, Alkemade23' stated the criteria

for the detection of individual species. These works mostly concerned single atom

detection (SAD), but apply also to SMD. He mentioned that SAD involves two basic

requirements: 1) an efficiency of detection of unity and 2) attainment of the intrinsic

noise limit. These two factors are defined below.

As in the case of defining a "single molecule," this definition must be modified

as well. In his papers, Alkemade indicates that SAD refers only to the spatially and

temporally probed region and does not account for sampling efficiency. By his

definition, achieving SAD by focusing a pulsed laser to a very small region of a flame

remains possible, despite that in this system, for each atom that is detected, hundreds

of atoms are not aspirated into the flame, thousands do not pass through the focus

volume, and thousands more pass unprobed during the time between laser pulses.

For the purposes of practical use of SMD in solution by LIF, more stringent

requirements for SMD are necessary. With this in mind, the third criterion for SMD

is that the sampling efficiency, E,, must be nearly 100%. By this definition, it is fair

to say that SAD/SMD has not yet been achieved.

Efficiency of detection. The first criterion of Alkemade for SAD is that the

efficiency of detection must be unity, or -d = 1. This means that each time that a

single species appears in the probe volume during the probe time, it must be

detected. In many cases, it is possible that single molecules can be detected, but only

a certain percentage of the times that a molecule is present is it detected. In these

cases, the researchers cannot make a valid claim to SMD.

Extrinsic and intrinsic noise. According to Alkemade's second condition for

SAD, extrinsic noise must be eliminated at which time intrinsic noise becomes

prevalent. Extrinsic noise is the background produced by external factors such as

stray light, thermal fluctuations, and electric and magnetic fields. These factors can


be virtually eliminated experimentally with the techniques to be described later in

this chapter.

Intrinsic noise is the result of fluctuations of the signal itself. Sources of

intrinsic noise include the noise of the detector and power fluctuations of the source.

These are inherent features of the detection process that can be greatly reduced, but

will prevail as the limiting source of noise in the absence of extrinsic noise.

Limit of detection. One of the major figures of merit for any analytical

technique is limit of detection (LOD). LOD is defined as the concentration at which

the signal is 3 times larger than the standard deviation of the blank (O), or

LOD = 3q,/sensitivity, (1-1)

where sensitivity is the linear slope of the analytical calibration curve of the detection

system. This definition was developed at a meeting of the International Union of

Pure and Applied Chemists (IUPAC) in 1976 to settle differences in the subjective

way in which detection limits were previously determined.3

In the signal domain, the measure of LOD is given the symbol X, which is

defined as,

Xd = bl + 3bb (1-2)

where Cbi is the mean signal of the blank. When using this expression, a calibration

curve is not necessary; the analyte concentration at which this criterion is met is the


Limit of guaranteed detection. Despite the IUPAC definition, LOD does not

always correspond to the best measure of lowest level of analyte detectability for a

system. In many circumstances, it is very difficult to actually observe a difference in

the signal at the limit of detection. Kaiser35 was aware of these problems and

introduced a term known as the limit of guaranteed detection (LOGD). Based on

statistical concepts, LOGD is set at twice the standard deviation requirement chosen

for LOD, which means

LOGD = 60b,/sensitivity, (1-3)

or in the signal domain,

X, = ibl + 6abl, (1-4)

where X, is the signal produced at the LOGD.

False positives and negatives. Another way of looking at detection in a

chemical analysis is the occurrence of false positives and false negatives. For

example, at the LOGD, the probability of the occurrence of a false positive is

essentially zero. A false positive, or type I error, occurs when the data exceed the

criteria for the detection of the analyte, but in actuality, noise, not signal, has been

the cause of the occurrence. False positives occur at a probability a. False

negatives, or type II errors, transpire when the analyte is present at a sufficient

concentration to be detected, but the signal does not exceed the detection level.

Type II errors occur with probability B.

In approaching the definition of detection limits based on the consideration

of types I and II errors, Xd (LOD in the concentration domain) corresponds to the

signal level that exceeds the background with a confidence of 1-a, and Xg (LOGD

in the concentration domain) is the signal level that gives a confidence of 1-f that


the analyte is actually being detected. Based on the definitions of LOD and LOGD,

the minimum confidence level is 99.86% which means that a and f must be 0.0014

or less.

Destructive and nondestructive probing. As mentioned previously, LIF has

a greater chance of success in realizing true SMD than other analytical methods.

The reason for this given earlier is the potential for high sampling efficiency by LIF.

However, an equally important factor is that LIF is a nondestructive probing method.

This means that the molecule is not destroyed in the detection process and more

than one detected event can occur per molecule. In fact, LIF is capable of producing

106 photons per fluorophore in a timespan of a few milliseconds which permits the

possibility of a higher noise level for an experiment that is still able to attain SMD.6

In destructive methods of detection, only one detection event can result from

each molecule. Destructive methods of detection include mass spectrometry,

radiometric methods, laser-enhanced ionization, and resonance ionization. In these

methods, the noise level must be extremely low, or signal of that one event must be

very high, to quantifiably detect a single molecule.

Symbols. Based on the definition of SMD, the laser must be continuous-wave

or very high repetition rate and probe the entire sample flow region. The focused

region of the laser is termed the probe volume, Vp, and the transit time an analyte

molecule in the Vp is the residence time, t,. During this interaction time, the number

of individual molecules) in the Vp, symbolized by Np, may give rise to a number of


detected events, N, (photoelectrons in the case of a photomultiplier tube). The mean

background level during t, is symbolized by jbI*

Statistics of Data in Single Molecule Detection

Poisson distribution. At low concentrations, Np follows a Poisson distribution,

and when photon counting is used for data collection at low levels, N, and ^bl also

follow a Poisson distribution. The Poisson probability distribution is given by'

P(X) = () ,(,), (1-6)

where P(X) is the probability of X events occurring (with X being Np, N0 or noise),

and t is the mean value of X. In Poisson distributions, the variance equals the mean

(o2 = I, where a refers to the standard deviation)" which applies at higher means

when Gaussian and Poisson distributions have a large overlap.28

In typical analytical measurements, the occurrence of noise, signal, and

numbers of analyte species in the detection region follow Gaussian probability

distributions and the expressions for LOD and LOGD were designed for these types

of analyses. However, in the case of SMD, a problem with the determination of

LOD and LOGD through the calibration curve method is that the slope of the

calibration curve at concentrations much higher than the detection limit, which are

Gaussian in nature, may lead to errors at near-SMD levels, which follow a Poisson


Criteria of signal and noise for SMD. Based on a Poisson probability

distribution,37 the values for Xd and X for a low-level counting experiment at various

background levels are given in Table 1. The table was constructed as follows:2

1) The Poisson probability distribution with a mean equal to the chosen

background level was found in reference 37. The detection limit, Xd, was

determined as the number of counts (minus background) at which the sum of

the remaining probabilities of the distribution beyond Xd did not exceed

0.0014 (confidence = 1-a or 99.86%).

2) A distribution was then found such that the sum of the probabilities greater

than Xd exceeded 0.9986 (1-3f). The mean of this distribution is X.

Table 1. Signal levels required to achieve single molecule detection with 99.86%
confidence at given mean blank levels.

Mean Blank Level Detection Limit Guaranteed Limit
(Ibb, in counts) (X, in counts) (XY, in counts)
a < 0.0014 1 f 0.0014

0.001 1 6.6

0.25 4 12.7

1 6 16

5 14 28

10 22 39

25 42 64

50 73 102

100 132 169

All parameters defined in text.


Based on these confidence levels, Table 1 gives the signal level required for

the realization of SMD at a given background level. Because of the differences in

practical aspects and statistics of data at near-SMD levels, Curie8" has advocated that

the confidence level be set to 99.5% (1.65a) rather than the IUPAC level. This

would significantly lower the values for X, and X, reported in Table 1.

The only remaining theoretical topic is whether the proposed LIF system is

able to meet the signal and noise levels presented in Table 1. These considerations

of LIF and its sources of noise are discussed in the next section.

Theory of Laser-Induced Fluorescence

Fluorescence is a physical phenomenon involving the absorption of a photon

of light by a molecule, causing an electron to climb to a vibronic level of an excited

singlet state, followed by emission of a photon of typically lower energy as the

electron returns to the ground vibronic level. Not all molecules undergo

fluorescence, and the ones that do often are highly conjugated and contain aromatic

functional groups. Fluorescence is useful as an analytical procedure because the

intensity of the emission is dependent upon concentration, and the excitation and

emission wavelengths of the light give some information as to the identity of the

fluorescent species. Due to factors to be discussed below, fluorescence is often a

very sensitive type of analysis.

As the name implies, laser-induced fluorescence (LIF) simply uses a laser as

the excitation source for fluorometric analysis. Lasers are able to produce higher

spectral irradiance (W/cm2) than common broad-band sources such as the xenon arc

lamp. Thus, LIF typically gives lower LODs for the analysis of fluorescent

compounds. Also, the narrow emission bandwidth of the laser is useful in many

situations that require selective excitation and detection of a fluorescent probe

species added to a system.

The practical formula that gives the average number of detected events, N.,

in the type of IUF system used for SMD is3

No = ( )a,Y,( -)?Tt, (1-7)
hv *S 4r

where ~L is the laser power (W), hvL gives the energy per laser photon (J) with h

being Planck's constant (6.636 x 10"' J-s) and VL being frequency of the light (Hz),

SL is the cross-sectional area of the focused laser beam (cm2), oa is the cross section

of absorption for the molecule (cm2), YF is the fluorescence quantum efficiency, Op

is the solid angle of collection of the fluorescence (sr), t is the cathodic efficiency of

the detector dimensionlesss), T is the transmittance of the optical components

dimensionlesss), and tr is the residence time of each molecule in the probe volume

(s). These parameters are given by the manufacturer (as in the case of q), easily

measured (tL, T, and S), referenced in the literature (YF), or calculated from other

known parameters (A, OF, and t,). The following paragraphs discuss the

determination of these parameters for the system to be described in this thesis.

The cross section of absorption. The absorption cross section, OA, is not as

much of an actual "size" of a molecule as it is a statistical quantity. The units of cm2

arise from the fact that irradiance (W/cm2) is used in the expression to determine

the availability of light. The value for oa is the probability that the light available in

a certain area is absorbed.

The simplest way to determine o^ for a molecule is to measure the absorbance

of a known concentration in solution. Assuming the concentration falls within the

region of linear response, the molar absorptivity, EA (M1cmH), at the chosen

wavelength can be found from Beer's law,

A = ^AC, (1-8)

where A is absorbance dimensionlesss), f is path length (cm), and C is concentration

(M). By knowing ^A, A can be found from,

aA = 1000eA/N, (1-9)

where N is Avogrado's number (6.02 x 103). The absorption cross-section can also

be determined from fundamental parameters,9 but this method was used in this

project based on its simplicity and its use of an experimental measurement. The

absorption coefficients for the dyes to be tested in this project (given in Chapter 2)

are 200,000 M-cm7': therefore, the values for oA are approximately 3 x 1016 cm2.

Fluorescence quantum yield. The quantum efficiency of fluorescence, or

quantum yield, Y,, is the probability of emission of a fluorescence photon once a

photon has been absorbed. In mathematical form, Y, is given by39

Y = k (1-10)

where kF is the rate of fluorescence (s'), and k, is the rate of nonradiative

deactivation (s-) of the excited singlet state (Si). Nonradiative decay of S, occurs

through the processes of external conversion (collisional deactivation), internal

conversion (nonfluorescent de-excitation), and intersystem crossing (S, to triplet,

T).39 These factors are difficult to quantify, and depend strongly upon the molecule

itself and environmental factors such as temperature, pressure, solvent, and presence

of other species. Therefore, the Y, for a particular system must be measured.

