advanced planar light guide solar concentrators by
TRANSCRIPT
Advanced Planar Light Guide Solar Concentrators
by
Michael J. Brown
Submitted in Partial Fulfillment of the
Requirements for the Degree
Doctor of Philosophy
Supervised by Professor Duncan T. Moore
The Institute of Optics
Arts, Sciences, and Engineering
Edmund A. Hajim School of Engineering and Applied Sciences
University of Rochester
Rochester, New York
2013
ii
Biographical Sketch
Mick Brown was born in Glendale, California. He attended Oberlin College, and graduated with a
Bachelor of Arts degree in Physics and Chemistry. He began his doctoral studies in Optics at the
University of Rochester in 2007. He received the Master of Science in Technical Entrepreneurship
and Management from The Simon School of Business, University of Rochester in 2012. He
pursued his research in Optics under the direction of Duncan T. Moore.
The following publications were a result of the work conducted during this doctoral study:
M. Brown, D. Moore, G. Schmidt, and B. Unger, "Measurement and Characterization of
Dimpled Planar Light Guide Prototypes," in International Optical Design Conference, OSA Technical
Digest (CD) (Optical Society of America, 2010), paper JMB45P
US Patent #8,189,970 B2 “Light Collecting and Emitting Apparatus, Method, and
Applications.” 2012
US Patent Application #13/462,047 “Light Collecting and Emitting Apparatus, Method, and
Applications.” 2013
iii
Acknowledgements
I would like to thank my advisor, Duncan Moore, for his guidance and support throughout
my graduate studies.
I would like to thank Professor Roger McWilliams for teaching me for many summers. The
experience in his lab was critical to my interest and growth as a scientist.
I would like to thank Greg Schmidt for his boundless ingenuity and kindness. His guidance
provided on this project and in all other areas of my graduate career was extremely valuable.
I would like to thank Blair Unger for his pioneering work on dimpled light guide
concentrators. Without his persistence, creativity, and communication abilities, this project never
would have gotten started.
I would like to thank Dan Williams, Pete McCarthy, Rebecca Berman, Eric Christensen, and
Xinye Liu both for their contributions to this project and for making the office as pleasant as it was.
I would like to thank Lynn Doescher for all her help with scheduling and logistics and
Evelyn Sheffer for keeping track of all the financial transactions of the group.
I would like to thank Tim McCollum, Gregg Podojil, and all the other members of the
Rambus team for their hard work and long hours spent fabricating the prototypes.
I would like to thank Ed White for helping connect me to his wide network of vendors and
resources.
I would like to thank my parents Sidford and Amy for their consistent support and
guidance through my graduate work and all the rest of my life. I would like to thank my brother
Casey for consistent support and empathy as we went through grad school.
I would like to thank my wonderful girlfriend Ellie Kilpatrick for her many years of support
and encouragement through this process.
I would like to thank Hans Hovanitz and all the rest of my friends back home in Los Angeles
for providing such rejuvenating respite from the rigors of my work in Rochester.
I would like to thank my friends in Rochester, specifically Talor Walsh, Tim Baran, Andrea
Baran, and Adam Heiniger for making my time in Rochester infinitely more pleasant than it would
have otherwise been.
iv
Abstract
Commercially viable solar energy is a critical part of meeting the energy needs of society
while sustaining the environment. Concentrating Photovoltaics provide a capability to produce
more power for a given area with a potentially lower cost. Dimpled light guide concentrators
couple a large input area to a small strip of solar cell. Several of the advantages for this type of
concentrator will be discussed. Following the pioneering work done earlier by Blair Unger, three
new designs of dimpled light guide concentrator are presented, each with distinct advantages and
drawbacks. Ray tracing models are described that illuminate the optical potential of each of these
three design families. In addition to optical performance, lifetime and durability models are
developed to inform material selection and cost modeling. Component parts of all three novel
designs are measured and characterized. Three concentrator prototypes are fabricated and
assembled into functional solar modules. The measured optical performance of these systems
shows geometric concentrations ranging from 60x to 71x with optical efficiencies ranging from
19% to 33%. It is expected that this efficiency can be improved through well known processes in
future iterations. In one of the designs, a small portion of the input aperture was found that
coupled 89.6% of the light to the cell, demonstrating the potential for optical performance of
future systems.
v
Contributors and Funding Sources
This work was supervised by a dissertation committee consisting of Professors Duncan
Moore, James Zavislan, and Julie Bentley of the Institute of Optics and Mitchell Anthamatten of
the Chemical Engineering Department at University of Rochester. The initial work described in
Chapter 1 was funded by Abengoa Solar under Award Number 052809-001 and describes
prototypes built in collaboration RPC Photonics. This work is described in more detail in the
doctoral thesis of Blair Unger, who contributed much of the data described in Chapter 1.
Investigation of the gradient index solar concentrator was done with funding from the DARPA
Manufacturable Gradient Index (MGRIN) project under Contract HR0011-10-C-0111. The research
presented in Chapter 3 was done with support from Rambus International Ltd. under Award
Number – 056105-002The team at Rambus fabricated the critical components of the concentrator
prototypes.
vi
Contents 1 Introduction ................................................................................................................................ 1
1.1 Solar Power Overview ......................................................................................................... 4
1.2 Photovoltaic Market Landscape .......................................................................................... 6
1.2.1 Current Photovoltaic Technologies ............................................................................. 9
1.3 CPV Systems ...................................................................................................................... 12
1.3.1 Concentrator Theory ................................................................................................. 13
1.3.2 CPV Families .............................................................................................................. 15
1.3.3 Critical Components of HCPV Systems ...................................................................... 18
1.3.4 Current HCPV concentrators ..................................................................................... 25
1.4 Light Guide Concentrators ................................................................................................ 28
1.4.1 Alternative Light Guide Concentrators...................................................................... 28
1.4.2 Dimpled Light Guide Concentrators .......................................................................... 31
2 Modeling and Design ................................................................................................................. 37
2.1 Overview ........................................................................................................................... 37
2.1.1 Modeling Software .................................................................................................... 37
2.1.2 Modeling Methods .................................................................................................... 39
2.1.3 General System Parameters ...................................................................................... 40
2.2 Concentrator Optical Design and Performance Modeling ................................................ 43
2.2.1 Phase Space Performance metric ............................................................................. 43
2.2.2 Lenslet Design ........................................................................................................... 48
2.2.3 Dimple Tree Performance and Sensitivity ................................................................. 51
2.2.4 Two Stepped Guide Performance and Sensitivity ..................................................... 59
2.2.5 Axial Index Variation Performance and Sensitivity ................................................... 64
2.2.6 Wedged Stepped Concentrator ................................................................................ 67
2.3 Material Durability and Lifetime Modeling ....................................................................... 73
2.3.1 Ultraviolet and Infrared Absorption Models ............................................................. 75
2.4 Integrated Module Performance Modeling ...................................................................... 85
2.4.1 Cell Models ................................................................................................................ 87
2.4.2 Yearly Energy Output Simulations ............................................................................. 89
vii
3 Fabrication and Testing ............................................................................................................. 91
3.1 Overview ........................................................................................................................... 91
3.2 Lenslet Arrays .................................................................................................................... 93
3.2.1 Lenslet Interferometer .............................................................................................. 94
3.2.2 Refractive Lenslet Arrays for Stepped Systems ....................................................... 100
3.2.3 Reflective Lenslet Arrays for Dimple Tree Systems ................................................. 109
3.3 Dimple Arrays .................................................................................................................. 115
3.3.1 Wedged Stepped Dimple Arrays ............................................................................. 116
3.3.2 Two Stepped Dimple Arrays .................................................................................... 126
3.3.3 Dimple Tree Arrays .................................................................................................. 131
3.4 Assembled Concentrators ............................................................................................... 136
3.4.1 Concentrator Module Testing Methods .................................................................. 137
3.4.2 Acrylic Wedged Stepped Concentrator ................................................................... 143
3.4.3 Polymer on Glass Wedged Stepped Concentrator .................................................. 147
3.4.4 Reflective Dimple Tree Concentrator ...................................................................... 158
4 Conclusion and Future Work ................................................................................................... 164
4.1 Improving Concentrator Prototypes ............................................................................... 164
4.2 Material Research ........................................................................................................... 167
4.2.1 Gradient Index Material Research .......................................................................... 167
4.3 Lifetime Durability and Performance Degradation ......................................................... 170
viii
Table 1-1: LCOE projections for several competing energy technologies. ......................................................... 3
Table 1-2: Standard cost breakdown of a solar installation. ............................................................................ 24
Table 2-1: Development of reflective dimple tree light guide variants showing an optimal design and then
tracking various concessions for manufacturing purposes. The results of the earlier Generation 2.5 light
guide prototypes are also shown for comparison purposes. ........................................................................... 55
Table 3-1: The shunt and parasitic series resistances were calculated for the three solar cells integrated into
the concentrator modules. Ideally a solar cell will have a high shunt and low series resistance value, so
LG003 was the best of this cell batch. ............................................................................................................ 141
Table 3-2: The various sources of loss for the acrylic wedged stepped concentrator. The facet scattering was
the dominant loss mechanism, though the defocus caused substantial losses. ............................................ 147
Table 3-3: Summary of loss mechanisms for glass wedged stepped concentrator. ...................................... 157
Table 3-4: Summary of loss mechanisms for the reflective tree concentrator prototype. ............................ 162
