Light Source Selection and Optical Design of a UV Absorption
Based Detector for Liquid Chromatography
by
Daniel P. Gillund
Bachelor of Science in Mechanical Engineering and in Physics
Massachusetts Institute of Technology, 2014
Submitted to the Department of Mechanical Engineering in
Partial Fulfillment of the Requirements for the Degree of
ARCHRIE
MASTER OF ENGINEERING IN MANUFACTURING
MASSACHUSETS INSTITUTE
OF TECHNOLOGY
at the
OCT 0 1 2015
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
LIBRARIES
September 2015
Copyright 0 2015 Daniel P. Gillund. All rights reserved.
The author hereby grants to M.I.T. permission to reproduce and to distribute publicly paper
and electronic copies of this thesis document in whole or in part in any medium now known
or hereafter created
Signature of Author:
Signature redacted
Daniel Paul Gillund
Department of Mechanical Engineering
August 10, 2015
Certified by:
Signature redacted
Jung-Hoon Chun
Professor of Mechanical Engineering
Thesis Advisor
Accepted by:
Signature redacted
David Hardt
Professor of Mechanical Engineering
Chairman, Committee for Graduate Students
[This page intentionally left blank.]
2
Light Source Selection and Optical Design of a UV Absorption Based Detector for
Liquid Chromatography
by
Daniel P. Gillund
Submitted to the Department of Mechanical Engineering on August 10, 2015 in Partial
Fulfillment of the Requirements for the Degree of Master of Engineering in Manufacturing
Abstract
Traditional sources of UVC light for absorption spectroscopy are bulky, inefficient, and
output tens to hundreds of watts of heat. In this paper we present the relative merits and
disadvantages of using AIN based UVC emitting LEDs for absorption detection in liquid
chromatography systems compared to traditional sources. We present the optical design for
a detector based on a modular LED architecture and employing a digital micromirror device
to modulate the light used in detection. The expected capabilities of a detector employing
the design are derived and then compared to existing UV absorption detectors for liquid
chromatography. The strategy presented in this paper holds significant advantages over
existing detectors, including increased resolution and dynamic capabilities.
Thesis Advisor: Jung-Hoon Chun
Title: Professor of Mechanical Engineering
3
Acknowledgements
Certain people deserve heartfelt gratitude for their involvement in this project:
My teammates and friends Aditya and Saksham without whom I and this thesis would be
just one leg of a three legged stool.
Our Advisor, Prof. Jung-Hoon Chun, who pushed us to think big.
Greg Puszko and Gabe Kelly for continually going out of their way for us.
Jim McPherson for his continual support of our project, even through some twists and turns.
Mark Basile for helping me through a crash course in optics theory and prototyping.
My wife Nicki, for her patience and support.
The many others at Waters Corp. who helped our team, for their patience and willing
assistance throughout
4
[This page intentionally left blank.]
5
Table of Contents
Abstract ................................................................................................................................................. 3
Ack-nowledgements ............................................................................................................................... 4
1. Introduction: ................................................................................................................................... 9
Detector Overview : ....................................................................................................................... 9
2. 1 Strategy .................................................................................................................................... 9
2.2 Critical Specifications ............................................................................................................. 9
2.2.1 Resolution ............................................................................................................................. 9
2.2. 2 Signal to Noise Ratio .......................................................................................................... 10
2.1.3 Optical Bandwidth .............................................................................................................. 10
2.1.4 Wavelength Range .............................................................................................................. 10
2 1.5 Dynam ic Capabilities ......................................................................................................... 10
Relevant Optical Theory ............................................................................................................. I I
.).1 Snell's Law ............................................................................................................................ 11
3. 1.1 Index of Refraction ............................................................................................................. 11
3. 1.2 Total Internal Reflection ..................................................................................................... 12
3.2 N umerical Aperture ................................................................................................................ 13
3.3 Etendue .................................................................................................................................. 14
3.4 Optical Power ........................................................................................................................ 15
3.5 Fraunhofer Single Slit Diffraction ......................................................................................... 15
3.6 Grating Theory ...................................................................................................................... 16
3.7 Digital M icro-mirror Devices ................................................................................................ 19
).8 Shot Noise ............................................................................................................................. 19
4. Light Source Selection ..................................................................................................................2 0
4.1 Overview of Available Light Sources ................................................................................... 20
4.2 Spectral Range ....................................................................................................................... 20
4.3 Stability ................................................................................................................................. 21
4.4 Optical Power and Spectral Irradiance .................................................................................. 22
4.5 Power Requirement ............................................................................................................... 24
4.5 Effi ciency ............................................................................................................................... 2 5
4.6 Lifetim e .................................................................................................................................. 2 6
4.7 Startup Time .......................................................................................................................... 28
4.8 Cost ........................................................................................................................................ 29
5. UVC LEDs .................................................................................................................................. 30
5.1 LED Operational Theory ....................................................................................................... 30
5.2 AIN Based UVC LEDs..........................................................................................................30
30
5.2.1 High Bandgap Sem iconductors .......................................................................................
5.2.2 Dislocation Densities..........................................................................................................30
5.2.3 Single Crystal AIN .............................................................................................................
30
5.3 Coupling LEDs to Optical Fibers .......................................................................................
31
6. Wavelength Selection..................................................................................................................31
31
6.1 Strategies Employed by Existing LC Detectors ...............................................................
6.1.1 Tunable UV Type Detectors............................................................................................31
33
6.1.2 Photodiode Array Type Detectors ...................................................................................
6.2 Alternative Strategy...............................................................................................................34
7. Unfolded Optical Path.................................................................................................................34
7.1 Initial Path .............................................................................................................................
34
7.2 Stray Light .............................................................................................................................
35
7.3 Path with Aperture Stops..................................................................................................
35
8. Optical Layout.............................................................................................................................36
9. Expected Performance.................................................................................................................38
9.1 Com ponent Specification ..................................................................................................
38
9.2 Bandwidth .............................................................................................................................
39
9.3 Optical Throughput ...............................................................................................................
39
9.4 Signal to Noise Ratio.............................................................................................................41
41
9.5 Sam pling Rate .......................................................................................................................
9.6 Resolution..............................................................................................................................43
9.6.1 Expected Resolution...........................................................................................................43
9.6.2 Resolution Sensitivity.........................................................................................................43
10. Comparison to Existing UV Detectors for LC .....................................................................
44
11. M echanical Design Considerations .......................................................................................
45
12. Conclusion.................................................................................................................................47
47
13. Future Work ..............................................................................................................................
14. W orks Cited...............................................................................................................................48
7
Table of Figures and Tables
Figure 1. Illustration of Snell's Law ................................................................................................
Figure 2. Total Internal Reflection ..................................................................................................