Basically three different methods are used to quantify Y,, the simplest and

most common of which is the Parker-Rees method.39 This involves the comparison

of the fluorescence emission of an unknown fluorophore, Y,,, with a fluorophore of

known quantum efficiency, YF,k. Assuming constant power of excitation, this method

uses the equation,

YF = YF,k EAk (1-11)

where the differences in the absorption coefficients and detection efficiencies (a

function of wavelength) for the known and unknown fluorophores must be taken into

account. The most commonly used reference fluorophore is quinine which has a

quantum yield of 0.59 in an acidic aqueous solution.


The Parker-Rees method is not always simple to use, and due to other

correction factors not included above, is not always accurate. An easier way to

determine YF for a particular system is to search the physical chemistry literature for

an accurate determination of Y, in the solvent to be used. THrough an intensive

literature search, it was found that that the dye has Yp 142

Solid angle of collection. A microscope objective was to be utilized for the

collection of fluorescence in this approach. The solid angle of collection, OF, can be

calculated from the stated specifications of the manufacturer and the distance of the

point source fluorescence emission to the collection optics. It should be stressed,

however, that erroneously large values for 0, result if one does not use the proper

equation for the calculation. The general expression normally used to calculate 0 for

a lens is given as9

,F = rtan20, (1-12)

where 0 is the angle defined by a line extended from the point source to the center

of the lens and a second line from the point source to the edge of the lens. The

problem with this equation is that it breaks down at large solid angles encountered

with microscope objectives. The calculated value of OF can exceed the true value by

a factor of 15% for 0 = 30* and by a factor of 300% for a 0 = 60*. The correct

expression for Ol is given by4

S= 27(1-cos0) = 4rsin 2(). (1-13)

With this equation, 0, is less susceptible to error than with the previous equation.

When using a microscope objective, 0 is found from the stated numerical

aperature, N.A. of the objective, in that

N.A. = n sine = -, (1-14)

where n is the refractive index dimensionlesss) of the medium between the object

and objective, 4 is the aperture (cm) and f is the focal length (cm) of the objective.

The calculated OF for the microscope objective to be used in this experiment is 1.5

sr which, when corrected for the 4r sr of a sphere, corresponds to a collection

efficiency of 11.9% of a point source. The magnification (40X) and N.A. (0.65) for

the microscope objective used in this project, as well as other considerations of the

collection optics, are discussed in Chapter 2.

Calculation of the expected LIF signal. Now that the parameters of IF have

been discussed, it is possible to estimate N, for the conditions of this experiment.

Table 2 contains the values of the the relevant parameters determined by the

methods described above and in Chapter 2. By incorporating these values into

Equation 1-7 above, the theoretical signal level of a single molecule in this project

is 81 counts per 2 ms measurement period of the photon counter.

Although the accuracy of these parameters is thought to be very good, and the

Equation 1-7 is theoretically sound, the determination of N, may not be truly valid

by this method for two reasons: 1) the potential for photodecomposition of the dye

before t, and 2) optical saturation of fluorescence. According to Equation 1-7,

simply increasing laser power, decreasing focus size, or increasing t, allows one to


obtain as large a signal as desired. However, physical limitations to negate this

possibility are described below.

Table 2. Parameters of the system to be used in the attempt of laser-induced
fluorescence detection of single molecules and the expected signal level
calculated from Equation 1-7.
Parameter Value
Laser Power, 'L 200 mW
Laser Frequency, vL 3.77 x 1014 Hz
Laser Focus Area, SL 3.5 x 10-5 cm2
Absorption Cross-Section, aA 3 x 10-16 cm2
Dye Quantum Yield, Y, 1
Collection Efficiency, QF/4T 0.119
Detector Efficiency, q 0.1
Optical Transmittance, T 0.5
Residence Time, t, 0.002 s
Expected Signal Level, N, 81 photoelectrons

Optical saturation. Optical saturation occurs when the fluorescence signal

becomes independent of laser power. This effect happens when the rate of

fluorescence is limited by the time it takes the molecule to cycle through the

excitation/de-excitation process before it becomes available to go through another

cycle. Optical saturation" is characterized by the relationship,

ELoAYF 2 (g )A,21 (1-15)
hvL g +g2

where EL' is the laser irradiance at optical saturation (W/cm2), g, and g2 are the

statistical weights dimensionlesss) of the ground state and excited singlet state, Si,

respectively, and A,2 is the Einstein coefficient of spontaneous emission of the

fluorescence (s"'). This rate of spontaneous emission is simply the inverse of the

fluorescence lifetime, rp (s). To determine EL', which is the laser irradiance when

the slope of a log-log plot of signal versus irradiance becomes 0.5, the relevant

equation is41

g2 4,rhc 2A,
EL = ( )(h ), (1-16)
gl+g, YX.s

where c is the velocity of light (3 x 1010 cm/s), AXF is the full width at half maximum

of the fluorescence excitation and emission bands (cm), and X is the excitation

wavelength of the laser (cm). The value of A2, for the dye to be used in this system

is 1.25 x 109 s"' (rp = 800 ps);42 the bandwidths of the excitation/emission spectra are

50 nm; the laser excitation wavelength to be used is 794.76 nm. The values for g,

and gz are assumed to be equal (for So and S, vibronic levels, this is usually a valid

assumption). When these values are put into Equation 1-16, the resulting saturating

irradiance is 5900 W/cm2, which corresponds to 207 mW laser power in the 3.5 x 105

cm2 focus size. By this account, the stated laser power above does not saturate the

transition and the calculated 81 photoelectrons per counting period is theoretically

obtainable with the stated parameters.

If laser irradiance, EL, is increased above EL', N, is no longer given by

Equation 1-7, but by

N, = ( )A21( F )Tt. (1-17)
g +g' 4,r

It is not desirable to require the use of this equation, because in an actual analysis,

the laser power should be kept just below saturating conditions. For EL > EL', the

signal changes by less than a factor of 2, but the noise continues to increase with EL.

Photodecomposition. The second pitfall of Equation 1-7 is degradation of the

the dye before it emits as many photons as the theory predicts in the 2 ms sampling

time. Other researchers performing SMD have encountered this problem4"4 and

developed a procedure to determine the optimum parameters to reduce dye

degradation. In this optimization technique, the end result is that t, should roughly

correspond to the time it takes the molecule to decompose under the laser irradiance

of the experiment. For the dyes in these experiments, the molecule typically

undergoes 106 fluorescence cycles before it degrades.43 If this value holds true for

the dye to be used in this project, the time it would take for the molecule to

photodecompose at the conditions listed in Table 2 is 146 ms, nearly 100 times

longer than the measurement time of this experiment. Based on this estimate, it is

hoped that photodecomposition will not become a factor in this attempt at SMD.

However, if optical saturation occurs to the point that the molecule undergoes

fluorescence at the theoretical limiting rate (Az2 = 1.25 x 109 for the fluorophore42

of Table 2), it would only take 0.8 ms to go through 1 million cycles.

Noise level required for SMD. Based on the theory of LIF presented here,

the average signal level should consist of 81 counts above background per counting


interval. Using the statistical theory applied to SMD presented earlier, the maximum

mean noise level, ^pb, permitted during the 2 ms interval to attain an LOD of 1

molecule with 99.86% confidence is 56 counts. For LOG = 1 with 99.86%

confidence to be achieved with 81 counts above background, IbI would have to be 35

counts or less. Remember in the calculation of these values it is assumed that the

data follow a Poisson probability distribution with the variance equal to the mean.

There is some question as to whether this is true when actual data are collected.3

A key to the success of this project is the reduction of noise to this required level.

The next section is a discussion of sources of noise in LIF and how they can be

eliminated or reduced.

Sources of Noise in LIF and Means of Noise Reduction

The three limiting extrinsic sources of noise for typical LIF experiments

consist of scattered light, Raman scatter, and background fluorescence. Other less

severe sources of noise exist, but by far, these three are the largest sources.

Laser Scatter

Laser scatter can be divided into two categories, specular scatter arising from

reflections from optics and other surfaces, and Rayleigh scattering from the solvent.

Laser specular scatter. In nearly every analysis utilizing LIF, scattered light

from the laser constitutes the most severe source of noise. This feature is not

surprising when one considers the amount of light that is delivered by the laser. For

example, the number of photons focused on the sample by the 200 mW laser at

794.76 nm to be used in this project is 8 x 1017 s' or 1.6 x 1015 photons during the 2

ms counting period. Of these, only 1 photon in 117 billion (a~SL) is absorbed by a

molecule in this time, which essentially still leaves some 1.6 x 10is nonabsorbed

photons. Realizing that 0,/4T is nearly 12%, qi is 10%, and T is 50% for the system,

the detector would produce = 9.6 x 1012 photoelectrons per 2 ms assuming that the

laser light is scattered isotropically from a point source. However, this assumption

is not true; the vast majority of the laser light continues unhindered through the

sample container and is not scattered within the 1.5 sr region collected by the

microscope objective (which is why fluorescence is collected 90 from the angle of

excitation). For the sake of argument, assume only 0.1% of the photons are

scattered. This still corresponds to nearly 10 billion photoelectrons. The 81

photoelectrons emitted by a single fluorophore in the same time period pales in

comparison. If the above assumptions are correct, an absorbance of > 10 is required

to reduce the scatter to a level compatible for SMD with this system.

Monochromators. As can be surmised based on this analysis, laser scatter is

an enormous problem experimentally. Even with the spectral selectivity of a

monochromator, stray light rejection is typically on the order of 10s which still leaves

some 50,000 photoelectrons produced under the conditions stated above (the 0.1%

isotropic point source of laser scatter at the focus of the collection optics is not valid

except in the case of capillaries as will be discussed later). Furthermore, with

monochromators, the spectral bandpass of the emission process is also greatly

reduced which lowers the signal as well as the noise. Therefore, the use of a

monochromator in the SMD approach is inadequate to the task.

Spectral filters. The use of spectral filters is a common approach to reduce

laser scatter in IF experiments. Typical rejection of interference filters and long

pass spectral filters is 10i, which is poorer than the rejection obtained with most

monochromators, but the filters generally have a much greater optical throughput of

the fluorescence signal. In most cases, two or more filters are used together or in

conjunction with a monochromator. Again, the problem of this approach is that as

the filters reject more laser scatter, the signal is also reduced. This type of filtering

is explored in more detail in Chapter 3.

Polarization. Lasers are typically highly polarized sources of emission.

Through the use of polarized filters, the spectroscopist can take advantage of this

trait to reduce laser scatter because the large majority of scattered light retains its

polarization. Conversely, fluorescence emission is nonpolarized. Optimally,

polarized filters are capable of 109 rejection of light polarized in the same direction

as the filter and are still able to pass light of the opposite polarization. This means

that approximately half of the nonpolarized fluorescence should pass through the


In practice, polarized light rejection does not work as well as expected from

theory. Problems arise with the purity of the laser emission polarity and maintaining

polarity when scattered from complex materials. Experimentation with stray light

rejection by polarization is discussed in Chapter 3.