ix
List of Figures
Figure 1-1: Energy consumption broken down by source. Renewables form a small fraction of the total
energy used by the United States. Reproduced with copyright permission of EIA. .......................................... 1
Figure 1-2: A photothermal trough from Acciona Solar (left, picture reproduced with copyright permission)
and a power tower from Abengoa Solar (right). ................................................................................................ 5
Figure 1-3: Regional breakdown of installed PV capacity. Image reproduced with permission of copyright
holder NPD Solarbuzz. Image available from http://www.solarbuzz.com ........................................................ 7
Figure 1-4: Global Demand Stack for PV materials. Image reproduced with permission of copyright holder
Greentech Media. Image available at http://www.greentechmedia.com/ ...................................................... 9
Figure 1-5: Production of PV systems by technology type in 2010. Crystalline Silicon is dominant, with CdTe
being the other material appropriating the greatest market share. Image reproduced with permission of
copyright holder Greentech Media. Image available at http://www.greentechmedia.com/ ......................... 11
Figure 1-6: Two example low concentration systems. The system on the left relies on flat mirrors to reduce
silicon use by approximately 2x(image reproduced with permission of copyright holder Zytech Inc.), while
the system on the right is a truncated CPC with a concentration of 3.5x(Image reproduced with permission
of copyright holder Intech Inc.). ....................................................................................................................... 16
Figure 1-7: A mid concentration photovoltaic module. The system only requires one tracking axis and uses a
trough mirror to focus incident light onto the solar cell. Image reproduced with permission of copyright
holder SunPower Inc. Image available at us.sunpowercorp.com ................................................................... 17
Figure 1-8: A schematic of a generic concentrating system. Assumptions are made about the other
components to inform targets for the concentrator that is developed in this thesis (component 3). ............ 19
Figure 1-9: The efficiency improvement of several different families of cells. Rapid growth, especially in the
multijunction cells, is continuing. Image reproduced with permission of copyright holder National
Renewable Energy Laboratories(NREL). Image available at http://www.nrel.gov.......................................... 23
Figure 1-10: The ideal Fresnel lens (left) will suffer from a minimum draft angle (center) and rounding of the
ideally sharp corners (right) which will substantially reduce optical efficiency. .............................................. 27
Figure 1-11: A large dish concentrator. This was a test module in Phoenix, AZ that is approximately 75 feet
in diameter. Image reproduced with permission of copyright holder Southwest Solar Technologies ........... 28
x
Figure 1-12: A schematic for the Morgan Solar rotationally symmetric concentrator system. ....................... 30
Figure 1-13: A schematic of the UCSD light guide concentrator. Light is focused onto coupling prisms that
redirect light to either side of the light guide. Image reproduced with permission of copyright holder The
Optical Society. ................................................................................................................................................. 31
Figure 1-14: Schematics for a refractive and reflective dimpled light guide system. ...................................... 32
Figure 1-15: A schematic of a dimple for use in a dimpled light guide. The injection prism is shielded from
light injected upstream by a bypass prism. ...................................................................................................... 33
Figure 1-16: When fabricated, the knife edge prototype had a rounded tail, substantial drafts on the side
walls, which were also rough. This substantially degraded performance. ...................................................... 34
Figure 1-17: The second generation light guide prototype (top), the designed dimple geometry (bottom left),
and an SEM image of the fabricated second generation dimple structure (bottom right). ............................. 35
Figure 2-1: A lenslet feature is created in Solidworks and then patterned appropriately using LightTools. ... 39
Figure 2-2: Tracker Accuracy is dependent on wind loading (left) and the acceptance angle determines the
fraction of energy available (right). Image reproduced with permission of copyright holder Green Mountain
Engineering. ...................................................................................................................................................... 41
Figure 2-3: ASTM G173-03 Direct Component. This is a standard spectrum representative of peak irradiance
in the United States .......................................................................................................................................... 42
Figure 2-4: The angular spectrum a) 1mm, b) 18mm, c) 36mm, and d) 52mm from the injection element.
The angular spectrum 1mm from the guide is that injected by the lenslet. The angular spectrum expands as
light travels further down the guide, and any light travelling with too large an angle (outside the blue
rectangle) is lost from the system. ................................................................................................................... 46
Figure 2-5: Offsetting the lenslet aperture causes the central ray to be deviated from normal to the plane
(Left). If the central ray is not deflected to travel in the plane of the guide the effective injection angular
spread will be much larger (Upper Right) than if it is deflected directly parallel to the plane of the guide
(Lower Right) .................................................................................................................................................... 49
Figure 2-6: A lenslet geometry designed to cater to a manufacturing process where the sag depth is
restricted. Allowing customized aperture shapes allows improved injection and guiding efficiencies. ......... 50
xi
Figure 2-7: Optimal optical performance requires a conic surface shape for the lenslets to achieve a tight
focal spot (left). A spherical lenslet surface will leave a large amount of spherical aberration and reduce
performance (right). ......................................................................................................................................... 51
Figure 2-8: The first generation "knife edge" dimple has been fused into a long strip to avoid the problems
associated with manufacturing the knife edge geometry. ............................................................................... 52
Figure 2-9: The dimple geometry of the "dimple tree" family of light guide concentrators. The injection
facets are highlighted in teal. ........................................................................................................................... 53
Figure 2-10: The efficiency fall off for various dimple tree light guide concentrators as the geometric
concentration increases. .................................................................................................................................. 56
Figure 2-11: The angular spread of a reflective dimple tree light guide with nearly ideal manufacturing
parameters (1µm fillet radii, 1° draft angle). The angular spectrum expands almost entirely laterally, with
the vertical expansion being due to the draft angles and fillets. The red rectangle shows the boundaries for
TIR containment in the light guide structure. .................................................................................................. 58
Figure 2-12: A Schematic of horizontal stepping. This represents a top view with the light purple squares
representing the lenslet apertures, the dark purple squares representing the injection facets, and the red
rectangles representing the chips. The relative size of the lenslets and injection facets determines both
concentration and how many horizontal steps can be taken. Shown are a 36x concentrator with 6 steps
(left), a 9x concentrator with 3 steps (center) and a 4x concentrator with 2 steps (right) .............................. 60
Figure 2-13: A schematic of vertical stepping. The guide layer increases in thickness so that each injection
facet will not interfere with light injected upstream. In this schematic, the thickness of the lenslet layer is
decreasing to compensate for the increased guide layer thickness. ............................................................... 61
Figure 2-14: A schematic of a two stepped guide. When a horizontal step would interfere with an upstream
injection facet, a vertical step is taken. ............................................................................................................ 62
Figure 2-15: The effect on the angular spectrum of a two stepped light guide for various levels of
manufacturing defects. More precise manufacturing tolerances allow isolation of vertical and horizontal
angles. .............................................................................................................................................................. 64
Figure 2-16: An axial index variation will reroute a ray injected with a steep angle to travel more directly
down the guide. This can be done with a mismatched guide layer or a GRIN in the guide layer. .................. 66
Figure 2-17: A model of a wedged stepped concentrator. The injection facets are joined to the facet directly
downstream by a slight wedge. This concentrator geometry uses horizontal stepping. ................................ 68
xii
Figure 2-18: The modeled optical efficiency for light injected into the guide at given distances. Overall the
system was designed to be 600mm long with a geometric concentration of 750x. Absorption and
manufacturing imperfections both reduce performance. ............................................................................... 71
Figure 2-19: The angular spectrum for a gradient index light guide. Manufacturing errors will circularize the
spectrum and reduce guiding efficiency of the device. .................................................................................... 72
Figure 2-20: The modeled effects of both a mismatched guide layer and gradient index guide layer compared
to the homogenous design ............................................................................................................................... 73
Figure 2-21: The UV transmission characteristics of PMMA compared with the solar spectrum incident on
the surface. The chain scission peak will cause damage to the material, but a radical scavenger can be added
to absorb this UV. ............................................................................................................................................. 76
Figure 2-22: The spectral overlap between solar radiation hitting the Earth's surface and PMMA. PMMA
shows excellent transmission in the visible, but significant absorption in the infrared at wavelengths longer
than 1100 nm. .................................................................................................................................................. 78
Figure 2-23: The thermal effects of a cone of light focusing onto an injection facet in PMMA. The
temperature differential is not projected to be more than a degree, even at the focal spot. ........................ 80
Figure 2-24: Simplified model of light guide IR absorption. Light is assumed to be injected evenly traveling
towards the chip at the left. ............................................................................................................................. 81
Figure 2-25: Heat dissipation by various lengths of PMMA light guide concentrators. Most of the absorption
is in the infrared. The absorption approaches the solar spectrum with a characteristic length determined by
the material absorption of the guide layer. ..................................................................................................... 83
Figure 2-26: Thermal model of temperature increase due to infrared absorption of a PMMA two stepped
light guide. In this model, the only means of dissipating heat is convection from the bottom surface
interacting with 300K air. ................................................................................................................................. 84
Figure 2-27: Characteristic summer and winter days in Phoenix (top) and Chicago (bottom). Phoenix has
much less cloud cover, and thus much more direct sunlight. High concentration systems are primarily only
able to collect direct sunlight. .......................................................................................................................... 86
Figure 2-28: Schematic for a triple junction solar cell. Each successive layer efficiently absorbs and converts
the appropriate spectral region while allowing lower energy light to pass through to the cells underneath.