Figure 3. Optical Coupling for Elements with Mismatched Numerical Apertures ..........................
Figure 4. Fraunhofer diffraction patterns for a single slit (a), double slit (b), and ten slits (c) ...........
Figure 5. Linear wavelength dispersion of a spherical grating.........................................................
Figure 6. Spectral range of UVC light sources...............................................................................
Figure 7. Stability of UVC light sources .........................................................................................
Figure 8. Average irradiance of UVC light sources ........................................................................
Figure 9. Spectrum of UVC light sources .......................................................................................
Figure 10. Power requirement of UVC light sources .....................................................................
Figure 11, Efficiency of UVC light sources........................................................................................26
Figure 12. Lifetime of UVC light sources.......................................................................................
Figure 13. Startup time of UVC light sources.................................................................................
Figure 14. Estimated cost of UVC light sources ............................................................................
Figure 15. Tunable UV type detector scheme .................................................................................
Figure 16. Photodiode array type detector scheme..........................................................................
Figure 17. Unfolded optical pathway for alterative scheme............................................................
Figure 18. Unfolded optical pathway with aperture stops...................................................................36
Figure 19. Detailed optical layout for UVC LED detector..................................................................37
Figure 20. CAD model of optics bench............................................................................................46
Figure 21. Full scale looks like prototype .......................................................................................
Table 1. Component Specifications................................................................................................
Table 2. Reflection, transmission, and coupling efficiencies of optics ...........................................
Table 3. Sample and reference photodiode responses for various scenerios..................................
Table 4. SNR and noise levels for various scenarios .....................................................................
Table 5. Max operating frequencies for dynamic elements.............................................................
Table 6. Comparison of UVC LED detector capabilities with tunable UV and photodiode array
detectors...............................................................................................................................................4
12
13
14
17
18
21
22
23
24
25
27
28
29
32
33
35
46
38
40
41
41
42
5
8
1. Introduction:
In "Development of a Product Architecture for a LED Based Detector for Liquid
Chromatography" by S. Saxena[1], a product architecture was laid out for a UV absorbance detector
for liquid chromatography. In this paper, the viability of LEDs as a light source for UV absorbance
spectroscopy is examined, a detailed optical design for such a detector is presented and then
compared to existing UV absorbance detectors.
2. Detector Overview:
2.1 Strategy
The issue this paper seeks to address is a complete lack of availability in the liquid
chromatography detector market of a system which is inherently dynamic in nature. A detector with
such characteristics would be flexible enough to take on a wide range of applications and would have
tremendous potential for both backwards and forwards compatibility.
2.2 Critical Specifications
2.2.1 Resolution
One of the most critical specifications for a liquid chromatography UV detector is the
resolution of the instrument. Resolution is the measure, in nanometers, of how far apart in
wavelength two light signals must be to guarantee that the detector can accurately and repeatable
distinguished one from the other. It directly impacts the capabilities of the instrument to uniquely
identify substances.
Many variables can negatively impact the resolution, however, the two factors which set the
maximum theoretical resolution of an instrument are its slit widths and the linear dispersion of its
diffraction grating.[2]
9
2.2.2 Signal to Noise Ratio
Often liquid chromatography is used as a method to determine not what a sample is
composed of, but rather how much of a given substance is present in the sample. For these
applications the signal to noise ratio, or SNR, becomes the most important attribute of the detector.
Nearly every component of the detector and of its environment has the potential to add some
amount of noise to the signal. The addition of a reference can reduce or eliminate many forms of
noise. As a result the noise arising from the quantized nature of light, called the shot noise, often
dominates the SNR in UV detectors.[3]
2.1.3 Optical Bandwidth
The optical bandwidth of a detector is defined as breadth of the minimum spectra a machine
is capable of detecting. It can be equal to, or more often, greater than the resolution of the detector.
For example, if an instrument has a resolution of I
nm and an optical bandwidth of 5 nm it
could reliably detect absorption between 250 nm and 255 nm or between 251 nm and 256 nm.
2.1.4 Wavelength Range
Wavelength range is the portion of the electromagnetic spectrum in which a detector can
operate. It is set either by the spectral distribution of the source, or the design of the optical system.
Many methods for absorption spectroscopy are designed around a very specific wavelength
or set of wavelengths.[ I] Therefore, the range of wavelengths over which a detector can operate
determines the set of applications which it can be used for.
2.1.5 Dynamic Capabilities
The dynamic capabilities of a detector are the set of features which can be used while
samples are running. This includes reference-based noise canceling, baseline adjustments, on-the-fly
self-calibration, timesharing among multiple wavelengths, and actively modulating the source or
entrance slit.
Dynamic features add flexibility and robustness to a detector. They can provide a means for
backwards and forwards compatibility, and serve as tools for the basis of future innovation.
10
3. Relevant Optical Theory
3.1 Snell's Law
3.1.1 Index of Refraction
For any transparent medium, the index of refraction for that medium, ?r, is defined as the
speed of light in a vacuum, c, divided by the speed of light within that medium, c,,,, as shown.[4]
C
77m = CM
tI)
Snell's Law describes the way light is transmitted at the interface of two mediums. As seen
in Figure 1, light will bend when passing between mediums which have different indices of
refraction. This phenomena is described by Snell's law, given in Equation 2, which says that the
ratio of the sines of the incident and transmitted angles is equal to the ratio of the indices of
refraction in the two mediums.[4]
7rt
77i
sin ;
sint
(2)
11
ot
T1 t
Frtgztre I. il/usiraton of SneL' Law
3.1.2 Total Internal Reflection
By solving Snell's law for Ot as shown in Equation 3. we can see that there is a critical angle
O at which Ot becomes imaginary, given in
Equation 4.[4]
Ot = sin-1(
sin i
(3)
7i
12
All light incident at or below this angle will reflect completely without any transmission
through the second medium. This effect allows optical fibers to transmit light over large distances or
along a specific pathway with very little losses, as shown in Figure 2.
Oc
Or
F-igure 2. 1t tal Internal Reflection
3.2 Numerical Aperture
The numerical aperture, NA, of an optical element is a dimensionless number which
is defined in Equation 5, as the refractive index of the medium, n, times the sine of the half
angle of the element's acceptance cone, a.[5] The numerical aperture squared may be
thought of as the light gathering power of the element.
13
(5)
NA = n sin a
In Figure 3, we can see what happens when we couple two optical elements together which
have mismatched numerical apertures. Working through the optical system in one direction, the
second element of the pair will be overfilled and the extra light will be unable to propagate through
the rest of the system properly. If, instead, we start from the other direction the second element will
be under-filled. The coupling efficiency, or fill factor, for the second element will be the ratio of
squares of their numerical apertures as shown in Equation 6.