Spatial filtering. As mentioned above, the focused laser light does not scatter

equally in all directions. The pattern of laser scatter produced at the sample greatly

depends on focusing and the shape of the sample container (this aspect becomes very

important later in this dissertation when considering the capillary container used in

this project). Spatial filtering exploits the nonisotropic feature of the scatter through

the placement of a small slit or pinhole between the collection optics and the

detector. With careful positioning and focusing, the aperture can collect a large

percentage of the fluorescence while blocking much of the laser scatter arising from

the edges of the container. The implementation of this simple concept by Dovichi

et a"45 (along with the use of 3 spectral filters and sheath flow cuvette sample

container) was a great break-through in initial studies on single molecule detection.

The problem with the spatial filter in SMD, however, is that it can limit the

probe volume of the analysis. By collecting emission from only a portion of the

focused region, the sampling efficiency is reduced which circumvents one of the

conditions for SMD. It is also very difficult to position the optics and spatial filter

for optimum effect.

Rayleigh scatter. Another drawback with spatial filters is that they do nothing

to limit the amount of Rayleigh scattering reaching the detector. Unlike specular

scatter, which arises at interfaces between media of different refractive indices,

Rayleigh scatter occurs in the medium itself due to light interaction at the molecular

level. Furthermore, the fraction of light that is scattered specularly (measured as a

percentage) is very high compared to Rayleigh scatter which typically has a cross-

section on the order of 1028 cm2. This corresponds to the production of 1

photoelectron at an average of every 2 ms for methanol in the 140 nL Vp of the

conditions presented in Table 2. Although this is very small, at the light levels

involved in this project, every additional noise photoelectron could become


The metal vapor filter. A possible way to eliminate the serious problem of

laser scatter is through the use of a metal vapor filter (MVF). This device is central

to the success of this project and its theory is presented in Chapter 2 and

experimental results in Chapter 3. The theory is too extensive to be presented now,

but according to theory, the MVF is capable of essentially totally absorbing the laser

scatter (or the entire laser emission for that matter) provided the laser emission

bandwidth is narrower than the absorption band of the metal vapor. Moreover, the

MVF is completely specific to the laser wavelength and does nothing the hinder the

transmittance of nearly the entire fluorescence emission band of the fluorophore.

Based on the calculated 10 billion laser scatter photoelectrons, it must suffice to say

that the required rejection of the MVF must be on the order of 1010 or higher to

achieve SMD, which is theoretically possible as shown in Chapter 2.

Even though the MVF is capable of eliminating the problems with laser

specular scatter and Rayleigh scatter under conditions of this experiment, other

extrinsic sources of noise exist that would thwart SMD. Both Raman scatter and

background fluorescence from the solvent and optics occur at removed wavelengths

from the laser, which allow their passage through the MVF to the detector.

Raman Scatter

Unlike Rayleigh scattering, Raman scatter is an inelastic process that occurs

with even lower probability, typically with a cross-section of 103 cm2 per molecule.39

This hardly appears significant until one realizes that in a large volume, the number

of molecules in a material is so great that the value for N, becomes significant. As

LIF researchers have learned, noise due to Raman scatter from the solvent becomes

the limiting source of noise when the laser scatter is reduced. The way around this

problem is the reduce Vp which limits the number of atoms and molecules in the

container that scatter light. This is one of the reasons why all LIF approaches to

single molecule detection utilize a small Vp. In this project, with the parameters

listed in Table 2, the Vp of 140 pL containing methanol would give rise to about 1

photoelectron per every 100 counting intervals. This is an acceptable level for SMD,

but a more difficult to quantify amount of Raman scatter arises from the quartz of

the capillary. This and other aspects of Raman scatter are discussed in Chapter 2.

Background Fluorescence

As discussed in the theory, fluorescence has a rather high absorption cross-

section (oA = 3 x 1016 cm2 for the fluorophore presented in Table 2) and can be

measured very sensitively. It is a selective technique as well, but it is often unable

to specifically detect one fluorophore at low concentration in the presence of

another. Furthermore, no solvent is absolutely pure, and even the presence of an

ultratrace concentration of fluorescent interferents can negate the possibility of SMD.


There are two main methods to avoid this problem. In the first case, the

purest available solvent should be used, and secondly, the detection should be

designed to be as specific to the analyte as possible. The former method is not trivial

in even the best available solvents,4 and attempts at SMD are consigned to basically

working with standard solutions of the purest solvent. However in the application

of these techniques to a real sample, interfering species become a severe problem

in the production of background fluorescence. In a real analysis, it is not realistic to

assume that the analyte will be the only detectable fluorescent species in the sample.

Excitation at long wavelength. The best way to avoid background fluorescence

is to specifically analyze the analyte. Very few species fluoresce at far-red/near-

infrared wavelengths, and the only known dyes to do so at the laser excitation

wavelength of this project are given in Chapter 2. Therefore, the method of

detection for this project is very selective to the molecule of interest.

An additional benefit of using laser light at 794.76 nm, as opposed to the

more commonly used 325 nm emission from a HeCd laser or the 514.5 nm line from

an Ar+ laser, is that Rayleigh and Raman scattering processes are reduced by a

factor of 1/X4 (and specular scatter is also reduced to a large extent).39 This is a

substantial reduction in noise when pursuing SMD.

History of Single Molecule Detection

Now that the pertinent concepts of SMD and LIF have been introduced, the

previous accomplishments of LIF analysis nearing SMD can be reviewed without

having to define terms or explain the rationale behind the design of the experiments.

Single atom detection. In the past, there have been several instances of SAD

most notably through the research of Letokhov"'47 and Hurst.48 Their separate work

concerns the use of resonance ionization spectrometry to detect atoms in an ion trap

or in an atomic beam. As discussed earlier, these experiments meet the

requirements of SAD as stated by Alkemade,"'2 but do not conform to the practical

definition of SMD of this dissertation.

LIF of solids. In the case of molecules, single molecule detection has been

accomplished under somewhat artificial circumstances. Hirschfeld49 implemented LIF

with a microscope to detect a single protein molecule (MW 20,000) tagged with

80-100 fluorescein molecules on a solid substrate. Kirsch et aL,5 in a similar type of

procedure, were able to detect 8000 rhodamine 6G molecules. More recently,

Moerner5'53 has detected single pentacene molecules in a solid matrix at low

temperature using laser-excited fluorescence.

Keller's approach to SMD. In the analysis of flowing solutions, Keller's group

performed several experiments leading to the claim of single molecule detection

(although none of the reports satisfies even Alkemade's definition of SMD).2' 45'"

The origin of Keller's project at Los Alamos National Laboratory, which is designed

for the application of SMD for DNA sequencing,' began with research by Dovichi

et aL45 who bested the previous lowest LOD by nearly two orders of magnitude in

obtaining an LOD of 35,000 rhodamine 6G molecules. The subsequent experiments

at Los Alamos concerned refinements of the basic set-up developed by Dovichi.45

As shown in Figure 1, this basic set-up utilizes a tightly focused argon ion

laser to excite a highly fluorescent dye flowing in a sheath flow cuvette. A

microscope objective collects the fluorescence, spatial and spectral filtering reduces

scattered light, and a cooled photomultiplier tube coupled with photon counting

electronics measures the signal.

Table 3 is a list of the parameters and detection limits of published results

reported by the Los Alamos group. From the table, it is apparent that the

researchers have been slowly lowering the detection limit with difficulty. In Ref. 4

and Ref. 6, in which single molecule detection was claimed, the sampling efficiency

(e,) was very poor, water was not used as the solvent, and the sheath flow cuvette was

abandoned. Also, the traditional method of determining the LOD was not used;

instead autocorrelation analysis of a single sample was performed. In Ref. 4, the

research team used a laser with 70 ps pulses and a microchannel plate detector with

sophisticated signal collection to help discriminate the fluorescence from the scatter.

The data presented in Ref. 4 do appear to be single molecule events, but the signal

to noise ratios are not reported so a statistical treatment to determine LOD cannot

be performed. The researchers base their claim of Ed = 70% (of those molecules

passing through the center of the Vp) on computer simulations that appear similar

to the actual data. Weighted quadratic sum plots of the data presented showed

1/2 Wave
Plate C



Figure 1. The instrumental approach to single molecule detection used by
Keller's group at Los Alamos National Laboratory. Redrawn from
references 2, 3, and 5.

Table 3.

Comparison of the reported parameters and results of the laser-
induced fluorescence experiments of Keller's group at Los Alamos
National Laboratory.

Parameter Ref. 2 Ref. 3 Ref. 4 Ref. 5 Ref.6
(1984) (1987) (1990) (1991) (1991)
Analyte R6G R6G R6G R6G R6G
Solvent H20 H20 H20/C2HsOH H20 C2FHOH
Cell SFC SFC flowcell SFC flowcell
Laser, Ar+ Ar+ Nd:YAG Ar+ Ar+
XL (nm) 514.5' 514.5 532b 514.5 514.5
SL (cm2) 3.8x10 1.1x106 4.4x10-7 2.1x10- 2.1xl05
EL (kW/cm2) 130 700 6.8 40 23.4
Stream Size 30 42 4000 44 250
F/4'r 0.06 0.045 --- --- -
E, 0.6 0.06 1.9x10-7 0.1 0.05
Vp (pL) 11 0.6 0.44 11 10.7
Flow Rate 25 0.012 5760 1.0 0.18
Flow Velocity 60 14.2 0.075 5.4 4.85
t, (ms) 0.037 0.085 10 1.8 2
rc (s) 1 1 0.004 1 0.0004
LOD (M) 1.3x10-13 2.2x10-13 -- 9x10-1 ---
LOD (#/rc) 33,000 1,200 1" 33 "1d
R6G = Rhodamine 6G; SFC = Sheath Flow Cuvette; e, = Sampling Efficiency;
7, = Time Constant of Measurement;
LOD (#/7,) = number of molecules passing through Vp during 7, at the LOD (M).
'Pulsed at 10 kHz, 50% duty cycle; bPulsed at 82 MHz, 0.57% duty cycle.
cTime discrimination method able to detect passage of single molecules with reported
70% detection efficiency (Ed).
dReported detection limit based on autocorrelation analysis; Ed not given.

photoelectron bursts arising from passage of single molecules which were not

presented in any of the other references. The same sort of basis was used in the

claim to single molecule detection made in Ref. 6, but in this case, passage of

individual molecules in the V, was not noticeable. The claim to single molecule

detection was based on the use of a 20 point sliding sum distribution which agreed

with theoretical results. No efficiency of detection can be calculated from such a

determination because the single events could not be counted.

In references 3 through 6, the spatial filter viewed only a small portion (5 Jtm)

of the laser focus volume to reduce the background sources of noise discussed

earlier. This technique limited the solid angle of collection of the microscope

objective which is why p0/4r is not reported in those cases (except Ref. 2). In Ref.

5, which reported an LOD of 9 x 10-' M based on conventional methods to

determine the LOD, the 4bI of the PMT was 212,764 +/- 461 counts/s. This high

noise level is a result of the large background sources of noise discussed earlier. At

this noise level, SMD is not possible with the approach presented in this thesis. I

a separate study utilizing B-phycoerythrin as the analyte, Nguyen et aL5 in Keller's

group claimed the first instance of single molecule detection in solution. It is

noteworthy that B-phycoerythrin is a very large (MW 250,000 g/mol) protein

possessing the equivalent fluorescence of 25 rhodamine 6G molecules."