Image courtesy of Solar Junction ..................................................................................................................... 88
xiii
Figure 2-29: Averaged monthly output from the system with insulation data taken in Phoenix in 2005. There
is substantial variability between the months, as this particular year had a unique distribution of weather in
addition to the annual seasonal variability. ..................................................................................................... 89
Figure 3-1: The wedged stepped concentrator model and illustrative schematic of the wedged stepped
concentrator (a), the two-stepped concentrator model and schematic (b), and the reflective concentrator
model and dimple schematic (c). ..................................................................................................................... 92
Figure 3-2: Definition of angles and transverse direction. The blades of lenslets were combined together to
form a master. .................................................................................................................................................. 94
Figure 3-3: Schematic of custom Twyman-Green interferometer used to characterize lenslet arrays. .......... 96
Figure 3-4: The output of a sample patch of the reference lenslet array. The Y Error is selected and shows no
notable systematic errors and a fairly small random variation. The phase profile is presented on the right,
from which an RMS error can be calculated. ................................................................................................... 99
Figure 3-5: The systematic spacing error can be seen in the reference lenslet array. The original
measurement showing an X spacing mismatch is shown on the right, and the left shows a consistent spacing
mismatch when the sample was rotated 90° and remeasured...................................................................... 100
Figure 3-6: A small patch of the refractive stepped lens array. The lenses have a rectangular aperture and
the optical axis is offset from the center of the lens aperture. In this picture, the light will be guided toward
the left. ........................................................................................................................................................... 101
Figure 3-7: White light interferometer measurement of the refractive stepped master. The red vertical line
on the surface map is fitted to both a variable radius and to the designed radius. While there is a slight
discrepancy, this only represents the 15% of the lenslet aperture near the apex. The microroughness of
these surfaces is approximately 10nm. .......................................................................................................... 102
Figure 3-8: An exaggerated schematic of the bowing of the lenslet array. This is the direction of bow as can
be seen visually. This will cause a focal shift in the Z direction as seen by the green arrow and will reduce the
X spacing shown by the black arrows. ............................................................................................................ 104
Figure 3-9: The surface profile of a representative refractive lenslet and the calculated RMS surface error of
each measured lenslet at the best focus. Both color scales are in mm. ....................................................... 105
Figure 3-10: The focus error in both Z (top) and X (bottom) directions. The Z error shows a low central
region consistent with a bowed array, while the X error shows a consistent spacing mismatch being too close
together. Both of these profiles are dominated by the bow in the lenslet array as expected. .................... 107
xiv
Figure 3-11: The Y coordinate error was dominated by an imperfection between the third and fourth row.
This is indicative of an imperfection preventing the two rows during assembly. .......................................... 108
Figure 3-12: The fitting of the bow of the refractive lenslet. This fit should yield a Z error of approximately
240µm at the corners of the lenslet array...................................................................................................... 109
Figure 3-13: The surface map near the apex of a reflective lenslet master (right) and the profile of a slice
through the apex. The surface roughness is quite low, and the fitting of the curvature shows a slightly
weaker curve than designed. ......................................................................................................................... 111
Figure 3-14: The RMS of the reflective lenslet array when used as a refractive system. The error is
dominated by the fourth row. It is believed that a scratch along the lenslet array caused a failure of the
phase unwrapping algorithm. ........................................................................................................................ 112
Figure 3-15: The RMS profile of a lenslet not in the fourth row. Substantial tooling marks can be seen in the
bottom section of this figure, but the profile is very good. There are highly visible tooling marks on the
bottom section of the lenslet, which is consistent with the two regions separated by the discontinuity in the
fourth row. ..................................................................................................................................................... 113
Figure 3-16: X (top) and Y (bottom) position errors of the focal spot for the reflective lenslet array. The Y
error is dominated by a scratch on the fourth row, while the X error is dominated by a bowing in the system.
........................................................................................................................................................................ 114
Figure 3-17: The first attempt at an injection facet (left) showed an obvious ripple. After refining
manufacturing parameters, the injection features were improved significantly (right). ............................... 117
Figure 3-18: Schematic showing dimple layer for the purposes of orienting white light interferometer
measurements. The colors of the channel are consistent with respect to height differences. .................... 118
Figure 3-19: White light interferometer measurements of a central region of an acrylic wedged stepped
dimple layer. The surface roughness of each channel in this region is approximately ±5nm. ...................... 119
Figure 3-20: An SEM image of the final wedged stepped dimple geometry. The feature replication is quite
good, though there is a considerable amount of debris that can be seen on the part. ................................. 120
Figure 3-21: White Light Interferometer measurements of polymer on glass wedged stepped dimple
geometry. Both the middle region and the high region are very smooth surfaces. The casting on glass
process introduces some roll off near steam height transitions, as can be seen in the bottom figure near the
abrupt height transition. ................................................................................................................................ 122
xv
Figure 3-22: White light measurement of facets of wedged stepped dimple array cast onto glass. Substantial
variations are observed from facet to facet, but there are several regions that will have a small amount of
scattering. ....................................................................................................................................................... 124
Figure 3-23: Scattering fraction from injection facet surface is a function of RMS surface roughness. ........ 125
Figure 3-24: A small patch cut of the two stepped dimple geometry. Channels are still visible, but the flat
regions are relatively consistent. ................................................................................................................... 127
Figure 3-25: White light interferometer measurements of the two stepped patch cut. The flat region has
clearly defined channels separated by a fraction of a micron while the injection facet cut shows signs of tool
wear. ............................................................................................................................................................... 129
Figure 3-26: The master cut for the two stepped dimple geometry showed substantial damage. The deeper
cut appears to have ripped away the base substrate through the nickel coating, and caused deep gouges
over a substantial portion of the dimple layer. .............................................................................................. 130
Figure 3-27: An SEM image of the dimple tree patch cut. While there is substantial debris and visible marks
ear the facets, the geometry is roughly correct. ............................................................................................ 131
Figure 3-28: The first cuts of the reflective tree design showed substantial ringing along the guiding axis and
a sharp height transition along the transverse axis. There was a substantial portion of the cut in which these
errors were much more pronounced and scattering from this section of the part was easily visible with the
naked eye (bottom). ....................................................................................................................................... 133
Figure 3-29: The final dimple tree cut showed substantial dampening of the ringing near the facets. While
the magnitude is not substantially lower for the master, the distance covered by the ringing will be
substantially reduced. The low frequency also will reduce the scattering effect. ........................................ 134
Figure 3-30: The injection facet of the reflective tree guide system showed substantial roughness, mostly
along the guiding direction (right to left in this figure). The marks were consistent from facet to facet. .... 135
Figure 3-31: The three final concentrator modules that were fabricated and characterized for this thesis.