(6)
NA2iement
77coupling
2
mNnA
source
10
=
Feiement
4
_14
Figure 3. Optical Couplingfar Elements wh Mismatcheth
Numerical Apertures
3.3 Etendue
The Etendue of a system is defined in Equation 7, as the cross sectional area of the
source, As, times the solid angle through which the light propagates, Q. It can be thought
of as the ability of an optical system to accept light.
14
K)
For small solid angles, Q will be linearly proportional to the numerical aperture
squared and the expression for etendue can be written as shown in Equation 8.[2]
E
(s)
= AsirNA 2
3.4 Optical Power
Once we know the system Etendue we can calculate the flux, (P, or optical power,
passing through it by simply taking the product of the Etendue and the radiance of the
source, as shown in Equation 9.[2]
P
QI
= ER
3.5 Fraunhofer Single Slit Diffraction
When coherent light passes through a narrow slit in a plane normal to the light's Poynting
vector, it produces interference patterns which in the far-field are well described by the Fraunhofer
diffraction equation, given in Equation 10.[5]
=)I
0 (sinx
X
2
(101
Where I is the intensity, I is the intensity at the center of the diffraction pattern, and with
minima occurring wherever x is an integer multiple of n.[5]
15
3.6 Grating Theory
If more than one slit are arranged in a pattern, the effects of the interference are
compounded. As we can see in Figure 4. as we add more slits the peaks become taller and narrower.
Single Slit Diffraction Pattern
(a)
Position
16
Double Slit Diffraction Pattern
(b)
>1)
Position
Ten Slit Diffraction Pattern
i>
(c)
rA
42.)
.
AA
-
m .
L-Al
i"
Position
Figure4. F-lrunhoter difricuion pasterns f]r a single slit (a), double slit (bl and ten slls
(c)
17
Because the location of the peaks is dependent upon the wavelength of diffracted light, this
effect can be used to sort light based on wavelength. If we use a repeating pattern of grooves in a
reflective surface rather than slits we will be able to retain more of the optical power. Such an optical
element is called a reflective grating. The diffraction off of such a grating is described by the grating
equation, as shown in Equation 11.[2]
sin c + sinf =knA
(11)
Where a is the incident angle,fi is the diffracted angle, k is the order of diffraction, n is the
groove density, and
2 is the wavelength. From this equation and the geometry as shown in Figure 5.
we can derive an expression for the linear dispersion of a spherical grating in nanometers per
millimeter, which is given in Equation 12.[2]
101 cos fl cos 2 y
dA
dx
knLH
La
HH
B
I
Figure 5. Linear wavelength dispersion ofa sphericalgrating
18
3.7 Digital Micro-mirror Devices
Digital micromirror devices consist of arrays of thousands to millions of microns scale
mirrors fixed to mems actuators which allow each individual mirror to be positioned at either t 00
from its neutral "flat" state. This allows arbitrary patterns of light to be imaged at high speeds.
3.8 Shot Noise
For a photodiode the shot noise power, i2, which arises from the quantized nature of light, is
given by Equation 13.[3]
(13.)
n =
2qB(Isignal + 'dark)
Where q is the charge of an electron, B is the measurement bandwidth in Hz, and I is current.
If we assume that dark current is much smaller than the signal current and solve for the root mean
square we can find the shot noise in the current coming out of the photodiode, as shown in Equation
14.
(/4)
in-ms
2
qBIsignal
By dividing the signal current by this expression we can find the signal to noise ratio, SNR,
for a system dominated by shot noise, given in Equation 15.
(05)
SNR
=
"'nR- 2qB
Isignal
19
Now we can clearly see that the signal to noise ratio is proportional to the square root of the
signal strength.
4. Light Source Selection
4.1 Overview of Available Light Sources
Several traditional options of Ultra Violet emitting light sources are available today
including various gas arc lamps and hot filament lamps. We will examine the relative merits and
shortcomings of deuterium arc lamps, mercury arc lamps, xenon flash lamps, quartz tungsten
halogen lamps, and UVC LEDs to compare their suitability for adsorption spectroscopy. The LEDs
are represented by a single 250 nm centered LED and by an array of eleven LEDs which are centered
10 nm apart from 220 nm to 320 nm.
There are many important factors to consider when selecting a light source for UV
absorption spectroscopy. These include the spectral range emitted by the source, total irradiance (or
optical power output) and its spectral distribution, output stability, efficiency, lifetime, startup time,
and the cost of the light source.[6]
4.2 Spectral Range
The spectral range emitted by a light source sets the limits for the wavelengths at which a
spectrometer can detect absorption. In Figure 6, we can see the range covered by each type of light
source.
20
Spectral Range (nm)
A
QTH Lamps
Xenon Flash Bulbs
Mercury Arc Lamps
X
Deuterium Lamps
250 nm Peak LED
UVC LED Array
100
150
200
250
300
350
400
500
450
Figure 6. Sptecr! ninge cf 'UClight sowces
As we can see, all of the sources except the QTH Lamp are capable of covering the UVC
range. The first three light sources have a large portion of their optical power concentrated in the
visible and infrared spectrums.
4.3 Stability
In UV absorption spectroscopy, the signal is a decrease in intensity of the light exiting the
flowcell caused by sample absorption. As a result, the process is extremely sensitive to short term
fluctuations in the intensity of the light source. In Figure 7, we can see the stability for typical light
sources of each type.
21
Stability (% p-p)
QTH Lamps
Xenon Flash Bulbs
Mercury Arc Lamps
Deuterium Lamps
250 nrn Peak LED
UVC LED Array
0.001
0.01
0.1
1
100
10
Figure 7. S'tbilityv oY U IC light SOUrCes
As we can see Deuterium lamps and LEDs have a similar level stability which is two orders
of magnitude better than any of the other sources.
4.4 Optical Power and Spectral Irradiance
Beer's law informs us that when we increase the light throughput of a flowcell at some point
the sample will saturate and be unable to absorb additional light. Therefore, arbitrarily increasing
light throughput does not always aid in detection. In fact, it can make detection more difficult. If the
steady state throughput is many orders of magnitude higher than signal, the sensor and associated
electronics will saturate if they have to small a small dynamic range, or if a large dynamic range is'
used, it could lack the sensitivity of a sensor with a much smaller dynamic range.[7]
However, if we consider varying the geometry of the flowcell, we will notice that the
etendue and, thus, the optical power throughput scales as the diameter of the flowcell squared, as
shown in Equations 8 & 9. Reducing the diameter of the flow cell reduces the minimum sample
volume required by the same factor squared, assuming the same absorption length is maintained.
Therefore, optical throughput must be increased proportionally when reducing flow cell volume to
maintain the same level of signal. As a result, the optical power provided by a source sets a lower
bound on the volume of flowcells used in conjunction with that source. In Figure 8, we can see the
22
average spectral irradiance, or optical power per nanometer of bandwidth across the UVC spectrum
typical for each type of source.