Mathies' research. At the University of California at Berkeley, the research

group of Mathies contested this initial claim to SMD made by Keller's group; they

repeated the LIF study with Through a more rigorous statistical


approach, their research effort showed that the previous work did not obtain SMD,

and their results demonstrated the detection of 15% of the passing single molecules.55

Based on the definition of SMD presented in this thesis, B-phycoerythrin, with

a MW 250,000 g/mol, does not qualify as a "single molecule," and the sampling

efficiency of the experiment is much less than unity. Keller's group2' and other

researchers43 are aware of these shortcomings and have worked to lower the LOD

for smaller fluorophores. In their proposal,' Keller's group' mentioned that sampling

efficiency must be increased to perform the desired DNA sequencing application, but

the reduction of noise through the use of the spatial filter was integral to the

detection limits they have achieved.

Also, the sheath flow cuvette as the sample container is inadequate to achieve

SMD with e, = Ed = 1. Further addressed in Chapter 2, sampling efficiency and

detection efficiency are diametrically opposed relationships with the sheath flow

cuvette. To increase sampling efficiency, the laser beam focus must be increased, but

to increase detection efficiency, the focus must be kept small. Furthermore, with

small probe volumes, the sample flow rate must be lowered to maintain the residence

time, but with a sheath flow cuvette, the flow stream becomes broader with

decreasing flow rate thus requiring a larger beam focus. This has been a difficult

problem with the sheath flow cuvette, and researchers using this device have resorted

to finding an optimum trade-off." In this respect, it is unlikely that research with a

sheath flow cuvette will ever achieve true SMD. In fact, the time discrimination

approach of the Los Alamos group with the frequency doubled Nd:YAG laser

appears more promising for true SMD than the approach exhibited in Figure 1.

Indeed, two of the recent papers by Keller do not use the sheath flow cuvette.4'6

Winefordner's approach. To avoid the problem of this trade-off with a sheath

flow cuvette and its high cost, Winefordner's group has decided to probe the entire

sample stream in their attempts at SMD. Furthermore, the diode laser was chosen

as the excitation source for purposes of greater analyte selectivity, lower noise from

Raman scatter, lower cost, simplicity, and the many other advantages of diode lasers

to be discussed in Chapter 2. In two separate studies (with different lasers), LODs

of 40,000 and 3,000 molecules flowing in the probe volume of the near-infrared dye,

IR 140, have been measured in a liquid jet (flow stream emanating from a


Despite the many advantages of working with diode lasers and the low

detection limits attained with their use, LIF with diode lasers was found to lack the

sensitivity to achieve SMD in a liquid jet. Furthermore, the liquid jet only operates

at high flow rates which gives unsuitable residence times for SMD. This flow

condition is one of the reasons why the method described in this thesis employs a

capillary for sample containment.

Ramsey's approach. Ramsey's group at Oak Ridge National Laboratory has

designed the first instrumental set-up capable of truly realizing SMD in solution as

defined in this dissertation.4 The design of this approach appears in Figure 2. The

key to the experiment is the use of the electrodynamic trap to contain droplets of the

sample instead of a flowing stream. In this way, the entire sample is probed (one

I 0

E =
S f


a b..















5 o
s ,D



drop at a time) and the interaction time can be on the order of days if necessary.

The noise rejection is not as high as in other experiments and the dye tends to

photodecompose before SMD can be quantitatively attained. The LOD currently

stands at an average of 25 molecules of rhodamine 6G per droplet.

The approach to SMD of this dissertation. It is hoped that the SMD approach

described in Chapter 2 will be the first method to truly achieve SMD in a solution.

Like Ramsey's approach, it is designed to sample the entire solution, and has been

shown to theoretically perform the desired assignment. More of the theory, design

and development considerations, and some results will be presented in the following



The Metal Vapor Filter

Because of its role to remove the laser specular and Rayleigh scatter from the

collected light, the metal vapor filter (MVF) is the most important component of the

experimental set-up to attain single molecule detection (SMD). With few exceptions,

laser scatter is the limiting source of noise in ultratrace analyses using laser-induced

fluorescence (LIF). Therefore, satisfactory performance of the metal vapor filter is

crucial to the success of the experiment.

Theory of the Metal Vapor Filter

The MVF typically is a 2-3 inch long glass cell that contains a surplus amount

of a metal in its elemental form enclosed in a nitrogen environment. Figure 3 is a

simple drawing of the MVF. Upon gentle to moderate heating of the cell, a portion

of the solid or liquid metal enters the vapor state depending on temperature,

pressure, and thermodynamic properties of the element. The absorption properties

of the MVF depend on the number density, or concentration, of the metal in the

vapor state, the length of the cell, and the pressure in the cell.



Metal Element

Figure 3. The metal vapor filter.

Absorption. Increases in the metal vapor number density and absorption path

length lead to directly proportional increases in absorbance assuming the linewidth

of the source is more narrow than the absorption bandwidth. This relationship is

exhibited by Beer's Law,39

A = 0.434Ao^n, (2-1)

where A is the absorbance dimensionlesss), A^ is the absorption cross-section (cm2),

f is the absorption path length (cm), and n is the number density of the absorber

(cm"). In the case of the MVF, the equation to determine the absorbance of an

atomic transition can be calculated from,58

e If.S,vnil
A = 0.434e (2-2)

where e is the charge of an electron (1.6 x 1019 C2), f is the oscillator strength for

the electronic transition from level i to level j (unitless empirical value for each

transition), S,v is the Voigt shape function of the absorption bandwidth (Hz-'), nA is

the number density of the atom in level i (m-3), e, is the permittivity of free space

(8.854 x 10-2 Ns'/C2), m, is the mass of an electron (9.11 x 10"3 kg), and c is the

velocity of light (2.998 x 108 m/s).

Choice of Metal for the Filter

For the MVF to be useful in practice, the electronic absorption transition

must start from a ground state due to the small fraction of the element existing in

an excited state as described by the Boltzmann distribution. Therefore, level i in

Equation 2-2 must be the ground state, and nA must be the number density in the

ground state. Another important consideration for the practical use of a MVF is the

thermodynamics of the chosen element. In order to attain a large enough number

density, n, to satisfactorily absorb the large light intensities associated with a laser,

the metal must have a low melting point and high vapor pressure. Table 4 is a list

of elements that could be of practical use in a MVF along with their melting points,

ground state transition wavelengths, and approximate o^ of the overall transition. Of

these elements, rubidium is an excellent choice for use in a metal vapor filter due to

its low melting point and strong absorption lines at 780.023 nm and 794.760 nm. For

these reasons and more (based on characteristics of lasers and fluorescence dyes

available, which will be discussed later in this chapter), rubidium has been chosen as

the element to include in the MVF for use in the SMD system.

Thermodynamic characteristics of several elements for possible use in
the metal vapor filter.








M.P. ("C)'







X (nm)a

gA (cm)b
9 x 10-13
2.5 x 10-"
8 x 10-1
2.1 x 10-"
1.3 x 10"1
9.5 x 10"1
4 x 10-"
8 x 1012
8.7 x 10-12
4 x 10-12
2 x 10-"
1 x 1011
3 x 10-13
6 x 1012
3.5 x 10-13
1.3 x 10-1
1 x 10-11
7 x 10-4
4 x 10-12
1.3 x 10-"
6 x 10-"
1.1 x 10-1
1.3 x 10-12
5 x 10-12
5 x 10-12
2.3 x 10-14
1.2 x 10-"

Table 4.

"Values from Lange's Handbook of Chemistry, 13th Ed., JA. Dean, Ed., McGraw-Hill,
New York, 1985.
bValues calculated by Ramee Indralingam from Equation 2-2.

Calculation of Spectral Linewidths and Absorbances for Rb

The two most important features for use of the MVF are the absorption

coefficients (or absorption cross sections) and spectral profiles for the atomic

transitions at 780.023 and 794.760 nm. These properties give an idea of the required

specifications of the laser to be used in conjunction with the MVF in SMD. In order

to quantify these two parameters of interest, some of the theory of line broadening

will be given.

Atomic spectral profiles. The emission or absorption spectral profile of a

single, stationary atom contained in a vacuum absent of electric and magnetic fields

would have a spectral linewidth determined by the lifetime of the electronic

transition as stated by the Heisenberg uncertainty principle." Typically, atomic

transition lifetimes are on the order of 10' s which corresponds to a natural, or

fundamental, linewidth of 10 MHz or 0.021 pm at 800 nm. However, in a real system

there are several factors, such as motion of the atom, atomic collisions, and presence

of electric fields, which act to broaden the spectral profile for a given transition.

Furthermore, fine and hyperfine structure exist due to quantum splitting of electronic

transitions and existence of isotopes for a particular element. These components

each have an individual, quantifiable effect, and after they have been factored

together, the overall peak shape, termed the Voigt spectral profile, can be calculated.

Rubidium has two main isotopes, Rb8 and Rb87, which exist naturally in the

ratio 2.59:1 for RbS:Rb7. The fine structure for transitions of these isotopes and

their fi values are given in Table 5 at 780.023 and 794.760 nm.59

5. Physical parameters of the rubidium hyperfine structure required for
the calculation of the Voigt spectral profile of the metal vapor filter.

For the Absorption Band at 780.023 nm (vo = 12.820 cm-'):

for Rb7 for Rb"8

V-p0 fij V-VQ f
(cm-l) (cm-')

0.1323 0.0417 0.0553 0.0833
0.1347 0.104 0.0563 0.108
0.1401 0.104 0.0585 0.0864
-0.0933 0.0208 -0.0451 0.0309
-0.879 0.104 -0.0430 0.108
-0.790 0.292 -0.0389 0.250

For the Absorption Band at 794.760 nm (vp = 12.582 cm^):

for Rb87 for Rb85

P-VP fij V-Po f4
(cm') (cm-')

0.1255 0.0208 0.0521 0.0309
0.1527 0.104 0.0642 0.108
-0.1025 0.104 -0.0493 0.108
-0.0753 0.104 -0.0372 0.0864

Each of these lines is broadened by the same extent as described below and

their normalized absorbances are added together at individual wavelengths to form

the Voigt profile.


Lorentzian broadening. There are two types of broadening taking place in the

MVF which factor into the Voigt profile, namely, Doppler broadening and collisional

(or pressure) broadening. The different types of collisional broadening can be

grouped together to produce the total Lorentzian profile. In the case of the MVF,

it is realistic to assume that all collisions of the atoms in the vapor state occur with

diatomic nitrogen (N2) and are adiabatic in nature (the atom remains in the same

electronic state during the collision). Thus, the Lorentzian linewidth, AvL (Hz), can

be calculated from the equation to determine the profile due to adiabatic collisions,

which is"

2a,An 2RT (2-3)
2L-j (2-3)

where R is the gas constant (8.314 J/K mol), T is the temperature (K), u is reduced

mass of Rb and N2 (kg), a, is the collisional cross-section of Rb (m2), and n, is the

number density of nitrogen in the metal vapor cell (mn3; in the case of nitrogen, n,

= 9.74 x 10u P/T, with pressure, P, in Torr and temperature, T, in K). For Rb in

a nitrogen environment, a, has been measured to be 2.49 x 10"9 m2 at 780.023 nm

and 2.30 x 10-19 m2 at 794.76 nm.58

Doppler broadening. Doppler broadening, APD (Hz), a result of the atoms

moving at different velocities upon absorption of light, can be determined from39

2p, 2(ln2)RT] (2-4)
c M

where R is the gas constant (8.314 J/K), M is the formal weight of Rb (kg/mol), and

vm is the center frequency of the overall transition (Hz).