The acrylic wedged stepped guide (left), acrylic reflective tree guide (center), and polymer on glass wedged
stepped guide (right). ..................................................................................................................................... 137
Figure 3-32: A schematic of the University of Rochester Solar Simulator. This produces an approximately
uniform 20cm diameter beam that matches both the spectral content and angular extent of the sun. ...... 138
Figure 3-33: The dual diode model of a solar cell. Image from pveducation.org .......................................... 139
xvi
Figure 3-34: Dark IV curves for the three silicon solar cells mounted to the three light guide concentrators.
The cells were chosen as the closest to ideal fits from an available batch of silicon cells. ............................ 140
Figure 3-35: Scanning the illuminated slit allows determination of how far injected light has propagated
within the concentrator before reaching the solar cell at the output face. ................................................... 141
Figure 3-36: The apparatus for measuring optical efficiency at different sections of the concentrator input
aperture. Two rotation axes and a translation axes are automated. ............................................................ 142
Figure 3-37: The optical efficiency falloff of the acrylic wedged stepped prototype. A small slit aperture
illuminated a small portion of each lenslet array and found the optimum pointing to determine the efficiency
falloff. ............................................................................................................................................................. 144
Figure 3-38: The field of view of the acrylic wedged stepped concentrator at the front lenslet array (15 mm
from cell) and at the rear lenslet array (90mm from cell). The profile is fairly broad and the center moves
substantially from lenslet to lenslet. .............................................................................................................. 145
Figure 3-39: The optical efficiency falloff for the polymer on glass wedged stepped guide. ......................... 149
Figure 3-40: The normalized field of view from a slit aperture of the polymer on glass wedged stepped guide.
The field of view much more closely resembled the design specifications than for the other prototypes. .. 150
Figure 3-41: The model of the bowed lenslet array. The bowing of the lenslet array was approximated from
the unattached lenslet array measured in the lenslet interferometer. The amount of delamination was
controlled by changing the thickness of the low index layer. ........................................................................ 151
Figure 3-42: The optical efficiency of each lenslet in the array for a system with the observed bowing defect
and no others. The observed immersed region covers slightly more than half the aperture....................... 152
Figure 3-43: The optical efficiency of the glass wedged stepped design near the sweet spot. The variation in
efficiency within the immersed region of a lens array is caused by a substantial randomness in the injection
facet roughness. ............................................................................................................................................. 153
Figure 3-44: Detailed field of view of sweet spot measured with collumated HeNe beam. The field of view
very closely resembled the designed system, and there is relatively little variation within the injection facet.
........................................................................................................................................................................ 154
Figure 3-45: Laser Scans at the front and back of the glass wedged stepped system. The guiding loss from
the best facet near the back of the guide to the best facet found near the front of the guide was 22.7%. .. 156
xvii
Figure 3-46: Efficiency falloff for reflective tree concentrator system, taken at optimum pointing angle. ... 159
Figure 3-47: Normalized field of view of acrylic reflective tree concentrator for the lenslet array closest to
the cell (left) and farthest from the cell (right). ............................................................................................. 160
Figure 3-48: The small slit scan of the field of view for the front lenslet of the reflective tree concentrator.
The field of view is substantially blurred, indicating a focal length error, but the center moves almost a full
degree and systematically, indicating a clocking error. ................................................................................. 161
Figure 4-1: A representative injection facet of the two stepped light guide system. The pattern of scratches
was consistent over all facets measured on this part and is thought to be due to tool wear. ...................... 166
Figure 4-2: A comparison of the clarity of the undiffused glass (right), the diffused GRIN sample pretreated
with K+ ions(center), and the diffused GRIN sample without pretreatment (left). Photo by G. Schmidt ..... 168
Figure 4-3: Surface heights at the edge of 2mm thick slab of glass that has not been diffused (left),
pretreated for 6 hours (center), and pretreated for 16 hours (right). The 2D plots show the vertical and
horizontal slices of the data represented by the corresponding line color.................................................... 169
Figure 4-4: Fringe pattern and calculated refractive index profile for diffused GRIN sample. ...................... 170
1
1 Introduction
One of the critical challenges to be confronted in the near future is sustainably generating
sufficient electrical power to meet the rapidly increasing demand of modern society. Demand for
electrical power continues to accelerate as the population increases and society modernizes. This
demand is largely met by consumption of fossil fuel reserves, primarily coal, natural gas, and
petroleum. These sources are not sustainable, as natural reserves are finite and become less
efficient to extract as they are depleted. The use of these power sources also poses an
environmental hazard with potentially dire negative consequences.
Development of renewable sources of energy is critical to fill the void that will be left by the
depletion of natural fossil fuel reserves. Renewable energy sources also help to alleviate the
environmental strain that consumption of fossil fuels creates. Renewable sources currently
account for a small fraction of energy generation, as shown in Figure 1-11. In order to make a
greater impact, renewable energy technologies must be advanced to the point where they can be
Figure 1-1: Energy consumption broken down by source. Renewables form a small fraction of the total energy used by the United States. Reproduced with copyright permission of EIA.
2
cost competitive with fossil fuel sources. There are many different sources of renewable power,
and the economic viability and scalability of each is currently uncertain. Thus concurrent
development of many potential technology pathways is critical to the widespread use of
renewable power generation.
Renewable energy sources have only captured a small fraction of the energy market, as they
are not currently cost competitive with current electricity generation methods. The number of
competing technologies that have potential for utility scale electricity generation makes
quantitative comparison somewhat uncertain. Forecasting energy prices in the future is imprecise,
as the behavior of coal and gas varies dramatically depending on the source of the information. In
addition, a number of disruptive technologies might cause substantial price fluctuations if they
should achieve cost parity first. While determining actual commercial viability is an extremely
complicated process, it is impractical to calculate accurately for a developing technology, and thus
a simpler metric is useful for guiding the development process.
The Levelized Cost of Energy (LCOE) is designed to measure the cost of generating power
over the expected lifetime of a plant ($/kWh). This is fairly simple to calculate for a given system
and provides a simple means of quantitatively comparing two systems. This provides a useful
target for development, as any competitive system must surpass the LCOE of currently available
power generation techniques in order to achieve widespread adoption. The LCOE is calculated by
3
where N is the expected life time of the plant in years, AO is the Annual Operations Cost, DR is the
discount rate, RV is the residual value, and SDR is the system degradation rate.2 This takes into
account the performance and costs of the plant over the course of its life time, and provides an
approximate scalar metric that can be compared with other technologies.
The Energy Information Administration has released projections for the LCOE of several
different technologies assuming plants coming online in 2017. These were based off a set of
assumptions taking into account current trends in each of these technologies3. Any disruptive
technology must at least meet these projections, and thus this provided a useful target for
designing and developing a system. The projections are given in Table 1-1.
Table 1-1: LCOE projections for several competing energy technologies.4
Minimum Average Maximum
90.5 97.7 114.3
102.5 110.9 124
107.2 111.4 118.7
77 96 112.2
57.8 88.9 147.6
119 152.7 238.8
176.1 242 386.2
Advanced Nuclear
Renewables
Wind
Hydro
Photovoltaic
Photothermal
Advanced Combined
Cycle Natural Gas
59.5 66.1 81
56.8 63.1 76.4
Cost Range for Total System Levelized Costs
(2011 $/MWh) for plants entering service in 2017
Dispatchable Technologies
Conventional Coal
Advanced Coal
Plant Type
Conventional Combined
Cycle Natural Gas
4
Current projections have natural gas-fired plants being the dominant source of energy
with an expected average cost of just over $0.06 per kWh. This is used as a goal for developing the
system described in this dissertation and provides a baseline for comparing alternative
technologies and assessing viability for future research and development. This is a long term
target for any power technology that aims to achieve grid parity in the near future, but before
widespread adoption in the utility scale market, many of these technologies can leverage other
advantages to penetrate smaller markets. This dissertation focuses on the development of a
specific solar technology, and follows the development efforts aiming at initially competing with
other solar power technologies with a proposed path toward competitive utility scale generation.
1.1 Solar Power Overview
Solar power encompasses a wide variety of technologies designed to convert the sun’s
energy into electricity. The two primary families for solar power are photovoltaics, which convert
the sun’s energy directly into electricity using the photoelectric effect, and photothermal, which
uses the sun’s energy to heat an intermediate to run a thermal engine. These two technologies
are growing areas of research and are being deployed with increasing frequency and scale.