2
Average UVC Irradiance (pW nm- cm- ) at 50 cm
QTH
Lamps
Xenon Flash Bulbs
Mercury Arc Lamps
Deuterium Lamps
250 nm Peak LED
UVC LED Array
0.1
0,01
Figre N
10
1
verage irradiance ofUt C light sources
QTH lamps emit very little power within the UVC spectrum. All of the other sources have
1
significant overlap at around 0.1 pW nm cM-2. This level of irradiance allows most detectors to
operate below the range where Beer's law limits absorption.
The output of traditional sources generally have high sensitivities to temperature changes
and therefore are designed to have a single operating point. AIN LEDs exhibit very little change to
their peak wavelength either from temperature changes or when increasing or decreasing their
driving current.[8] As a result, we can easily modulate the irradiance of the LEDs by changing their
driving current without adversely affecting their spectral output.
Figure 9, shows a small handful of representative irradiance spectrum of the sources as well
as some Xenon arc lamps, which are very similar in most respects to mercury lamps, but with
smoother spectra.
23
100
-m
MERCURY XENON LAMPS (200 W)
LO
E
0
HALOGEN LAMP (24 V-150 W)
_O
CO
10
.0
0
E
1
z
-
XENON LAMPS(150 W)
0.1
XENON LAMPS (35 W)
CC
XENON FLASH LAMPS (20 W)
)
DEUTERIUM LAMPS
I .....
0.01
2 00
300
400
500
600
.. I
700
I
800
1000
900
WAVELENGTH (nm)
Figure 9. Spectrn of IC light sources
4.5 Power Requirement
Figure 10, gives the range of power requirements for each type of source.
24
Input Power Required (W)
QTH Lamps
Xenon Flash Bulbs
Mercury Arc Lamps
Deuterium Lamps
250 nm Peak LED
UVC LED Array
1
0.1
Fignre l0.
100
10
Pmver requirement ofIJU1C
1000
I1ight sources
All UVC light sources have small enough wall plug efficiencies that we can assume that all
of their power will effectively be dissipated as heat. The traditional light sources all have significant
heat output, l0s to
100s of Watts, which will need to be managed in such a way that it does not
adversely affect the sample or the spectrometer. Often this fact drives large facets of product
architecture for existing detectors, driving their size, overall power requirements. and eliminating
any possibility of portability.
4.5 Efficiency
Determining a common definition for efficiency that can fairly and equally apply to all of
these light sources is very difficult. The approach implemented here is based off of the intended
application; to carry out UVC absorption spectroscopy. Selectable wavelength detectors tend to have
a bandwidth of 5 nm.[9], [10] Therefore, to compare the efficiency of these sources we have
estimated the total optical power output between 250 nm and 255 nm from each spectral irradiance
curve, see Figure 9, and divided that number by the respective input power. The result is the
approximate efficiency of each type of source as it would be used for UVC absorption spectroscopy,
see Figure 11.
25
Efficiency
QTH Lamps
Xenon Flash Bulbs
Mercury Arc Lamps
Deuterium Lamps
250 mriPeak LED
UVC LED Array
10
6
10-S
Figure ii,
10
10:
10
100
L;fjcwiencv of UVC light sources
Although the efficiency of an array of LEDs is comparable to that of a Mercury arc lamp or
a Xenon flash bulb, their real benefit is that the individual wavelengths can be turned off and on
again as needed. This gives LEDs an advantage of at least two orders of magnitude in raw efficiency
over any other source. When adding in the ability for all of them to be turned off when they are not
actually being sampled, they have the potential to improve their effective efficiency by another two
orders of magnitude over all traditional sources other than Xenon flash bulbs.
4.6 Lifetime
For these lamps, lifetime is defined as the average time it takes for the peak irradiance of the
source to fall to half its initial value. In Figure 12, we can see the range of expected lifetimes for
each type of UVC light source.
26
Lifetime
QTH
Limps
R
Xenon Flash Bulbs
Mercury Arc Lamps
Deuterium Lamps
250 nm Peak LED
UVC LED Array
u
1000
2000
Fwigre
I
3000
4000
5000
6000
7000
9000
8000
Lifitime of/Ut-C light sources
The lifetime for a typical traditional UV lamp is 2000 hours. The lifetime for a UVC AIN
LED starts around 3000 hours if it is run continuously at its maximum drive current. However, if the
drive current is lowered from 100 mA to 20 mA, its expected lifetime goes up to 8000 hours. The
tradeoffs these LEDs exhibit between lifetime, age, drive current, and irradiance could be harnessed
by creating an active driver for each LED which gradually increases the drive current provided as it
ages. Such a light source would have a consistent irradiance throughout its lifetime, a strong signal at
the end of its life, and a significantly extended lifetime compared to a light source which started at a
higher brightness and decayed steadily to half intensity over its life.
These tradeoffs have a great deal of potential for other applications. One additional example
would be a software widget for the detector which allowed the user to adjust the drive current
themselves. In this way one device could equally well serve customers who want the higher
brightness even if it means replacing a lamp every 500 hours, as well as, customers who wanted to
stretch out the life of their lamp and live with less signal.
27
4.7 Startup Time
In Figure 13, we can see the average startup times for the various light sources. For most of
the lamps these startup times are to reach thermal steady state, and thus maximal optical stability.
For the Xenon flash bulbs and the LEDs it is the rise time to achieve approximately full brightness.
Startup Time (min)
QTH
Lamps
Xenon Flash Bulbs
Mercury Arc Lamps
7
Deuterium Lamps
250 nm Peak LED
UVC LED Array
10
is
20
25
33
30
Figmre 13. Starup tilnae Of U fC light soutCes
Source startup time is one factor which has tremendous impact on both the lifetime and the
efficiency of the system. When the light source requires a longer startup time it is consuming more
energy, none of which is usable for absorption spectroscopy. This lowers its overall efficiency. It is
also using up minutes of its lifetime which reduces its effective lifetime for detection. These effects
are compounded when considering user behavior. Long startup times prompt users to start up the
machine well in advance of when they want to start using it, and it encourages them not to shut it off
for short-duration breaks in running samples. These behaviors further exacerbate the reduction in
overall efficiency and effective lifetime.
28
4.8 Cost
Finally, we must factor in the relative cost of each light source and its power supply, which
in some cases costs more than the light emitter itself. To approximate the relative cost of the light
sources, as shown in Figure 14, we assumed that each can be purchased for one tenth its retail unit
value when purchased wholesale in production quantities. Note: these are not negotiated prices.