The a-parameter. To determine the proportion of each broadening effect in

the overall linewidth, the "a-parameter" is used, where, a = 0.83(AvL/APD). In the

case of the Rb metal vapor cell at pressures > 75 Torr, the effect of collisional

broadening is greater than that of Doppler broadening, which is expressed by a > 2.

The Voigt integral value, b(a,0), at the line centers of both Rb resonance lines can

be calculated (within 10%) to be 6(a,0) m 0.56/a (for a > 2).39

Voigt linewidth. The multiplication of 5(a,0) by the Doppler shape function,

S,D (s), results in the Voigt shape function, S,v (s). For Doppler broadening, SD at

frequency v (Hz) can be determined from the relationship,39

2 1/2
S,D = 2 l e -4(n2)(,-,^/(A, (2-5)

and the Voigt linewidth, Avv (Hz), is found from

APV [.2_+'2 ] 1/2 (2-6)

All linewidths denoted by the subscript v correspond to the full width of the

peak at half of the maximum intensity. To convert any of these linewidths from

frequency, Av (Hz), to wavelength, AX (nm), the value is multiplied by wavelength

squared and divided by the velocity of light, c (3 x 107 nm/s). For example, AXv at

800 nm is found from the relationship, Avv(800 nm)2/c.39

Rb number density. At this point, all of the parameters, except nA and e, have

been defined that are necessary to determine the absorbance, as described by

Equation 2-2, and the Voigt spectral shape function, as described by Equation 2-5.

For the metal vapor cell I is fixed and n, is dependent on temperature and pressure.

For the evaporation of liquid Rb, nA follows the expression,"

log(n.) = -A/T (B+ 1)log(T) + C + DT +18.985, (2-7)

where A = 4529.6, B = 2.991, C = 15.8825, and D = 0.00059 with T in K. There

have been many different measurements of the values of these constants by physicists

with variability in nA as great as 25%, but the above values were chosen from

reference 60 because of their close agreement with other referenced values61 and

high precision.

Voigt spectral profile. The overall absorption line shape of Rb vapor was

calculated with the help of a computer spreadsheet and a computer program (written

in basic language by Michael Wensing). The program summed the absorbances of

the individual components of the profile versus wavelength based on values

calculated from the above equations in spreadsheet format. Plots of the Voigt

profiles at different temperatures and pressures were then generated (with I of 4.4

cm). These plots are exhibited in Figures 4 and 5. As the figures show, increasing

cell temperature has a large effect on the absorbance, which essentially mirrors the

increasing Rb vapor number density, and increasing pressure increases linewidth at


the expense of absorbance. The presence of more than one peak at low pressure

exhibits the fine structure of the overall transition.

For this work, two different Rb metal vapor cells were available for use in the

project. Both cells contained 500 mg of 99.99% pure rubidium metal in nitrogen

(this is more than enough to produce a large vapor number density without

expending all of the condensed metal). One cell (to be referred as cell #1),

manufactured by Rudy Strohschein in the University of Florida glass shop, has an

absorption path length of 4.4 cm and a pressure of 200 Torr at room temperature.

The other cell (cell #2, manufactured by Opthos, Rockville, MD) has a 4.7 cm path

length and a room temperature pressure of 500 Torr. These parameters were

entered into the computer programs to calculate the Voigt profiles for these cells at

100C for both the 780.023 nm and 794.760 nm lines. These profiles appear in

Figures 6 and 7. The Voigt linewidths are 21 pm (at 780 nm) and 20 pm (at 794.76

nm) for cell #1, and 36 pm (at 780 nm) and 36 pm (at 794.76 nm) for cell #2. Cell

#1 is capable of a larger absorbance at constant temperature than cell #2, which

makes cell #1 preferential for use in the experiment provided that the laser emission

linewidth is less than 21 pm.

Tye Barber41 experimentally verified the Voigt profiles generated from

calculations using a single-mode diode laser with cell #1. He found the absorption

bandwidth to be 21 pm at 780 nm which is in excellent agreement with the calculated

bandwidth. The shape and intensity of the absorption peak also closely agreed with

the computer generated plot at the same temperature.41


o -


"- a

o 0.
o 4
/ I '-"

0 0

o -

o T

o0 0 000 O' -
eo-- oq \-
CUJ 0 cO (0 t1 C J




0 0
0 0

o \
0-- 8'



0 ou -q-o, \ r


0 o

O o noL m o m o LO O1
It I- Co C, C'j CJ '-

eoueqjosqv 4


4 )


a) 0


-- ^----------oa a- O


-a l

\\ i'

CD LO 9t CO C\J C0
eoueq N






o 4
z 0

Lo 0 o

CM s

o .
0M -0


o 0


o 0
I I I \ I ,- ,

eoueqJosqv a
\\ o O
\ \ "" l
\ \ ^il

The Laser

Instrumentation for fluorometric analysis. To perform molecular fluorescence

analysis, five basic components are necessary: 1) a source of light, 2) a fluorescent

sample, 3) a container for the sample, 4) a spectral filter, and 5) a detector. In the

SMD project, the MVF, which is able to remove the source light while still allowing

the passage of the fluorescence to the detector, serves as the fourth component listed.

The excitation source is the subject of this section, and each of the remaining

components will be discussed individually in the following sections.

Light sources. For selective analysis, the excitation source should have a

narrow emission band, and for improved sensitivity, the source should have high

emission intensity. Arc lamp sources are capable of very large radiances, but in

order to obtain narrow bandwidth, the light must be passed through a

monochromator which filters the excitation source into less intense bands. With the

invention of the laser, a source became available that was capable of high power in

a narrow spectral bandwidth. However, a problem with the laser as an excitation

source is its lack of tunability. Only those analytes with an excitation band at the

laser emission wavelength can be determined by LIF. Of course, there are many

different lasers with many different laser lines, but the cost and practicality of using

so many different sources makes LIF an analytical technique for unique


Criteria of the Laser for Single Molecule Detection

The choice of laser to be used in the detection of single molecules must be

made based on several criteria. These criteria include:

(1) the laser must be capable of sufficient irradiance (W/cm2) at the excitation

wavelength of a given fluorophore to attain SMD;

(2) the laser emission wavelength must fall at or be able to be tuned to the

absorption wavelength of the metal vapor cell and remain at this wavelength

over time;

(3) the emission profile of the laser must be Gaussian-shaped with a linewidth

less than the absorption bandwidth of the metal vapor;

(4) the source must be continuous or have a pulse rate of at least m 10 kHz in

order to efficiently probe a flowing sample.

Other desirable traits of the source include low cost, simple maintenance, and easy


Choosing the laser. Most lasers meet criterion (1) if the beam can be focused

to a very small area. Because no non-tunable lasers happen to emit light at an

absorption band of an element for practical use in a MVF, criterion (2) establishes

that a tunable laser must be used. Only dye lasers, diode lasers, Raman-shifted

lasers, and the Ti:sapphire laser are tunable. Requirement (3) negates Raman-

shifted lasers and all pulsed dye lasers except the copper vapor-dye laser system.

Criterion (4) precludes the use of a pulsed laser unless it is capable of a repetition

rate greater than = 10 kHz. The upper limit of the repetition rate of the Cu vapor-

dye laser system is 10 kHz, thus single-mode diode lasers and Ti:sapphire laser have

the best characteristics for SMD.

Diode lasers. Diode lasers are solid state electronic emitters of

electromagnetic radiation much like light-emitting diodes except the diode laser

possesses a cavity to induce lasing.2 The lasing material is usually gallium doped

arsenide which has an energy gap that corresponds to wavelengths in the red to near-

infrared wavelengths. Diode lasers have several advantages over conventional lasers

that make them generally accepted among spectroscopists as the "laser of the future."

The favorable characteristics62 of diode lasers include: (1) inexpensive, (2) easy to

operate, (3) maintenance free, (4) long-lived, (5) robust, (6) small, (7) tunable, (8)

powerful, (9) efficient, (10) versatile, (11) high power stability, and (12) narrow

spectral linewidth.

Despite these advantages, diode lasers will remain the "laser of the future"

unless the following problems can be corrected. Firstly, diode lasers are tunable, but

not continuously tunable allowing coverage of all wavelengths in the tuning range.

Diode lasers commonly exhibit mode hopping which is the tendency of the laser to

jump instantaneously from one wavelength to another and exhibit lack of tunability

in the region between these modes. Secondly, since diode lasers have a small laser

cavity, the laser beam has a large beam divergence which makes focusing difficult.

Lastly, diode lasers are available over a limited wavelength range of 650-1300 nm

with each laser tunable over a 20-30 nm range. If the laser had a low divergence, it

would not be difficult to frequency double the emission to wavelengths half of the


fundamental output utilizing nonlinear optical properties of certain crystals. This

would enable diode lasers to be capable of emission from 325-1300 nm, but the poor

doubling efficiency currently obtained results in laser powers too low for general


However, it is fair to say that the current intensive research in the area of

diode lasers in the brief time since their invention has resulted in more powerful

lasers, wider range of lasing wavelengths, more efficient single-mode operation, and

lower cost which will soon result in a laser useful for SMD. Already, diode lasers

can be used in conjunction with a MVF.41 Special electronic control of the diode

laser current with active feedback virtually eliminates mode hopping, but the low

power and focusing difficulties of these lasers does not yet meet all of the criteria

listed earlier. Undoubtedly, diode laser excitation would be the preferential

approach for SMD due to the many advantages of diode lasers, but until the above

problems are solved, another laser must be used.

The Titanium:Sapphire Laser

Of all possible sources currently available, only the titanium:sapphire

(Ti:A1203) laser meets all of the outlined criteria. Table 6 lists the specifications of

the manufacturer of the Ti:sapphire laser purchased for use in this project. As the

table shows, the laser is capable of two lasing arrangements, the standing-wave and

ring cavity configurations. The specifications for both configurations are the same

except that the spectral linewidth is narrower in the ring configuration.

Figure 8 is a drawing of the design of the Ti:sapphire laser. The diagram

shows the laser in the ring cavity arrangement with the beam passing through the

optical diode. In the standing-wave configuration, the two flat mirrors are tilted to

return the beam to the curved mirrors on either side of the crystal instead of through

the optical diode.

Table 6. Specifications of the titanium:sapphire laser used in the single
molecule detection project.

Parameter Specified Value
Maximum Power 750 mW
(at 800 nm) (10% of pump power up to 7.5 W)
Tuning Range 700-1000 nm
Spectral Linewidth <2 GHz standing wave
(at 5 W Pump Power) <40 MHz ring configuration
Spectral Profile Gaussian
Spatial Profile TEMoo
Polarization Horizontal
Beam Size 1 mm at exit

The Ti:sapphire laser is a passive device with no moving or electronic parts.

An important aspect of the Ti:sapphire laser is that it requires the services of a pump

laser to initiate the lasing action. The source of the lasing is a solid state AO13

crystal doped with titanium. The crystal has broad excitation and emission bands

with maxima = 530 nm and 800 nm, respectively." Due to peak absorption in the

green part of the spectrum, argon ion, Nd-YAG, and copper vapor lasers are used

to pump the Ti:sapphire laser. The Ar+ laser is used most often to take advantage













of its continuous laser action, but pulsed operation can be applicable in many

situations. The output power of the Ti:sapphire is proportional to the power of the

pump laser and, according to the manufacturer specifications, is capable of 10%

efficiency of conversion, but in the SMD experiments, nearly 20% efficiency was

attained with careful alignment.