Current levels of development and deployment show uncertain, though promising, potential for
both families of technology. The two technologies rely on fundamentally different scientific
principles, each with its own merits and drawbacks for commercial use.
Photothermal technologies rely on a concentrator to focus light onto a heat exchanging
medium. There are many different incarnations of photothermal technologies, but the most
common are the trough receiver and the power tower. Photothermal troughs are designed to
focus light at a relatively low (approximately 20x) concentration onto a tube containing an
5
absorbing fluid, which is then used to run a heat engine.5 These typically have efficiencies of
between 12% and 18% due to the relatively low temperature differential they can achieve with
these concentrations. The advantages of these systems are that they only require one tracking
axis, which substantially simplifies mounting and installation, and these systems can be scaled to
relatively small installations. The power tower employs a large area of mirrors designed to focus
light onto a receiver situated atop a tower in the center of the mirror array, which is heated to
temperatures of up to 1,000 °C. These power towers are much more efficient, as the efficiency
scales with the temperature differential. These plants can achieve efficiencies in the range of 22-
26%.
Photovoltaics rely on the photoelectric effect to excite an electron into the conduction
band of a semiconductor. A junction of p-type and n-type semiconductor cause the excited charge
carriers to flow preferentially in one direction which creates a voltage across the junction. The
spectral region that excites a carrier and the energy that is gained from each exciting photon are
intrinsic properties of the semiconductor junction used. A material with a higher band gap more
efficiently extracts energy out of the high energy region of the sun’s spectrum, but lower energy
Figure 1-2: A photothermal trough from Acciona Solar (left, picture reproduced with copyright permission) and a power tower from Abengoa Solar (right).
6
photons do not excite an electron, and thus cannot be converted into electricity. In order to
optimally convert as much of the sun’s energy as possible, different PV junctions can be stacked
(multijunction cell), so that each junction utilizes the part of the spectrum it can optimally convert.
PV systems comprise a much larger market currently than photothermal systems due to
several key advantages. PV systems have a much wider range in the scale at which they can be
deployed. For many applications such as remote stations, residential, or portable power,
photothermal systems are too large to be practical. PV systems also do not require the large heat-
to-electricity conversion infrastructure. The primary advantage of photothermal systems is that
the heated medium can be stored much more easily than the power generated by a PV system,
which must be stored in a battery or some other means. This allows the capacity factor of the
photothermal system to be much higher, as the power is available for use when needed for a
lower price than storage for PV systems.
1.2 Photovoltaic Market Landscape
While photovoltaics have not become a major player in the utility scale power market,
they are being used currently in a wide variety of smaller markets. Photovoltaics are
advantageous for off grid applications and distributed generation, as they scale down to almost
arbitrarily small size. Thus PV materials are currently being used be the construction and
transportation industries for powering small installations where grid connectivity is impractical.
On a smaller scale, PV materials are frequently used to power consumer electronics that have low
power consumptions such as calculators and watches.
7
The market for photovoltaics has continued to grow rapidly, with new technologies and
manufacturing development continuing to drive industry growth. As of 2011, slightly over 37GW
of installed PV capacity existed around the world. Most of installed PV capacity has been centered
in Europe, where strong government support has driven widespread implementation, especially in
Germany. This dominance was expected to weaken over the coming decades as the markets in
North American and Asia expand while economic uncertainty was slowing adoption in Europe.
Figure 1-3 shows the regional breakdown of installed PV capacity.6
Figure 1-3: Regional breakdown of installed PV capacity. Image reproduced with permission of copyright holder NPD
Solarbuzz. Image available from http://www.solarbuzz.com
The price of solar energy varies substantially by region, technology employed, and
financing agreements. Many photovoltaic technologies are relatively immature, and thus it is
difficult to get an accurate levelized cost of energy from these systems. It is difficult to accurately
quantify the performance over the lifetime of many PV technologies, and thus the value of
implementing these technologies is uncertain. An alternative performance metric is the cost per
8
watt produced during peak conditions. This metric does not take into account the lifetime or
degradation of the plant and also fails to consider performance at suboptimal conditions.
Problems with this metric can make comparisons misleading, but it provides a simple and readily
calculated approximation to judge performance.
PV systems provide advantages beyond strict power generation, and thus there is still
demand for PV modules even though the system is more expensive than alternative fuels. These
systems have been adopted at remote locations where grid connectivity is unavailable and some
other specialized applications. Customers who are sensitive to environmental impacts or seeking
to market a “green” image may choose photovoltaics even if more expensive than alternative
forms of power generation. The demand for photovoltaics increases as the price decreases and
more markets become viable. The demand stack for PV systems shown in Figure 1-47 illustrates
the increase in demand for each region as the price of this energy varies.
9
Figure 1-4: Global Demand Stack for PV materials. Image reproduced with permission of copyright holder Greentech
Media. Image available at http://www.greentechmedia.com/
In order to compete with conventional fuel sources, PV technologies must surpass $1/W.
The US government launched the Sunshot Initiative in 2011 with the goal of reaching this cost
target for PV systems.8 If this cost target can be eclipsed, then PV technology can begin to
penetrate the utility power generation market, which expands the potential greatly. To compete
on a utility scale, more accurate information about the lifetime performance of a solar plant will
have to be gathered and compared with alternatives.
1.2.1 Current Photovoltaic Technologies
A great deal of research and development is being focused on new solar technologies, yet
the dominant technology for terrestrial solar power, both residential and utility, remains flat plate
10
silicon. Silicon is a fairly mature technology that has been used for solar cells for decades.
Advances in processing have increased the efficiency of deployed modules9, but most of the
development has been aimed at improving manufacturing methods in order to reduce costs.
Flat plate silicon modules efficiently convert photons with wavelengths between 500nm
and 1100nm and typically have efficiencies of approximately 15% when deployed in the field. This
technology is relatively mature, and thus the risks associated with deployment are minimal and
well known. The relatively high conversion efficiency and process maturity are the primary
reasons that silicon continues to dominate most PV applications. The cost of silicon is primarily in
the actual material and material processing, and thus the module price will continue to be tied
closely to that of raw silicon. This has been an extremely volatile commodity, as silicon is widely
used by the electronics industry, creating extreme fluctuations in price.
The maturity of Silicon photovoltaics is an advantage and a limitation, as the performance
is approaching a fundamental theoretical maximum. The cost is unlikely to be dramatically
reduced, as the primary cost driver is the cost of the raw Si. Other technologies aim to avoid these
limitations, and thus potentially dramatically lower the cost of producing solar power. The
technology that has reached the most widespread deployment is Cadmium Telluride (CdTe) films
deployed primarily by First Solar10.
CdTe has substantially lower conversion efficiency when compared to Silicon, but requires
only a thin film of material to be deposited on the substrate, and thus the cost of producing a
module is drastically reduced. This is a much newer technology than flat plate silicon, and thus
research continues to increase the efficiency11. These systems have achieved substantially lower
$/W than available Si modules, but the lifetime of these systems is unclear, and thus the LCOE is
11
uncertain. Active research is continuing in quantifying and extending the lifetime of these systems,
though they are still deployed on a large scale (Approximately 22% of the PV systems in the US).
Silicon and CdTe account for the vast majority of solar power plants that are currently
running. Other systems are primarily still in the pilot plant phase while the manufacturing
readiness level increases towards large scale production. Concentrating Photovoltaics are among
these fledgling technologies that have only been deployed in small demonstration facilities. Figure
1-5 shows the installation level of key solar technologies, and highlights the dominance of Silicon,
with CdTe being the only competing technology to appropriate significant market share.
Figure 1-5: Production of PV systems by technology type in 2010. Crystalline Silicon is dominant, with CdTe being the
other material appropriating the greatest market share. Image reproduced with permission of copyright holder
Greentech Media. Image available at http://www.greentechmedia.com/
12
1.3 CPV Systems
Concentrating photovoltaic systems aim to achieve parity with alternative systems by
concentrating incident light onto a relatively small area of active photovoltaic material. The
primary driving cost of current PV technologies is the cost of the actual photovoltaic material.
CPV systems aim to reduce the material cost and employ relatively inexpensive optical
components to collect sunlight and concentrate onto the reduced area of photovoltaic material.