Cost
QTH
Lamps
Xenon Flash Bulbs
Mercury Arc Lamps
Deuterium Lamps
250 nrm Peak LED
UVC LED Array
0
50
100
150
200
250
300
350
400
450
Figure 14. E'stinuedcost olt/UC lighi sources
The array of LEDs currently exceeds the cost of any traditional light source. However, when
comparing the current cost for these light sources, it is important to remember the differences in the
product life cycle stage between the LEDs and the other sources. While the traditional sources are
very mature technologies, AIN based LEDs have only been commercially available for a few years.
If one were to be making decision regarding new product development it is important to keep in
mind that this graph is likely to look very different five years from now.
29
5. UVC LEDs
5.1 LED Operational Theory
LEDs produce light by taking advantage of the bandgap in semi-conductors. Electrons in a semiconductor junction are electrically excited across its bandgap and subsequently de-excite, radiating
light in the process.
5.2 AIN Based UVC LEDs
5.2.1 High Bandgap Semiconductors
The wavelength of light emitted by an LED is determined by its bandgap. To create LEDs
which emit at lower wavelengths requires semiconductors with larger bandgaps. Group III nitrides
have bandgaps from 2 to 6 eV which make them ideal candidates for LEDs emitting UV
radiation.[ II] AlN has a particularly broad bandgap, 6.1 eV, making it ideally suited for the
construction of low wavelength LEDs. In 2006, researchers from NTT Basic Research Laboratories
were able to develop a PIN type AIN LED with an emission wavelength of 210 nm.[12]
5.2.2 Dislocation Densities
As AIN LEDs emitting lower wavelengths have been developed, defects in the crystalline
structure have become a prominent problem. Defects can provide pathways for electrons to relax
thermally without radiating. As a result, the efficiency and light output of an LED go down while its
heat output goes up. Many AIN LEDs are built upon a substrate of either Silicon Carbide or
Sapphire. Mismatch between the crystal lattices of AIN and these substrate lead to dislocation
densities of 108 to 10" defects per square centimeter or even higher.[12]-[14]
5.2.3 Single Crystal AIN
UV LED chips grown on single crystal AIN and the substrate itself have nearly identical lattice
structures and therefore dislocation densities of only 103 to 104 defects per square centimeter is
possible, a tremendous improvement over LEDs grown on other substrates.[l 3]
30
5.3 Coupling LEDs to Optical Fibers
LED chip sources have viewing angles of
numerical apertures, NA =0.87 for
120' or more which gives them extremely large
120'.[ 15] This presents a challenge when coupling to UV
transmitting optical fibers which have numerical apertures of only 0.22.[16] Assuming the diameter
of the fiber is larger than the diameter of the source (LED chip) and not including Fresnel losses, the
theoretical maximum coupling efficiency as given by Equation 6. is 6.4 percent.
Assuming additional losses, a value of five percent LED to fiber coupling efficiency could
have been fairly assumed for the purposes of this paper. However, the use of specially designed
refractive-reflective micro lenses have the potential to increase the coupling efficiency, raising it as
high as forty percent[17]. If instead we assume the use of such a micro lens, we can assume a
coupling efficiency of twenty percent.
6. Wavelength Selection
6.1 Strategies Employed by Existing LC Detectors
UV absorption spectroscopy detectors for liquid chromatography can be grouped into three
major groups:
*
Single wavelength detectors such as older models based on the 254 mercury arc lamp line
*
Tunable UV detectors[9], [10]
*
Photodiode array detectors[ 18], [19]
Tunable UV and photodiode array detectors must both implement a strategy to distinguish
between absorption at different wavelengths of light. We will examine a typical strategy for each
type of detector and discuss their merits and shortcomings. After this discussion we can move on to
forming a strategy appropriate for use with UV LEDs.
6.1.1 Tunable UV Type Detectors
Tunable UV detectors use a rotating planar grating and a bandwidth aperture to select
wavelengths of light for detection. As shown in Figure 15, light passes through an entrance slit onto
a mirror which focuses the light onto the grating. The light is diffracted from the grating and
31
bounced off a second mirror which focuses the light through a bandwidth aperture onto the flow cell.
Detectors of this type tend to have a fixed bandwidth of 5 to 10 nm.[9], [10]
Entrance Slit
Focusing Mirrors
Grating
Bandwidth Aperture
Figure 15
TIunable U-V tvpe detector scheme
In a tunable UV detector, only one small section of the spectrum is sent through the sample
at a time. This greatly reduces the possibility of noise from multiphoton interaction events. It allows
the complete light signal to be referenced immediately before passing through the sample which can
noticeably increase the signal to noise ratio.
32
Because these detectors use motors to rotate their gratings, they are not able to dynamically
any
switch between wavelengths in any meaningful way. Fixed width apertures also do not permit
flexibility in the bandwidth sent through the flowcell.
6.1.2 Photodiode Array Type Detectors
Photodiode array detectors direct the light from the source straight to the flowcell, sending
slit onto
through the entire spectrum. After the light exits the flow cell it is directed through a narrow
at least
a concave grating. The grating diffracts and focuses the light onto an array consisting of
several hundred photodiodes. See Figure 16.
Entrance Slit
Flow Cell
Photodiode
Array
Grating
F1igllure 16. Photodiode arraY type defector scheme
33
Photodiode array detectors collect absorption data for the entire spectrum at one time.
Obtaining the overall spectral response of a sample can be a very powerful analytical method,
especially compared to gathering information about only a single wavelength for each sample run.
These detectors still reference the optical signal, however, since the reference pulls from the
entire spectrum at once it cannot eliminate as much noise as the reference in a tunable UV detector
can. Multiphoton absorption is a second order effect, nevertheless it has been observed in liquid
chromatography detectors.[7] By sending through the entire spectrum at once we lose the ability to
de-convolve multiphoton absorption and any other higher order interaction effects from the
conventional absorption signal. Although photodiode array detectors are not limited to one
wavelength per run, they still have fixed bandwidth bins and lack the ability to dynamically
modulate their characteristic specifications.
6.2 Alternative Strategy
At this point, we would like to present an alternative strategy for the UVC LED Detector
which allows dynamic manipulation of both the bandwidth and the specific wavelengths passed
through the sample during detection.
Light from the LEDs will be coupled to optical fibers which terminate at the entrance slit to
the optics bench. Light entering the slit will be incident upon a spherical grating, which will diffract
the light across the breadth of a digital micro-mirror device. Once the light is spread across the DMD
in this fashion, selecting the optical bandwidth and center wavelength is simply a matter of tilting
those columns of mirrors to the on position. From the DMD the selected light will be reflected by a
focusing mirror through a beam splitter and the flow cell to arrive at the reference and sample
photodiodes.