Comparison of the Ti:sapphire laser with dye lasers. The Ti:sapphire laser

was invented in the 1980s and is poised to replace the dye laser for many

applications. In the early 1970s, dye lasers were hailed as the solution to the lack of

laser tunability. There are several problems with dye lasers, however, concerning

both scientific and practical aspects in their use. Each dye has a limited tuning range

based on concentration, solvent, pH, and pump laser. Power is strongly dependent

on wavelength and is greatly changed when a dye laser is scanned; efficiency of

conversion is often less than 1% for dye lasers. For these reasons, dye lasers are

seldom used to scan over a wavelength range, and the more common application is

to tune a dye laser to a fixed wavelength. In practice, dye lasers are noted for the

problems: dye decomposition, solvent flammability, chemical toxicity, waste disposal,

chemical spills, and extended down time (often due to solvent pumps). For these

reasons, spectroscopists have been anxious for a laser to replace the dye laser.

The Ti:sapphire laser has many advantages over the dye laser, but is not the

panacea for the laser spectroscopist. Since the Ti:sapphire is a passive, solid state

laser, it has none of the practical problems associated with using laser dyes. No

moving parts translates to no down time for mechanical reasons, and no dyes means


no messes, chemical hazards, or waste. Also, the Ti:sapphire laser is capable of

continuous tuning with moderate change in power over a continuous range of 120 nm

(fundamental) or 60 nm (frequency doubled) before requiring a change of optics.

Few individual dyes can match this feat; however, the overall wavelength coverage

of 350-500 nm and 700-1000 nm is still inferior to a network of dyes that allows for

coverage of the spectrum from the ultraviolet to the near-infrared. Therefore, the

Ti:sapphire laser is not able to replace the dye laser in all applications.

The Ti:Sapphire Laser/Rb Metal Vapor Filter Combination

The stated linewidth specifications of < 2 GHz in the Ti: sapphire standing-

wave configuration and < 40 MHz in the ring cavity configuration corresponds to <

4.1 pm and < 0.081 pm, respectively, at the 780 nm Rb transition. For the 794.76

nm transition, these linewidths are < 4.2 pm and < 0.084 pm, respectively. Due to

the Gaussian nature of the spectral profile for the laser as opposed to the Lorentzian

wings of the Voigt profile, the laser beam for either laser configuration should be

virtually totally absorbed by the Rb metal vapor cell at temperatures > 100*C. This

aspect is shown in Figure 9 in which the emission profiles of the Ti:sapphire in both

configurations are superimposed onto the Voigt profile of cell #1 at 794.76 nm. As

shown, the specified laser linewidths are much narrower than the profile of the

absorption line. This condition is especially notable in the ring cavity profile where

the laser line is a thin line on the same scale as the other profiles.

A4!SuGIuI Jesei








C) e-



-0 r-


T- 0


Absorbance is the log of the inverse of transmittance, T, (A = log(l/T))

which means that the absorbance of 10 for the Rb metal vapor filter at 100*C

corresponds to a transmittance of 10-10, or 99.99999999% of the light is absorbed.

Therefore, with a metal vapor absorbance of 10, the calculated maximum percentage

of light scattered by the capillary is 5.9% (if it is an isotropic point source at the

focus of the microscope objective) that would still allow for an LOD = 1 molecule

based on the parameters listed in Table 2 of Chapter 1. In other words, the laser

scatter collected in the 1.5 sr solid angle of collection (0,/47r) must be limited to less

than 0.7% (5.9% x 11.9%) to achieve a background count rate of 56 counts per 2 ms

increment (maximum theoretical AbM possible in order to attain SMD for Xd = 81

counts. These calculations are based on the discussion given in Chapter 1.

It should be pointed out that the Ti:sapphire laser/Rb MVF combination has

been used previously to perform Raman spectroscopy.6', As in LIF, the limiting

source of noise in Raman spectroscopy is typically laser specular scatter. With a

similar laser/MVF set up to be used here, Raman spectra were obtained that showed

no evidence of the laser despite viewing at the laser wavelength."'" This is a major

accomplishment in Raman spectroscopic analysis because it is now possible to

measure Raman bands very close to the excitation wavelength. Furthermore, the

conditions for optimum performance of Raman spectroscopy are very similar to those

of the SMD; since the Ti:sapphire laser and Rb MVF combination was successful in

that application, there was little reason to believe it would not be successful in

detecting single molecules by LIF.

Ti:sapphire laser configuration. Because the linewidth in either the standing-

wave configuration or ring cavity configuration is much narrower than the Voigt

profile of the Rb vapor cell, it does not matter which laser configuration is used for

single molecule detection. Both configurations were tested for use in the project with

no significant difference in results.

The Sample

After the excitation source, the sample constitutes the second of the five basic

components of a fluorometer. In the case presented so far, the analyte must absorb

strongly at 780 or 795 nm (Rb ground state electronic transitions) and fluoresce with

high quantum efficiency at removed wavelengths. Furthermore, as discussed in

Chapter 1, the analyte should not be a polymer or other chemical species with a

molecular weight greater than = 1000 g/mol. Another consideration is the solubilty

of the molecular species because this analysis is to be attempted in solution phase.

Choice of Analyte

Polymethine dyes. Very few chemicals fluoresce beyond 700 nm. In fact, the

only known molecules to fluoresce strongly in the far red part of the spectrum are

the cyanine dyes known as polymethines.67 Table 7 is a list of polymethine dyes that

includes their common names, chemical formulas, excitation and emission maxima,

and fluorescence quantum yields of those available in the literature. As expected for

fluorescent molecules, these dyes have highly conjugated, semi-symmetrical structures

Fluorescent dyes for possible use in the single molecule detection project.

Dye Molecular x X, C eA" Yp
Formula (nm) (nm) (MN'cm'1)

Rhodamine 800 C26H26N30 Cl 682" 700' 89,500 0.39'
Methylene Blue CiXH{N3S Cl 668' 683" 66,60 --
Nile Blue C2oH2N30 Cl 640" 672' 77,500"
Oxazine 750 C24HN3O Cl 673 691 82,500 ---
IR 125 C43H47N2S, Na 780 806 150,000 0.13c
IR 132 Cs2H48N304S2 C104 810 846 210,000
IR 140 C38H34C12N3S2 C104 800 833 180,000 1.0d
IR 144 CoH8sN408S2 NEt3 698b 708b 127,000 ---
DTTC CH25HN2S2 I 746b 777b --- 0.38c
DTDC C2H23N2S2 I 647 668b --- 0.73
DOTC CsHN202 678b 703b --- 0.63'
HITC C28H33N2 I 736b 764b 240,000 0.28c
HDITC C36H37N, C104 771b 805b
DDTC C32HN2 765" 855 -- 0.16'
DQDC CH27N2 765' 835C -- .001'
DQTC C29H2N2 825' 865' -- .035c

'Values in water from T. Imasaka, A. Tsukamoto and N. Ishibashi, Anal Chem., 61. 2285
bValues in methanol from D. Andrews-Wilberforce and G. Patonay, AppL Spectrosc., 43
1450 (1989).
"Values in dimethylsulfoxide from R.C. Benson and H.A. Kues, J. Chem. Eng. Data, 22, 379
dValue in ethanol from D.J.S. Birch, G. Hungerfold, B. Nadolsi, R.E. Imhof and A.D.
Dutch, J. Phys. E: Sci Instrum., 21, 857 (1988).
All other values determined experimentally in methanol.

Table 7:


containing several aromatic rings with the length of the structures correlating to

excitation and emission wavelengths. In fact, Benson and Kues6 have worked out

an empirical relationship between the structure of polymethine dyes and their

fluorescent characteristics.

Of the dyes listed in Table 7, only three, IR 125, IR 140, and IR 132, are

potentially useful for single molecule detection with excitation at 780 or 794.76 nm.

Figure 10 gives the structures of these polymethine dyes. It is rather fortuitous that

X," of IR 125 falls at the 780 nm Rb line, and that IR 140 has k" at 800 which

is very near the 794.76 nm transition. Meanwhile, IR 132 is the least favorable of

these dyes in this respect with X,", of 810 nnm.

Another important parameter for SMD is the absorption cross-section, Ao

(cm2), of the analyte which is proportional to the molar absorptivity, cA, of the dye

in bulk solution as shown in Chapter 1. In a simple experiment using Hewlett-

Packard 8450A and Varian 634 (for wavelengths greater than 800 nm)

spectrophotometers, the absorption spectra of known concentrations of the dyes in

methanol were measured. The wavelengths of maximum absorption corresponded

to the fluorescence excitation maxima for the dyes, and the measured EAm for IR

125, IR 140, and IR 132 were 1.5 x 10', 2.1 x 105, and 1.5 x 10' M-Ncm', respectively.

These figures are in agreement with values reported in the literature.69

The emission spectra of the dyes in methanol solution are shown in Figure 11

with excitation at 780 and 795 nm. For this experiment, the Ti:sapphire laser was

used as the excitation source of the flowing dye solutions contained in a 1 cm path

N j CH=CH)3 -CH Jg
(CH2)4 IR 125
80- 803 Na

CI H5 IR 140 i
C2 C2H5

8 N 8
Q/[OH=-CH j CH-H hh I
N 0 0
o4 (CH)3OC IR 1 32 o
4 (CH2 OCCH (CH2)30CCH3

Figure 10. Chemical structures of the polymethine dyes to be tested for use in
the single molecule detection project.
length cuvette. A Spex 1680 double monochromator with 1 mm slits was used to
collect the spectra and a cooled Hamamatsu R636 photomultiplier tube served as the
detector. These spectra have been normalized to a laser power of 145 mW and dye
concentration of 4.6 x 107 M. The X maxima for IR 125, IR 140, and IR 132 with



-0 0
o 1.
co c


6( 0

I I 0 co


(tyu) luejjno I|/4d .
0 00

this system were 806, 833, and 843 nm, respectively. Each of these dyes was tested

for application to the single molecule detection project as presented in Chapter 3.

More detailed characteristics of each dye will be discussed separately below.

IR 125. The formal chemical name for IR 125 is anhydro-l,1-dimethyl-2-[7-


sulfobutyl)-1H-benz(e)indolinium hydroxide sodium salt; its Chemical Abstracts

Services (CAS) number is [3599-32-4]. The compound is more commonly known by

many chemists as indocyanine green, or ICG, which is associated with its use as an

indicator. Laser spectroscopists, on the other hand, are more familiar with the name

IR 125 which is associated with its use as a laser dye. IR 125 is the only water

soluble dye of the three species, and is also readily soluble in most organic solvents.

In water, IR 125 has a pKl of 3.27.70 Like the other polymethine dyes, IR 125 is a

zwitterionic salt due to the presence of an aromatic heterocyclic ring containing


Applications. IR 125 is the most widely used of the mentioned polymethine

dyes, mainly due to its solubility in water. By far, the most common uses of IR 125

are as a laser dye" and as a clinical indicator dye for testing of in vivo blood flows

and hepatic functions in animals and humans.7 It is useful in clinical applications

due to its large molar absorptivity at long wavelengths where blood does not absorb

strongly. It has also been used in angiography73 and many other studies in blood.2

A very interesting aspect of IR 125 for future applications is that it has been bound

to surfactants74 and proteins75 for analytical purposes. Patonay at Georgia State

University is actively pursuing the use of IR 125 and other polymethine dyes for

tagging purposes.67 If his research is successful, this SMD technique could become

very important in many tagging applications.