In order to surpass conventional flat plate PV technologies, the benefits of concentrating
systems must make up for the added system complexity and costs. The cost savings comes in the
form of reduced chip area, and thus cost. If the concentration is high enough, more efficient cells
become cost effective. Though these materials would never be viable in a flat plate system with
no concentration, but if the required area can be reduced by a factor of several hundred, the
increased efficiency can make up for the increased cost.
Concentrating systems add a varying degree of complexity to the system, and the costs of
these additional components must be minimized to make CPV systems viable. The concentrator
optics constitute an additional cost, though different designs of concentrator can vary orders of
magnitude in cost per area. The mount for a concentrating system is much more expensive than
for traditional flat panel systems, as many designs require some sort of active tracking mechanism
to maintain their alignment with the sun. Many systems also require a thermal management
system to keep the temperature of the PV material and potentially other components within a
designed operating temperature range.
13
1.3.1 Concentrator Theory
One of the critical system design decisions for CPV systems is the design of the concentrator,
specifically what level of geometric concentration to operate at. The concentration ratio
(collected area/PV chip area) dramatically affects what PV materials can be used in the system, as
high efficiency cells are expensive enough that they only become cost effective at high
concentrations. The primary drawback for high concentration systems is that a high concentration
dictates a system with a small acceptance angle. High concentration systems must be mounted
onto a tracking system to keep them aligned with the sun. This adds an additional cost to the
system that can be a driving cost for the system in many cases.
The tradeoff between concentration and acceptance angle is a physical limit described by
Étendue. Étendue is a property that takes into account both the spatial and angular extent of light
travelling through an optical system. Given a ray passing through an entrance plane at (x,y) with
direction cosines (L,M,N) in a material with refractive index n, the beam passing through the
system can be described by a ray bundle with spatial shifts (dx,dy) and angular shifts (dL,dM). The
generalized étendue is described by
Let the same be true of the exit space, with coordinates and material being represented by the
same symbols prime (‘). For a nearby ray with coordinates (x+dx, y+dy, L+dL, M+dM), the
generalized étendue is invariant, and thus for a lossless system, the étendue at the input plane
and the output plane will be equal.
14
An illuminating way to interpret étendue conservation is by combining the direction
cosines and index of refraction into an “optical momentum”. The flux through an optical system
can then be considered as a volume in the 4 dimensional space given by (x,y,p,q). This has been
likened to a four dimensional space in fluid dynamics, and thus has taken the name phase space.
As this input beam travels through an optical system, the volume cannot be compressed in phase
space, though it can be distorted and/or diluted.
The goal of an optical concentrator is to transfer an input plane with a large area and a
comparatively narrow angular spread to an output space with a small area and a larger angular
extent. For a simple system in which the angular extent is spatially invariant at both entrance and
exit surface, the concentration is given by the equation below, with the maximum concentration
occurring when the angular extent at the output face is a full 2π steradians.
The major implication of this equation for CPV systems is that increasing the concentration
ratio necessitates a smaller system acceptance angle. The sun itself has a fairly small angular
extent (approximately ±0.26°), but the rotation of the earth causes the sun to effectively sweep
over a large range of angles. Concentrating systems must either have a large enough acceptance
angle to produce power at various times throughout day, which requires a low concentration ratio,
or the system must be mounted on a tracker that adjusts the pointing of the module to maintain
the module’s alignment with the sun. Higher concentration systems have narrower acceptance
angles, and thus demand more accurate trackers. These high precision trackers are substantially
more expensive.
15
1.3.2 CPV Families
While CPV systems can have geometric concentrations anywhere from just over one up to
potentially several thousand, most systems under investigation fall into one of three different
families. There are reasons to investigate each of these different concentration regimes, and each
has potential for success and barriers that must be overcome before widespread deployment.
1.3.2.1 Low Concentration Photovoltaics (LCPV)
Low concentration systems are closely related to flat panel systems. The sun typically only
covers most of the East-West range in the sky over the course of a day, but the North-South
variation of the sun is defined mostly by the tilt of the Earth’s rotation axis. This limited variation
allows a small degree of concentration without any tracking systems. With a static mount, the
large angular variation of the sun dictates that concentration ratios must be low, and these
systems typically operate between 1x and 3x concentration.
The primary advantage of this family of designs is the possibility of incorporating advances
from current flat panel technology. These systems do not require different mounts, different cells,
or additional thermal management, and thus can act as a simple retrofit for flat panel systems that
reduces the amount of actual PV material required. To be cost effective, these optical systems
must be extremely inexpensive and thus the optical design of these concentrators tends to be
relatively basic. A few sample LCPV systems are shown below in Figure 1-6. Simple TIR
concentrators that can be made through high volume processes such as injection molding are
common.
16
Figure 1-6: Two example low concentration systems. The system on the left relies on flat mirrors to reduce silicon use by approximately 2x
12(image reproduced with permission of copyright holder Zytech Inc.), while the system on the
right is a truncated CPC with a concentration of 3.5x13
(Image reproduced with permission of copyright holder Intech Inc.).
The viability of these systems is critically linked to the success of flat panel systems.
Currently, the cost of a Silicon flat plate module is heavily influenced by the cost of the actual PV
material, but the price of silicon is expected to drop relatively rapidly. The potential savings of
LCPV designs are in reducing the actual chip cost by a factor of up to three, but if the cell cost is no
longer the driving cost of the module, these systems lose the advantage over the simpler flat plate
technologies.
1.3.2.2 Mid concentration Photovoltaics
Mid concentration designs aim to balance the added complexity required for a
concentrating system with the potential benefits of more efficient cells. A key parameter in
determining the optimal concentration to design the system for is the future cost of high efficiency
solar cells. Projections of the cost of these cells are highly uncertain, and cost estimates can vary
by more than an order of magnitude. Mid concentration systems usually seem to employ a higher
17
efficiency cell to gain added power. An example of mid concentration designs that is being further
investigated is shown in Figure 1-7.14 This system uses a shaped trough reflector to concentrate
light onto the PV material and requires only one tracking axis.
Figure 1-7: A mid concentration photovoltaic module. The system only requires one tracking axis and uses a trough
mirror to focus incident light onto the solar cell. Image reproduced with permission of copyright holder SunPower
Inc. Image available at us.sunpowercorp.com
Mid concentration systems seek to gain the benefit of higher efficiency cells with
dramatically reduced cell costs while balancing the expenses of adding complexity to the optical
system and tracker. This family of systems operates in the general range of 25x to 75x
concentration. By keeping a moderate concentration, these systems place lower demands on the
tracking system, and thus either a single axis or simple two axis tracking system can be used. This
can keep mounting costs lower and can still use relatively simple high volume optical systems.
18
This concentration also frequently allows the use of passive cooling systems which are much less
expensive than active cooling systems.
Mid concentration systems frequently rely on projections that higher efficiency solar cells
will be available at a dramatically reduced cost. Advances in producing high efficiency cells make
this assumption plausible, but there are obstacles that must be overcome before these cells are
available at a cost that would be feasible at this concentration level.
1.3.2.3 High Concentration Photovoltaics (HCPV)
High concentration systems use much more complicated optics to achieve concentration
ratios between 150 and 1,000. Dramatically reducing the amount of PV material allows these
systems to use the highest efficiency chips available and not have the chip be the driving cost of
the system. These modules have to be mounted on precision solar trackers to precisely maintain
alignment to the sun for the system to produce significant power. These systems can harvest less
than 5% of diffuse sunlight, and are thus inefficient when the direct solar radiation is blocked.
High concentration systems provide a completely different cost structure to currently
available flat plate technologies, and the development of a new high concentration optical system
is the focus of this thesis.
1.3.3 Critical Components of HCPV Systems
The research presented here is focused around a concentrator design aimed for use in
high concentration photovoltaic (HCPV) systems. These are inevitably complex systems with a
large array of components working together to produce power. This research focuses on the
concentrator system. In order to justify performance and cost goals, information about the cost
19
and specifications of the other components of the system is necessary. Active research continues
in several areas aimed at both improving performance and lowering cost of other components of
the HCPV system, and these dramatically affect the cost competitiveness of the final system
incorporating the proposed light guide concentrator. A brief overview of the system components
is described below in order to inform the performance goals of the concentrator.
Figure 1-8: A schematic of a generic concentrating system. Assumptions are made about the other components to inform targets for the concentrator that is developed in this thesis (component 3).