7. Unfolded Optical Path
7.1 Initial Path
The unfolded optical pathway for the strategy described in Section 6.2 can be seen in Figure
17, with the distance between elements, their orthogonal direction, and NA labeled. All of the
distances for the layout have been set by using the specifications for commercially available
34
components which closely matched the requirements for each respective optical element. Although a
more optimal design might be achieved using custom components, this approach permitted the
design process to proceed quickly and lends itself to rapid production of prototypes for initial testing
of the overall strategy. The reference photodiode and associated elements, such as the beam splitter,
have been omitted from the diagrams for clarity.
5636
.63
Mirror
PDI
LGFC
13710
131
13 45
4.31
170
70A3
0
,
Groing
Entrcance Sli
L)
5.80
NA.2
NA .22
3.81
NA.28
NA.49
NA .25
0NA
.21
Figure 17.
1
66NA .24
ifddec optical pathwav for alterative scheme
7.2 Stray Light
As shown in Figure 3, a mismatch between the NA for two optical elements in series can
result in either overfilling or under-filling the second of the pair. Both cases can result in excess stray
light propagating through the system. One easy way to mitigate this problem is to add aperture stops
at some of the elements to closely match the NAs throughout the system.
7.3 Path with Aperture Stops
An updated optical pathway including aperture stops which correctly couple the elements is
shown in Figure 18.
35
56.36
170
70.83
131.45
Mirror
PD1
LGFC
13710
137.
Grating
Ll
5.80
22
NA.28
NA.48
NA .21
.2
8NA
-
NA.2
Figure 18 Un/ided optica plthwaY with aperxure stops
Now that we have an optical pathway, the elements can be arranged in their folded
configuration, which is a packaging for the optics that provides a compact footprint for full
implementation in a consumer version of the detector.
8. Optical Layout
The detailed optical layout of the optics bench for the UVC LED Detector is shown in
Figure 19.
36
70.83
PD2
Designation
24
-2ASR I
126
LGFC
Mirror
BS
8.11*
5E: PD I
Ll
LEDs
Fiber
Entrance Slit
Grating
45
137.10
DMD
Mirror
BS
0 Grating
Li
\
Lght Dump
LGFC
PD]
2
DMD
PD2
Light Dump
,ibe.
DETAIL
Entrance Slit
B
SCALE 3: 1
ASG
ASM
LEDs
ASR
Description
Bank of modular LED
light sources
0.22 NA fused silica
fibers, 100 pm 0
0.040 x 2 mm aperture
285 g/mm
Rcurvature = 139.19
AXoze = 250 nm
7 mm
Lactive
Pitch 5.4 pm
120 orthogonal tilt
quarts window
Spherical mirror
Lf = 100 mm
Beamsplitter plate
0.50 mm
Refractive-reflective
lens, FC entrance
NAin =.48, NAout = .28
CT = 5.512 mm, 6.60
mm 0, 3.81 mm 0 CA
Light guided flow cell
Lhiuid = 10 mm, 0.275 NA
Sample photodiode
5.8 mm square active
area (PD1 and PD2 are
purchased as a
matched pair)
Plano-convex lens,
reference photodiode
ROC = 7.8 mm, CT =
5.25 mm, 12.7 mm 0
Reference photodiode
5.8 mm square active
area (PD1 and PD2 are
purchased as a
matched pair)
Aluminum member
coated with highly UV
absorbing coatinq
Grating aperture stop
55 mm 0
Mirror aperture stop 68
mm o
Reference photodiode
aperture stop 2 mm 0
Figure 19. Detailed optical kayout finr UVC LEI) detector
37
9. Expected Performance
This section lays out the calculations supporting the expected specifications for a UVC LED
Detector and then compares them to commercially available detectors of the tunable UV and
photodiode array variety.
9.1 Component Specification
Table I presents a list of specifications for commercially available components which could
be used to build this detector. The list is not exhaustive, but contains only the specifications relevant
to calculating the expected performance of such a detector and is provided purely for this purpose.
Two notable exceptions found in this table are the AIN LEDs with peak emissions below 250 nm
and the UVC reflecting DMD. Neither of these are commercially available at the time we are
writing, however, each has been demonstrated in the laboratory and can reasonably expected to be
commercially available in the near future.[12], [20], [21]
7iibie !. Component
LEDs:
Specificatins
DMD:
0
Peak wavelengths 220, 230,..., 320 nm
*
Pattern rate ~ 10 kHz
0
Optical power output > I mW each
*
Width of active array > 7 mm
Thermal output ~ I W
*
Micromirror pitch 5.4 pm
FWHM l2 nm
*
Overall efficiency at least 60%
*
Tilt 120 orthogonal to package
*
Window material quartz glass
S
S
0
S
Rise-time <
1ms
S
Packaging < 10 mm 0
Lifetime > 2000 hours
Spherical Grating:
Photodiodes:
*
Focal length ~ 140 mm
*
*
Linear dispersion ~ 25 nm/mm
& Length ~ 6 mm
*
Blaze X250 nm
0
Rise time less than 5 ps
"
Wavelength range 200 nm - 350 nm
.
Photosensitivity ~ 0.1 A/W
*
Efficiency at least 40% across entire range
*
Dark current less than 50 pA
*
Window material quartz glass
Spectral response at least 200 nm to 350 nm
38
Flowcell:
Fibers:
*
Core 0 100 pm
*
"
Material fused silica
& Length of flow 10 mm
*
NA
0.22
Light-guided
& Overall length ~ 50 mm
0
Entrance Slit:
NA.28
Analog to Digital Converters
*
Width 40 pm
*
Sampling Rate>1 MHz
*
Height 2 mm
*
Bits > 20
9.2 Bandwidth
The minimal possible bandwidth for the detector is simply the entrance slit width multiplied by
the linear dispersion of the spherical grating. For a slit width of 40 ptm and a linear dispersion of 25
nm/mm the minimal bandwidth is I nm. By tilting more columns of the DMD to the on position this
bandwidth can be increased in steps of 0.135 nm, which is set by the micromirror pitch of 5.4 pm.
9.3 Optical Throughput
We can calculate the optical power incident on the sample photodiode by multiplying the
cumulative irradiance of the LED sources,
I, by the bandwidth, A2, and by the net efficiency of each
optical element as shown in Equation 16 for n elements. The net efficiency is the product of the
reflectance or transmission efficiency of an element q; and the fill factor for an element F, which is
the ratio of the numerical apertures squared.
<'photodiode =
(10 AA)
77i Fj
(116)
i=1
39
Table 2 shows the approximate values of il and F for each element in the light path. F is listed as
I for all fill factors greater than unity. Total coupling efficiency from the LED to the fiber is
assumed to be 20 percent, which should be possible given use of a refractive-reflective lens as
mentioned in Section 5.3, and is shown distributed between the fill factor for the LED lens and the
fiber.