Indirect fluorometric detection. Another interesting use of IR 125 concerns

its chromatographic properties. It has been analyzed by high-performance liquid

chromatography (HPLC) for the clinical applications previously mentioned,7678 and

based on these studies, it was thought to be an excellent choice as a visualization

agent for indirect fluorometric detection in HPLC.79 Indirect detection works on the

principle that by monitoring the concentration of a continuously present indicator

species, termed the visualization agent, the presence of other species can be

determined by fluctuations in the concentration of the visualization agent.8 In this

manner, it is a universal method of detection with detection limits based on three

factors: 1) the size of the effect of the analyte on the signal of the visualization

agent (known as the transfer ratio); 2) the ability to measure these signal fluctuations

(or dynamic reserve); and 3) the concentration of the visualization agent.7 In HPLC,

ion exchange chromatography and capillary electrophoresis, the magnitude of the

transfer ratio can be very large based on separation properties, and IR 125 is a good

choice for this purpose because it can be detected very selectively and sensitively due

to its fluorescence at long wavelengths. A diode laser is suitable as the excitation

source due to its very high power stability which greatly reduces noise on a large,

constant signal. Indirect fluorometric detection with diode laser excitation of the IR

125 visualization agent was used in the detection of alcohols using reversed-phase

HPLC,79 and current work is underway to use indirect fluorometric detection in

capillary electrophoresis with the system. If successful, it will be possible to obtain

a LOD of 10 pM for an analyte with a transfer ratio near unity.: Indirect

fluorometric detection is of interest to separation scientists because it has the

potential of being a sensitive and universal detector which is a rare combination.

IR 140. IR 140 is formally known as 5,5'-dichloro-11-diphenylamino-10,12-

ethylenethiatricarbocyanine perchlorate; its CAS number is [53655-17-7]. The major

use of IR 140 is as a laser dye," but it has also been used as the analyte in several

LIF experiments with diode laser excitation.56,s7'",81 The goal of these projects was

very much similar as this one, which was to attain the smallest possible limit of

detection of the dye in a flowing stream. Based on these earlier experiments with

this dye and the very high fluorescence quantum efficiency (YF = 1) and a high

spontaneous emission (A21 = 1.26 x 109 s~1 based on 791 ps fluorescence lifetime,

rT),42 IR 140 is emphasized in the experimental studies.

IR 132. Like IR 140, the only known use of IR 132 is as a laser dye.71 It's

formal name is 3,3'-di(3-acetoxypropyl)-11-diphenylamino-10,12-ethylene-5,6,5',6'-

dibenzothiatricarbocyanine perchlorate (CAS # [62669-62-9]).

With the advent of diode lasers, these dyes have become very important as

probe species. Several analytical chemists have used these dyes in an attempt to

apply diode laser source to analytical techniques." Since diode lasers are only useful

at wavelengths longer than 650 nm, and polymethine dyes are one of the few

molecules fluorescent at these wavelengths, they have been thrust into several


applications.81 In the SMD project, the optimum dye is to be chosen based on the

conditions of the excitation wavelength and filtering to be discussed later in this

chapter and the next.

Stability of the Dyes. Due to the highly conjugated structures of these dyes,

it was anticipated that these dyes decompose readily. There have been numerous

studies involving the decomposition of IR 125 in blood plasma, water, and electrolytic

solutions, and indeed, IR 125 degrades in a matter of hours in these solutions.2

However, the decomposition rate is vastly reduced in organic solvents. A study was

performed to determine whether decomposition of the dyes would be a problem for

the single molecule detection project. Figure 12 shows the fluorescence of the dyes

in methanol over a period of 25 days. These measurements were made with a Spex

Fluorolog 2 spectrofluorometer with 1 mm slits and a cooled Hamamatsu R928

photomultiplier tube detector. The Xx was set to 764 nm because the 500 W Xe arc

lamp used as the excitation source has a greater emission intensity (and produces a

larger signal) at this wavelength than at the wavelengths of ECA of the dyes. Stock

dye solutions of m 1 x 10' M were kept at room temperature in the dark. Each time

a spectrum was taken, a 0.1 mL aliquot of solution was pipetted into a 1 cm path

length cuvette and diluted with 3 mL of solvent. As the figure shows, there was quite

some day-to-day variation in the procedure, but the overall fluorescence did not

decrease significantly. Therefore, the dyes do not readily decompose in organic

solvent (assuming that the degradation products do not fluoresce at the same

wavelengths as the dyes).



I + 2


,o *
0 0

S+ oJo

C-o .a
7*- 2


-T z:, *
0 0 0

8- 3 0-r------ 0 Sn 4

Effect of degassing the solutions. In a similar study utilizing the same

instrument, the effect of bubbling different gases on the photodecomposition of the

dye solutions in methanol was examined. In this study, = 0.1 jiM dye solutions were

prepared, split into 3 volumes, and continuously sparged with nitrogen, helium, and

air. Their fluorescence emission spectra were measured after m 1 minute. These

solutions were then placed in a chamber radiated by light of wavelengths greater than

700 nm. For this purpose, a 500 W Eimac arc lamp was placed in the chamber, and

its white light emission was filtered by a series of long pass spectral filters. The

emission spectra of 3 mL aliquots of these solutions were measured periodically over

a period of 2 hours. Figure 13 shows the effect of degassing on the

photodecomposition rate of IR 140 in methanol. This figure clearly shows that air,

or more specifically oxygen, causes increased dye degradation as opposed to an

oxygen free solution. Figure 13 further shows that nitrogen and helium greatly

reduced the photodecomposition with respect to air, therefore, all solutions used in

the single molecule detection project should be sparged with one of these two gases.

Based on cost and ease of access, N2 was chosen for this purpose.

The Choice of Solvent

Since the goal of this project was to achieve single molecule detection with no

particular sample type in mind other than a solution, the choice of solvent was based

mainly on signal to noise of the dye fluorescence in that solvent. Of course, water

would be the ideal solvent due to its great applicability to many types of analyses, but














".f .



since IR 140 and IR 132 are water insoluble, and IR 125 decomposes rapidly in

water,82 H20 is not one of the choices as solvent for this experiment. Instead, several

organic solvents covering a wide range of properties, were tested.

Table 8 is a list of six solvents, and their pertinent characteristics. Of the

enormous number of organic solvents, these six were singled out due to their general

availability and history of use in fluorescence analyses. Polarity tends to be the most

important factor in the fluorescence intensity and maxima emission and excitation

wavelengths. For most fluorescent solutes, nonpolar solvents, such as hexane, tend

to shift the spectra to longer wavelengths and reduce intensity. Nonpolar solvents

are useful when a red-shift is desirable, or when the analyte is soluble only in

nonpolar solutions.

Table 8.


Physical properties of

)xide CH3SOCH3

the solvents tested for use in

Density n'
0.7908 1.3588
0.840 1.3420
1.100 1.4783
0.7894 1.3614
0.6594 1.3749
0.7913 1.3284
1.0000 1.3330

this project.


"n is refractive index at 589 nm; be is dielectric constant
All values for 200C from Lange's Handbook of Chemistry, 13th Ed., J.A. Dean, Ed.,
McGraw-Hill, New York, 1985.


The emission spectra of the IR dyes were taken with the laser tuned to 780

and 795 nm using the same instrumental system and normalized to the same

conditions as described in Figure 11 for methanol. The results of these spectra with

,, = 795 nm are compiled in Figure 14 which shows the normalized peak emission

intensities of IR 125, IR 140, and IR 132. Acetone and acetonitrile consistently yield

the largest fluorescence signal for the dyes whereas the nonpolar solvents, hexane

and dimethylsulfoxide produce less intense fluorescence. This is a typical behavior

for fluorescent compounds.

Raman spectra of the solvents. Based on Figure 14, acetone or acetonitrile

would be the best choices for use in the SMD project. However, signal is not the

only criterion on which the selection of solvent is based. Signal to noise ratio is the

most important parameter in any analysis to minimize limits of detection. As

discussed in Chapter 1, Raman scatter is the second most severe source of noise after

laser specular scatter, and if the MVF removes the laser scatter as it should, Raman

scatter from the solvent will then be the limiting source of noise. Since Raman

scatter is shifted in wavelength from that of the laser, it is transmitted through the

MVF to the detector. Therefore, it must be reduced at the source, by the reduction

of V, and choice of solvent, or filtered by a spectral filter other than the MVF.

Figure 15 contains the Raman spectra of the 4 solvents that gave the highest

fluorescence signals in Figure 14. These spectra were taken under the same

conditions with the Spex Raman microprobe instrument utilizing an Ar+ laser source

and cooled RCA C31034 detector. In Figure 15, the Raman shift in cm"1 from the

*I V I I

1 f I




( *II





leu6!s Need p9z!lweJON

_ __







514.5 nm Ar' laser line has been converted to wavelength as if the excitation

occurred at 794.76 nm. By referring to the fluorescence of the dyes in Figure 11, one

can see that the Raman spectra of the solvents overlap with the fluorescence signal

of the dyes. Only methanol is devoid of Raman peaks at wavelengths less than 860

nm. Assuming Raman scatter is the limiting source of noise, methanol is then the

best choice of solvent, despite the slightly lower intensity of the IR dyes in methanol

than in acetonitrile and acetone. These solvents give several very intense Raman

peaks around 820 nm.

To ensure that the extension of the Ar' excited Raman spectra to the

Ti:sapphire was valid, the spectra of methanol and acetonitrile were measured in a

1 cm square flow cell with Ti:sapphire laser excitation and the Spex 1680 double

monochromator/Hamamatsu R636 PMT combination. Figure 16 shows the resulting

spectra from this analysis with excitation at 795 nm. The wavelengths of the peaks

correlates with the wavelengths shown in Figure 15. It is unknown why the

background for the methanol spectrum is higher than it is for the acetonitrile in this

case. In the case of the argon ion laser excitation, the high background for methanol

is believed to be fluorescence of an impurity.

Dimerism. Another reason to choose methanol over acetonitrile as the

solvent is that the polymethine dyes dimerize in acetonitrile and not in methanol.

Dimerization is the tendency of solute molecules to associate with each other in pairs

instead of exist individually in solution. In the literature, other researchers have

c O
Oo N

6 co .

a) C

-0 4)

<1 o a P


U I,


oo 0

(vd) .UJJn4 1 1
0cc 0

0 0 0 02
+ + + +
w w w w

noted this problem with polymethine dyes in certain solvents.67'983 Dimerism is

exposed in calibration curves of signal versus concentration. In acetonitrile, the

calibration curve was nonlinear, and when converted to log-log scale, the slope of the

line was 2. Figure 17 is a calibration curve of IR 125 in acetonitrile taken with

Ti:sapphire laser excitation at 795 nm; based on the log-log slope of 1.99,

dimerization is clearly shown in this figure. This situation is much like the analysis

of sulfur with a flame photometric detector in which signal is proportional to

(concentration)2 due to the production of S2 in the flame. This is an unacceptable

situation when calculating limits of detection based on the sensitivity of an analysis.

Methanol, on the other hand, produces linear calibration curves and log-log plots

with a slope near unity. This indicates that dimerism of the dyes does not occur in

methanol solutions, and that LOD calculations based on sensitivity can be made of

this dye/solvent system.