1.3.3.1 Land for System Deployment
By targeting the utility market, the location for these systems should be inexpensive and
have as much direct sunlight as possible. An ideal location for these systems is the desert in the
Southwest United States. An ideal location for a solar power plant is on extremely inexpensive
land with high direct solar incidence that is near a population center to minimize transmission
losses and keep component transportation costs down. States such as Nevada and Arizona are
actively seeking solar projects to provide clean power, and these represent ideal locations for
constructing a solar plant.
20
The annual cost to rent a parcel of land for a solar power plant varies substantially by
location. According to the Bureau of Land Management, the annual cost to rent land in
designated “solar areas” ranges between $15 and $315 per acre per year. This represents a
negligible cost compared to the rest of the installation, as this is equivalent to between 0.3¢ and
7.8¢ per square meter, while the cost of the system is orders of magnitude greater.15 The final
location of a solar power plant is more heavily influenced by legislative policy decisions and tax
incentives.
1.3.3.2 High Precision Solar Tracker
High concentration photovoltaics must be mounted on a precision solar tracker in order to
maintain alignment with the sun over the course of time. When working with concentrations of
several hundred suns, the system is limited to a two axis tracker. The two critical specifications
that the tracker will place on the concentrator are the pointing error that the concentrator must
be able to compensate for and the weight that the tracker is capable of bearing while maintaining
its alignment.
The accuracy of commercially available solar trackers varies based on the manufacturer
and the location of installation. Wind loading causes substantial pointing errors, and thus any
concentrator mounted on these trackers must have an acceptance angle that can compensate for
this misalignment. While the accuracy of several trackers is quoted to be 0.3° or less, this is
frequently just the drive accuracy and does not take into account misalignments from wind or
other factors. From a study conducted by Green Mountain Solar, the measured pointing error on
current commercially available solar trackers is determined to be approximately 1.2°. Ongoing
research has produced prototypes that have improved on this substantially16, but pricing
21
information and longer term performance require further testing. In order to accommodate a
reasonable error in tracking systems, the light guide concentrators are designed to have an
acceptance angle of ±1.0°.
The weight of HCPV systems that are currently deployed greatly exceeds that of the
proposed concentrator systems, and this is not a limiting factor for deployment within currently
available trackers. The light guide concentrator is much thinner and lighter than current HCPV
systems and therefore has the potential to be deployed with alternative tracker systems that
might substantially reduce cost.
The tracker system is currently a driving cost in HCPV systems which costs approximately
$150/m2. Current trackers are large and bulky systems relying on mechanical gears and crowns
that are expensive and difficult to maintain. Research is ongoing into simpler, cheaper, and more
reliable tracking methods, and this cost is expected to fall substantially in the near future. In order
to reduce this cost, an ideal optical system will have a reasonable acceptance angle to
accommodate small misalignments, be light enough to lower the mechanical requirements of the
tracker, and be scalable to allow the size of the tracking system to be an independently optimized
parameter.
1.3.3.3 Concentrating Optics
The concentrator will be the focus of this dissertation. The performance and design of the
concentrator directly affect the requirements of the other components of the system and the
performance of the module as a whole.
22
1.3.3.4 Photovoltaic Cell
A critical quality for the solar cells in an HCPV system is the getting the maximum
conversion efficiency. The high concentration causes the cell cost to not be a critical driver in most
cases, which enables use of the highest efficiency cells available at the time of production. Rapid
advances continue to be made in many different families of high efficiency solar cells, which
makes predicting the efficiency or cost of these cells imprecise. Records for solar cell efficiency
have been set and broken many times in recent years17. The most common cells that are designed
to be used with HCPV systems are III-V multijunction cells. These cells are significantly more
efficient under concentration, and thus HCPV systems have the highest potential efficiency of any
PV system currently known. Under high concentration, these cells have reached conversion
efficiencies of 44.0%18.
One issue that there is significant disagreement over is the cost of these high efficiency
cells. Current production methods for these cells involve growth on an extremely expensive
substrate that must be discarded after use. A relatively new method for reusing the substrate to
grow more than one cell called epitaxial lift-off has promise of reducing the cost of these cells.19
Manufacturing high efficiency cells is still a subject of extensive research, and thus approximating
a final cell cost per area is imprecise. Predictions range from approximately $5,000 to $50,000 per
square meter, which affects the optimal balance of system concentration and tolerances on the
system components.
23
Figure 1-9: The efficiency improvement of several different families of cells. Rapid growth, especially in the multijunction cells, is continuing. Image reproduced with permission of copyright holder National Renewable Energy
Laboratories(NREL). Image available at http://www.nrel.gov
Multijunction cells are stacks of multiple p-n junctions with different materials designed to
convert different bands of the solar spectrum. These junctions convert a portion of the solar
spectrum efficiently into electricity while being transparent to the rest of the spectrum, which
passes through to the junction where it is most efficiently be converted. This allows a broad
spectrum to be converted more efficiently than for any single junction device, which yields the
high efficiency of these cells. One potential drawback for these cells is the added design
requirement that the cells be optimized to convert the actual spectrum delivered by the
concentrator. Vertical junction stacks with no internal contacts require the current at each
junction be matched, or the junction producing the least current limit the cell as a whole.
24
1.3.3.5 Back End Electronics and Storage
Photovoltaics produce a direct current and only do so when exposed to sunlight. In order
to be used by the end consumer, this must be converted into an alternating current of the proper
frequency, which requires the use of an inverter. A larger challenge for solar energy remains the
intermittency of energy generation. Solar energy must be stored for use on demand in order to be
a viable alternative to fossil fuels which can be burned to meet demand. This system must also be
incorporated into the grid, which represents an additional cost, though this is consistent for all
power generation methods. The approximate component cost breakdown of a solar installation is
shown in Table 1-2.
Table 1-2: Standard cost breakdown of a solar installation.
Photovoltaics produce a direct current while the electrical grid and most power
applications require an alternating current. To convert a direct current into an alternating current
of the appropriate frequency requires the use of a solar inverter, also referred to as an electric
drive. These come primarily in two forms: a micro-inverter and a string/central inverter. Micro-
inverters convert power from an individual panel and made up over 98% of inverter revenues in
201020. These systems dominate the current market because they can avoid high voltage wiring,
reducing line losses. Micro-inverters are much more expensive than string inverters,21 and also
have multiple points of failure whereas a string inverter is consolidated into a single system. String
Component Cost %
Module 40-60
Inverter 10-18
Battery 15
Charge Controller 10
25
inverters are much more cost effective for a utility scale system, and thus continuing development
of these systems is critical for utility scale solar power production.
Solar panels only produce power when the sun is out, and therefore this power must be
stored for use on demand in order to be a commercially viable replacement for fossil fuels. There
are a variety of potential storage mechanisms for PV systems, but one of the most rapidly
emerging and promising technologies is Lithium ion batteries. Extensive research and
development is being done on this technology for use in consumer electronics, as these batteries
have an extremely high energy density and can be recharged many times without being depleted.
These systems face substantial competition from more mature technologies such as Pb batteries,
but market penetration is expected to improve rapidly over the next few years as Lithium ion
technology becomes more readily available in utility scale capacities.22
While rapid advances in battery technology provide opportunities for less expensive storage
for solar power generation, this makes cost projections quite variable. Currently, no high capacity
(>1MW) Lithium ion battery system is available, yet projections of these systems have the cost
ranging from $350 to $600 per kW for utility scale systems23, which increases the cost of an
intermittent solar plant providing consistent power.
1.3.4 Current HCPV concentrators
HCPV systems currently make up a very small portion of solar power technologies. Flat
plate silicon or other thin film technologies are much more mature and have reached a stage
where large deployments are being made. Some HCPV systems have gone through construction of
either functional single units or small pilot plants, with a few utility scale projects under
construction that are set to come online in the next decade.24
26
The optical designs for these systems are dominated by Fresnel lens based concentrators.
These systems employ a large Fresnel lens as a primary element, which then focuses incident
sunlight onto a secondary optical element, which increases the concentration slightly and
homogenizes the light over the PV cell. Despite a relatively simple optical design, these systems
can reach geometric concentration ratios of nearly 1,000x while maintaining optical efficiencies of
greater than 70%. The relatively simple optical design allows manufacturers to take advantage of
previous experience fabricating Fresnel lenses, which has accelerated this design to a later stage of
deployment than potentially competitive systems.