Table 2. Reflection, transmission, and couplingeiciencies of optics
Element
I
F
qnet
LED lens
0.9
0.45
0.41
Fiber
0.95
0.45
0.43
Entrance Slit
n/a
0.5
0.5
Grating
0.55
0.8
0.45
DMD
0.6
1
0.6
Mirror
0.9
1
0.9
Sample Path
BS
0.9
1
0.9
LI
0.98
1
0.98
LGFC
0.8
0.7
0.56
PDI
n/a
I
I
Entire Sample Path
0.011
Reference Path
BS
0.1
1
0.1
L2
0.98
1
0.98
PD2
n/a
1
1
Entire Reference Path
0.002
By solving Equation 16 using the values given in Table 2, we can find the flux incident on
the sample and reference photodiodes. Using the photosensitivity specification for the photodiodes
we can then find the signal and reference current. A summary of the outcome for several cases can
be found in Table 3.
40
Jable 3. &nple and reference photodiode responses for varriotsscenerits
Source and
A10
Bandwidth
nm
(mW/nm)
Single LED, 1 nm
I nm
Single LED, 5 nm
,$r
@0
Is
Ir
mW
mW
mA
mA
0.08
8.8 x 10-'
1.6 x 10-
8.8 x 10-5
1.6 x 10'
5 nm
0.08
4.4 x
10-
8.0 x 10'
4.4 x 10-
8.0 x 10-
Array, I nm
I nm
0.11
1.2 x 10-
2.2 x 104
1.2 x 10
2.2 x
Array, 5 nm
5 nm
0.11
6.1 x 103
1.1 x 10 3
6.1 x
10 4
10-'
1.1 x 10 4
9.4 Signal to Noise Ratio
Using the currents from Table 3 and Equation 15 for the shot noise dominated signal to noise
ratio, we can calculate the limiting SNR values for the same set of conditions. The result of these
calculations can be found in Table 4.
Table 4. SV R and noise levels f1r various scenarios
Source and Bandwidth
SNR
Noise Level
80 Hz Sampling Rate
Single LED, I nm
58,600
+ 1.7 x 10-5 AU
Single LED, 5 nm
131,000
+ 7.6 x 10-6 AU
Array of LEDs, I nm
68,500
Array of LEDs, 5 nm
154,000
1.5 x 10-i AU
t 6.5 x 10-6 AU
9.5 Sampling Rate
The detector would use a sampling rate of 80 Hz to display the chromatographs, mimicking
current methodology and keeping the shot noise down. However, the electro-optical components in
the detector are themselves capable of operating at much higher frequencies. As demonstrated in
Table 5 the DMD is actually the limiting factor for the system.
41
hible 5. A/ax aperaitingfrequencies
dynami clements
er
Element
Relevant Spec.
Value
Max Frequency
LED
Rise time
< I ts (dominated by
driving electronics)
I MHz
DMD
Pattern rate
9.5 kHz
9.5 kHz
Photodiode
Rise time
~ 2 pLs
0.5 MHz
ADC
Sample rate
1 MHz
1 MHz
If we consider using the DMD to timeshare among multiple wavelengths, it is apparent that
9.5 kHz will support continuous sampling of 118 channels at 80 Hz. Since the instrument has a
minimum bandwidth of I nm and a range of 110 nm, this feature can allow full sampling at every
wavelength simultaneously, mimicking the functionality of a photodiode array detector while
maintaining the benefits of the referencing strategy it shares with tunable UV detectors. Even at full
sampling over the entire range, the DMD has seven percent excess capacity which can be used for
automatic wavelength and dark current calibrations in the loop while sampling.
After the capabilities of the DMD have been fully utilized the rest of the components have at
least 50 times the ~ 10 kHz rate available. This excess capacity can be used to sample each pattern
created by the DMD multiple times and use averaging to filter out noise. Shot noise increases
proportionally to the square root of the sampling rate, and averaging decreases noise by the square
root of the number of samples, so the shot noise will not be improved. However, other sources of
noise present in the signal can effectively be filtered out by this method.
Designing a detector to have an electronic sampling rate of I MHz may not be directly
useful for chromatography, but it would be high enough to accurately characterize the instrument
dynamics. Such information could be used to optimize machine specific operation and troubleshoot,
resulting in increased performance and reliability for the end user. Such detailed knowledge about
the instrument dynamics could also be heavily leveraged for future improvements both in software
and hardware.
42
9.6 Resolution
9.6.1 Expected Resolution
As mentioned in Section 11.2, each column of micromirrors on the DMD corresponds to a
0.135 nm step in the wavelength. The width of the entrance slit prevents a bandwidth this small,
however this inherent step size in resolution is in fact the theoretical lower bound for the instrument.
In reality, several factors will increase this number to an empirically realizable resolution.
9.6.2 Resolution Sensitivity
In order to accurately estimate the resolution of the detector we will have to characterize its
sensitivity to several variables:
*
Distance between the entrance slit and the spherical grating
*
Distance between the spherical grating and the DMD
0
Width of the entrance slit
0
Angular misalignment between the DMD and the spectral plane of the grating
All of these can be determined to first order purely by the geometry of the light path.
If we take L, to be the position of the entrance slit relative to the grating then we can find the
relation between changes in L, and changes in resolution as shown in Equation 17 where R is
resolution and d, is the linear dispersion of the grating.
dR
= 2 IdLaI NA d,
dLa
The very same relation holds for the position of the DMD relative to the grating,
(17)
LH.
We already know the relationship between the entrance slit width and the resolution; they
are directly proportional to each other where dl is the constant.
43
Finally, we can define the relationship between angular misalignments,
dOIAII,
and
resolution. Using the small angle approximation, we find the expression shown in Equation 18 where
LDMD
is the length of the DMD array.
(1)
dR
dODMD = dODMD(LDMDNAdl)
Putting all of these together we can find an expression for the sensitivity of resolution to all
of these factors, given in Equation 19.
7
t
DMD
= 2 NA di(I3La + IaLHI) dws1 td, +
+ODMD(LDMDN
Ad1
0 9)
)
dR
dR9
aLa aL~awsite
This expression for sensitivity can serve as a basis for error budgeting during mechanical
design to ensure that the actual resolution of detectors with this optics bench meet their target
specifications.
10. Comparison to Existing UV Detectors for LC
Table 6 provides a comparison of the specifications of our detector design, as shown above,
with detectors currently in the market.[9], [10], [18], [19]
44
Table 6. Comparison of UVC
LED detector capabilities with tunable
UVC LED
PDA
TUV
Specifications
l' and photodiode array detectors
Wavelength Range
190-700 nm
190-700 nm
Variable within 210-320 nm
Sampling Speed
80 Hz
80 Hz
Variable from 80 Hz - 9.5 kHz
Noise level
2 x 10-6 to 6 x 10-6
3 x 10-6
6.5 x 10- (shot noise, no filter)
Bandwidth
5 to 6.5 nm
1.2 nm to 4 nm
Variable from
Resolution/Accuracy
I nm
I nm
1 - 110 nm
0.2 nm
Irradiance of source
0.1 ptW nm-1 cm- 2
0.1 pW nm' cm- 2
Variable up to 1.5 iW nm-' cm-2
Self-referencing
Yes
No
Yes
Self-Calibration
5 min. at warmup
5 min. at warmup
0.05 s (any time)
Start-up time
20 min.