Sample Containment

In an ideal system, the method of sample containment should not perturb the

sample in any way or introduce noise into the measurement. Despite this seemingly

innocuous task, sample contact with the sample cell may cause problems such as

introduction of reaction surfaces and matrix interference. In spectroscopy, light

transitions from one medium to another decreases the amount of light reaching the

sample due to scatter, fluorescence, and absorption by the containing medium.



0- |
\- 8 -

U) 00
\ en q,

S~- |S

\0 0
( U I1- 'I d

I I e5

\- 0

\4 a-


LO '+ 0+J 0
+ + + + + + 0 '

Lx ^

*s 1^
o i 3

Furthermore, laser specular scatter from the container constitutes the major source

of noise in most IF analyses.

Liquid jets. Some of the problems with sample containers can be eliminated

by using a liquid jet, which is merely the term used for a flowing stream emanating

from a nozzle. The optimal probe region occurs at a point just after the orifice

where the flow stream narrows before it spreads again. Previous studies in this

group'657 used a liquid jet in the analysis of IR 140 in methanol with diode laser

excitation. In these studies, noise due to laser scatter was greatly reduced by exciting

the stream outside of the capillary, and the sensitivity of the measurement was

increased because more of the laser light was reaching the sample. However, the

major problem with a liquid jet is that it only operates at high flow rates. At lower

flow rates, the sample drips from the orifice one drop at a time which also causes

measurement difficulties. Another problem is that residence time of the analyte in

the probe volume is < 1 ms at flow velocities of liquid jets. For SMD, the desired

residence time (t,) is in the range of 1-10 ms, depending on laser power and dye


Levitation. Ideally, the analytical chemist would like to ensure that the

sample is isolated from external factors and positioned in the probe region for a

length of time sufficient for complete analysis. The most obvious way to accomplish

this goal is through sample levitation. Means of sample levitation involve the use of

physical forces to counteract gravitational attraction, or as in the case of performing

analyses in outer space, the reduction of gravitational effects. Of course, since space-

based research in analytical chemistry is overly expensive and generally impractical,

the former means of levitation are far more common than the latter. Practical types

of sample levitation possible make use of aerodynamic, acoustic," photophoretic,"

electrodynamic,6 and magnetic forces.7 As discussed in Chapter 1, Ramsey's

approach to SMD makes use of the advantages of electrodynamic levitation.4

Although sample introduction into the trap is complicated and time consuming, it

does remedy some problems related to a flowing sample contained in quartz. The

way in which Ramsey contains the sample is reviewed in the following paragraph in

order to give solid grounds for comparison with the other methods.

Electrodvnamic levitation. In electrodynamic levitation, a charged species

(atom, molecule, particle, or droplet) is introduced into a chamber containing a ring

electrode, to which is applied an ac (radiofrequency) potential, and top and bottom

electrodes to which are applied dc potentials." A field applied to the ring electrode

controls the lateral position of the charged species, and the dc potentials applied to

the top and bottom electrodes control the vertical motion. Figure 2 in Chapter 1

shows an electrodynamic trap in the drawing of Ramsey's set-up. With samples

having high mass/charge ratios (> = 10 kg/C), such as the charged, micron-sized

droplets introduced into the chamber by Ramsey, the droplets can be stably levitated

at the center of the trap. Sample introduction is accomplished with a piezoelectric

droplet generator at the top electrode of the chamber; the electrode voltages are

then altered until the droplet is trapped.43 The size of each droplet must be

determined from the complicated analysis of scattering patterns of a HeNe laser

beam focused onto the droplet.43 Furthermore, the analyte must be contained in a

solution partially consisting of glycerin to reduce solvent evaporation under the laser.

After a few minutes required for introducing, trapping, and determining droplet

volume, the fluorescence analysis begins. Based on these aspects, one can see that

practical analysis of large volume samples would be tedious by this method and rapid

measurements in a flowing stream would be impossible.

Sheath flow cuvette. The approach to single molecule detection of Keller's

group makes use of a sheath flow cuvette for sample containment.2" The sheath flow

cuvette is a type of flow cell in which an external solvent stream, or sheath, is used

to compress an internal sample stream.8 Figure 1 in Chapter 1 contains a partial

drawing of a sheath flow cuvette used in Keller's set-up. The narrowing quartz walls

of the flow cell cause the sheath flow, originating from outside the internal capillary,

to compress the sample stream as it effuses from the internal capillary. The solvent

in the sheath is usually the same solvent as the sample in order to avoid changes in

refractive index, but use of different solvents has certain advantages in some


In the set-up at Los Alamos, the laser probes a region in the cuvette where

the quartz tube is square (flat surfaces do not scatter as much light as round

surfaces). In a situation much like the liquid jet, but not as severe, the flow rates

required to induce the desired narrowing of the sample stream, greatly decrease t,

When the flow rate is lowered in an attempt to increase t, diffusion of the analyte

into the sheath occurs and the irradiance of the laser required to cover the entire

sample stream is not sufficient for SMD. To deal with this problem, Keller's group

decided to avoid it altogether by lowering flow rate to a point where t, is adequate

for the detection of the single molecules as they flow through Vp,2'3'5 but the laser

probes only 6% of the sample flow region (and the spatial filter only views a

portion of that region). This does not meet the 100% sampling efficiency

requirement of SMD expressed in Chapter 1.

The capillary. Unlike the other sample containment methods discussed, the

capillary is simply a glass or quartz tube requiring no instrumental adjustments.

Capillaries of various sizes are commonly used in several types of analytical methods.

In separation techniques using capillaries (gas chromatography, supercritical fluid

chromatography, microcolumn HPLC, capillary electrophoresis), the inner diameter

(i.d.) of the capillary is an important characteristic because it effects column volume,

flow, pressure, heat dissipation, and separation factors.9" In gas chromatography,

capillary i.d. is on the order of hundreds of micrometers, whereas in capillary

electrophoresis, the capillaries range from 10-100 /im i.d. It is for this latter

technique that this experiment is designed, so the sample capillary is on the order of

those used for capillary electrophoresis. Furthermore, narrower capillaries have

smaller probe volumes which is desirable in the reduction of Raman scatter from the

solvent. The key in choice of capillary i.d. is the laser focus size and t, at obtainable

flow rates which will discussed in later paragraphs. As it turns out, a 50 Jm i.d.

capillary is nearly ideal for the system.

Capillaries commercially available for capillary electrophoresis are made of

fused silica, or quartz, with an outer diameter (o.d.) of 150 or 360 Gim. For this

application, the thinner walled tubing was chosen because: 1) thinner walls absorb

less heat and better dissipate heat than thicker walls; 2) thinner walls have a smaller

interaction volume with the laser beam (less absorption, less Raman scatter); 3)

focusing with narrower capillaries is easier; 4) the microscope can be positioned

closer to the source of the fluorescence emission, if necessary, with thinner walls; and

5) the greater lensing curvature of the narrower capillaries focus the laser to a

greater extent at the sample than thicker capillaries. Some of these points will be

discussed in more detail in the section below.

Capillaries used for electrophoresis are commercially available and are coated

with polyimide which endows the brittle quartz with flexibility. At the probe region,

the polyimide coating must be removed, which is usually accomplished through

stripping or burning it off. Burning is advantageous because it does not create

grooves or scratches on the capillary walls as stripping may do. However, burning

does leave a sooty layer on the capillary, but this artifact is easily removed by wiping

with a wet tissue. Once the polyimide coating is removed, the capillary must be

handled delicately because the quartz tube snaps easily upon slight bending.

Optical Considerations Regarding the Capillary

Focusing in optical systems is paramount in attaining the optimum conditions

in a spectroscopic analysis. The rounded surfaces of a narrow capillary create several

considerations involving focusing of the laser onto and collection of emission from

the sample stream not normally encountered. Previously, researchers performing LIF

for detection in capillary electrophoresis have resorted to special means to reduce

the problems associated with round surfaces.12 In many cases, the sample cell is

made to be rectangular,88 or a separate cell is used altogether such as a sheath flow

cuvette.9 An interesting approach is to manufacture an immersible cell of a desired

shape containing a fluid of the same refractive index of the capillary.9' However,

in this SMD approach, confidence has been placed on the ability of the MVF to

remove the laser scatter, so the capillary is used without alterations. The subject of

ways to reduce laser scatter by the capillary will be discussed in more detail in

Chapter 3.

Laser specular scatter. Specular scatter from a rounded surface occurs in all

directions arising from both the outer and inner walls of the capillary. The modeling

of this system is very difficult due to the shapes of the focus beam, capillary,

collection optics, and the difficulty of predicting where rays will end up from a

capillary. As mentioned in Chapter 2, up to 0.7% laser scatter of the total laser

intensity can be tolerated in the direction of the collection optics according to

calculations presented in Chapter 1. Through elaborate experimentation and optical

ray tracing, Bruno et aL9 have shown that the least amount of scatter from a round

capillary occurs in the direction perpendicular to the excitation beam. This is

because the increased angle of incidence of the incoming light at the edges of the

capillary causes closely spaced rays to disperse more widely at 900 to the laser beam.

Fluorescence is to be collected from this direction (90*) in this approach.

Rayleigh scatter. Laser specular scatter produced at interfaces of different

refractive indices is by far the most intense form of scatter produced from the

capillary, but Rayleigh and Raman scatter also arise from within the quartz to a

lesser extent. Rayleigh scatter is very weak and occurs at the same wavelength as the

laser emission. These forms of scatter have slightly broader spectral profiles than the

source, but due to the small Rayleigh cross-section, small capillary volume

illuminated, and broad absorption profile of the metal vapor filter (with respect to

the laser), this form of scatter is expected to be negligible as shown from theory in

Chapter 1.

Raman scatter from the capillary. Raman scatter from the capillary may

prove to be a more formidable problem than Rayleigh scatter. Quartz has a Raman

emission band" at 464 cm'n which corresponds to 825 nm with excitation at 795 nm

and 809 nm for 780 nm excitation. The Raman spectrum overlaps the fluorescence

emission spectra for the polymethine dyes, and there is little to be done to filter this

radiation unless a portion of the fluorescence is to be filtered as well. The possibility

of using a capillary made of a different type of glass does exist, but commercially

available capillaries of this sort are made of fused silica, and there is no guarantee

that other glasses will not have the same problem. The cross section for Raman

scatter is typically on the order or 10-30 cm2 per molecule,3 and the volume of fused

silica illuminated is about 100 nL. Without knowing the MW of quartz, the number

of expected photons arising from the quartz cannot be calculated. It is estimated

based on the parameters presented in Table 2 in Chapter 1 that the maximum signal

from the quartz of the capillary is on the order of 10 photoelectrons per 2 ms

counting interval.

Focusing the Laser

Laser focus onto the capillary. Figure 18 is a diagram, generated by a

commercially available optical ray tracing program (Beam 4, Stellar Software,

Berkeley, CA), that represents the laser focusing onto the 50 im i.d., 150 Jim o.d.

quartz capillary to be used in this system. The incoming light is collimated and was

assumed to be of 786 nm wavelength. The quartz has a refractive index, n, of

1.45356 at this wavelength4 and the inner tube was said to contain methanol (n =

1.327 at 589 nm). The focusing of the laser beam by the capillary wall is an

Figure 18.

Focusing aspects of the 0.05 mm i.d., 0.15 mm o.d. quartz capillary
containing methanol with a 0.075 mm beam of collimated light at
786 nm entering from the left. Diagram generated by an optical
ray tracing program.