20 min.
< 1 min.
11. Mechanical Design Considerations
A mechanical design of the optical system given here was undertaken by A. S. Prasad, which
is described in detail in "Mechanical Design of Optics Bench for UV-LED Based Liquid
Chromatography Detector".[22] On the basis of that design, a rapid prototype was created for
demonstration purposes as shown in Figures 20 & 21.
45
F.igure20. C4D nodel a/optics bench
Figure 21. Full scale looks-like prolotype
46
12. Conclusion
AIN based LEDs hold many advantages over traditional UVC sources. LEDs can provide a
level of stability previously only available from high-end deuterium lamps. They emit ten times the
irradiance of deuterium lamps, use less than one tenth the amount of power, and can do so for at least
twice as long. They are more efficient, produce less heat, are smaller, and can be switched on and off
instantly.
Digital micromirror devices can provide detectors with dynamic capabilities unimaginable
with more traditional optics. These include the ability to timeshare among over one hundred different
wavelengths, an almost instantly variable bandwidth, and the potential for multiple types of autocalibration.
The primary challenges currently preventing widespread adoption of AIN based LEDs are
the cost and the lack of commercialization of lower wavelength emitters from 210 nm to 240 nm. If
this technology follows a trend similar to many previous semi-conductor technologies, including
visible spectrum LEDs, both of these obstacles will quickly be overcome, paving the way for a new
wave of scientific instruments which are smaller, faster, more efficient, more accurate, and more
robust.
13. Future Work
The design we have lain out here is intended to operate in conjunction with existing liquid
chromatography systems. This was a very conscience decision intended to minimize the amount of
time until this technology is fully implemented, commercialized, and has gained acceptance through
the cumulative experience of many satisfied users. However, the technologies and techniques we
have discussed could be applied towards a design freed not only from deuterium lamps, but from the
entire stack of benchtop LC instruments. It is our hope that in the future, such optical systems will
form the heart of ultra-miniaturized LC systems which can be powered by a battery, carried in one's
pocket, and used in the field rather than in a lab.
In the more immediate future, the design presented herein has not benefited from an optical
CAD program, such as ZeMAX. There is undeniably a fair bit of further optimization of the design
which could be undertaken with such a software package.
Finally, we would consider it very worthwhile to build a fully functional prototype of this
design, perform all tests necessary to fully characterize its performance, and compare the results in
detail with the predictions. To date, we have been prevented from accomplishing just that by a lack
of expertise in the electrical engineering which would be quite necessary to bring a prototype to life.
47
14. Works Cited
[1]
S. Saxena, "Development of a Product Architecture for a LED Based Detector for
Liquid Chromatography," MIT, 2015.
[2]
J. M. L. and A. Thevenon, "A TUTORIAL ON SPECTROSCOPY," 2003.
[3]
A. Yariv, Introduction to optical electronicsHolt, Rinehart & Winston series in
electrical engineering, electronics, and systems, 2nd ed. Holt, Rinehart and Winston,
1976.
[4]
J. W. P. Peatross, Physics ofLight and Optics. 2008.
[5]
B. Wolf, Principles Of Optics. 1970.
[6]
Hamamatsu, "Xenon lamp, Xenon flash lamp, Mercury-Xenon lamp, Deuterium lamp
and Hollow cathode lamp."
[7]
T. M. Vickrey, Liquid chromatographydetectors. Vol. 23. CRC Press, 1983.
[8]
0. U. V. C. Leds, C. Is, and C. Is, "High Performance UVC LEDs for
Instrumentation."
[9]
T. Details and 0. Details, "Agilent 1260 Infinity Variable Wavelength Detector
Features , Technical Details, Specifications and Ordering Details Highest sensitivity
and selectivity for HPLC and RRLC."
[10]
Waters Corp., "ACQUITY UPLC TUV Detector Spec Sheet," pp. 1-3.
[11]
F. a. Ponce and D. P. Bour, "Nitride-based semiconductors for blue and green lightemitting devices," Nature, vol. 386, no. 6623, pp. 351-359, 1997.
[12]
Y. Taniyasu., M. Kasu, and T. Makimoto, "An aluminium nitride light-emitting diode
with a wavelength of 210 nanometres.," Nature, vol. 441, no. 7091, pp. 325-328,
2006.
[13]
J. Carlos Rojo, G. a. Slack, K. Morgan, B. Raghothamachar, M. Dudley, and L. J.
Schowalter, "Report on the growth of bulk aluminum nitride and subsequent substrate
preparation," J. Cryst. Growth, vol. 231, no. 3, pp. 317-321, 2001.
[14]
Y. Muramoto, M. Kimura, and S. Nouda, "Development and future of ultraviolet
light-emitting diodes: UV-LED will replace the UV lamp," Semicond. Sci. Technol.,
vol. 29, no. 8, p. 084004, 2014.
[15]
Crystal Is, "Crystal IS SMD UVC LEDs." pp. 1-8.
[16]
P. Bell and T. Wiggs, "Multimode Fiber," vol. 10, pp. 1-20, 2005.
48
[17]
C. Rooman, "Reflective-refractive microlens for efficient light- emitting-diode-to-
fiber coupling," vol. 44, no. September 2005, pp. 1-5, 2015.
[18]
Waters Corp., "ACQUITY UPLC PDA Detector Spec Sheet.".
[19]
T. Details and 0. Details, "Agilent 1290 Infinity Diode Array Detector Features,
Technical Details, Specifications and Ordering Details World ' s most sensitive and
fastest diode-array detector."
[20]
J. T. Fong, T. W. Winter, and S. J. Jacobs, "Advances in DMD-Based UV
Application Reliability Below 320nm," Proc. SPIE, vol. 7637, pp. 763718-7637186,2010.
[21]
J. Thompson, E. Braddom, J. Farris, and R. Mezenner, "DIGITAL PROJECTION OF
UV LIGHT FOR DIRECT IMAGING APPLICATIONS, DLP® TECHNOLOGY IS
ENABLING THE NEXT GENERATION OF MASKLESS LITHOGRAPHY," 2008.
[22]
A. S. Prasad, "Mechanical Design of Optics Bench for UV-LED Based Liquid
Chromatography Detector," MIT, 2015.
49