The design of the Fresnel lens primary element and the secondary optical element vary
from system to system, but there are several factors common to these Fresnel systems that limit
potential performance. The optical efficiency of these systems is limited by two primary loss
mechanisms inherent to the design, which are Fresnel losses and “tooth” losses caused by
manufacturing limitations related to the Fresnel lens primary. The Fresnel losses resulting from
light entering and then exiting the lens are added to the Fresnel loss entering the secondary
concentrator for a combined loss of approximately 15%. Antireflection coatings can reduce this,
but broad band coatings provide a substantial extra expense to the system and are potentially
damaged by environmental factors.
The other limiting factor of Fresnel system’s optical efficiency arises from manufacturing
tolerances. Fresnel lenses are made by attaching a structured polymer (either thermoplastic or
silicone) to a glass substrate. Two imperfections are inherent in such a system: a rounding of
sharp corners and a draft angle on surfaces that are ideally vertical, as shown in Figure 1-10. Light
that hits a rounded feature or a draft wall is not coupled to the secondary concentrator and thus is
27
lost. The magnitude of these losses is be dependent on the designed tooth geometry and
manufacturing process, but is typically between 7% and 12%.
Figure 1-10: The ideal Fresnel lens (left) will suffer from a minimum draft angle (center) and rounding of the ideally sharp corners (right) which will substantially reduce optical efficiency.
While the Fresnel lens based concentrators account for the majority of currently installed
HCPV systems, other designs rely on an obscured reflector to focus incident light onto the receiver.
These can take the form of a large dish or more complicated designs involving multiple reflective
surfaces. An example design for high concentration reflective systems is shown in Figure 1-11.
These systems can achieve high geometric concentrations well over 1000x. In addition to the
obscuration caused by the receiver, most of these systems put a tremendous strain on the tracking
system, as the heavy receiver presents a heavy load that is difficult to balance. These systems are
also frequently too large to hermetically seal, and thus maintenance costs are higher than for
alternative systems.
28
Figure 1-11: A large dish concentrator. This was a test module in Phoenix, AZ that is approximately 75 feet in diameter. Image reproduced with permission of copyright holder Southwest Solar Technologies
25
1.4 Light Guide Concentrators
Light guide concentrators are designed to trap light within a solid medium where it is
coupled to the output of the system. These systems usually trap light within a relatively thin layer
using total internal reflection at the top and bottom boundaries to contain light within the guide.
There are two critical qualities that differentiate the various families of light guide concentrator.
First, the injection method is how the specific concentrator traps the light within the guide layer.
The second critical characteristic is how the light trapped within the guide is coupled to the
desired output.
1.4.1 Alternative Light Guide Concentrators
While there are many alternative light guide concepts such as fluorescent26 and dye
sensitized concentrators, microoptic concentrators use precisely engineered optical features to
deflect light into the guide substrate. These systems typically take advantage of an array of
29
primary microoptics to focus incident light onto a small injection feature, which then couples the
light into the guide material where it is trapped.
An example of such a microoptic concentrator with a dramatically different design than
the systems detailed in this research is Morgan Solar’s Sun Sumba27 concentrator. This light guide
is rotationally symmetric and is designed to couple light to the center of the part where a
secondary concentrator further concentrates the light onto the solar cell. The thickness of the
light guide increases near the center, which allows light to more efficiently couple towards the
center without interacting with injection faces closer to the center. Figure 1-12 shows a schematic
for the Sun Simba concentrator.
30
Figure 1-12: A schematic for the Morgan Solar rotationally symmetric concentrator system.
A light guide system closely related to the systems developed at University of Rochester
has been investigated at the University of California, San Diego28. This system uses a lenslet array
to focus incident light into small injection patches. These patches have small extruded triangular
prisms which then deflect light toward two edges of the guide layer as shown in Figure 1-13.
These systems have been successfully prototyped and have achieved a geometric concentration of
40x with an optical efficiency of 32.4%. These systems rely on the injection elements being small
and having a small vertical extent, which allows them to shadow less of the light injected from
31
facets upstream. Light contained within the guide that interacts with another injection feature
downstream with be coupled out of the guide.
Figure 1-13: A schematic of the UCSD light guide concentrator. Light is focused onto coupling prisms that redirect light to either side of the light guide. Image reproduced with permission of copyright holder The Optical Society.
1.4.2 Dimpled Light Guide Concentrators
The dimple light guide concept was originated at University of Rochester. Initial designing,
modeling, and prototyping were done previously, and were detailed in the PhD thesis of Blair
Unger. These were layered systems, with a lenslet layer that focused incident light onto a set of
injection prisms on the other side of the concentrator. The lenslet layer was made up of either
refractive or reflective elements. These injection prisms redirected light approximately 90° into
the guiding layer. A low index layer separated the lenslet layer from the guide layer, which
allowed light that had been injected to avoid successive interactions with the lenslet array. A
schematic of the dimpled light guide concentrator is shown in Figure 1-14.
32
Figure 1-14: Schematics for a refractive and reflective dimpled light guide system.
The injection element in this system is an air prism that reflects incident light via total
internal reflection. As light propagated down the guide layer, further interactions with the
injection prisms would cause light to couple out of the guide layer and be lost. To prevent this, a
bypass prism is combined with the injection prism to redirect light away from the facet. Light
hitting the bypass prism will continue propagating in the guide, though at a slightly increased angle
relative to the guide axis. A schematic of an injection-bypass prism combination (referred to as a
dimple) is shown in Figure 1-15.
33
Figure 1-15: A schematic of a dimple for use in a dimpled light guide. The injection prism is shielded from light injected upstream by a bypass prism.
The first dimpled light guide prototypes used employed a conic lens array with hexagonal
aperture and the “knife edge” dimple design shown below in Figure 1-16. The manufacturing
process used was not able to accurately recreate this geometry, as both the vertical side walls of
the bypass prism and the sharp tail were not fabricated accurately. These defects were modeled
to determine the relative sensitivity of each manufacturing defect, and it was found that the
rounding of the bypass prism tip and rough vertical side walls had the greatest detrimental effect
of the dimple defects.
34
Figure 1-16: When fabricated, the knife edge prototype had a rounded tail, substantial drafts on the side walls, which were also rough. This substantially degraded performance.
The second generation of dimpled light guide prototypes was designed to avoid the
defects that most dramatically reduced the efficiency of the first generation prototypes. The
steep side walls were removed, and the whole bypass prism was replaced by a rounded structure
that smoothly blended all sharp corners. The second generation prototype, along with the
designed dimples and the dimple structures that were actually fabricated are shown in Figure 1-17.
35
Figure 1-17: The second generation light guide prototype (top), the designed dimple geometry (bottom left), and an SEM image of the fabricated second generation dimple structure (bottom right).
The second generation of light guide prototypes was designed to be 50mm x 60mm with a
1mm thick glass guide layer and a 100µm dimple layer. The optical efficiency of the prototype was
measured to be 72% at a geometric concentration of 55x. The acceptance angle of the
concentrator prototype was measured to be approximately ±1.0°. This system provided a baseline
for the dimpled light guide concentrators described in the remainder of this thesis.
This first generation of light guide concentrators was intended to provide a proof of
concept. These systems validated the design methods and demonstrated the capability of the
36
modeling software for assessing performance errors. While the cost of high efficiency cells used
with CPV systems is expected to decline, these cells currently cost approximately $50,000/m2.
These concentrators were only designed to be approximately 55x geometric concentration, and
with current cell costs, this low a concentration would prevent these systems from ever being
commercially viable. The focus of this thesis is to redesign the dimpled light guide concept into a
system with concentrations of several hundred, which would allow these concentrators to
compete economically.
37
2 Modeling and Design
2.1 Overview
Efficient development of new concentrator systems requires accurate models to guide
fabrication efforts. Dimpled light guide concentrators have complex microstructured features and
complicated light paths through multiple layers of material. Modeling these complicated
structures requires the use of multiple software packages and the use of key simplifications to
make optimization feasible. In addition to predicting the optical performance of concentrators,
the lifetime of these devices and performance degradation when deployed in the field are
investigated. These devices are designed for deployment outside in a desert setting, and they
must be able to survive thermal cycling, high solar flux, and environmental conditions associated
with decades of deployment in such a setting.
The design process was used determine which concentrators were most viable to fabricate,
but the models were also used to improve manufacturing processes. Initial prototypes were
fabricated, and then characterized as discussed in Chapter 3. Prototype features were measured
and characterized, and then they used to update the model parameters. When the concentrator
prototype was sufficiently characterized, the measured optical performance and the modeled
performance mirrored each other. The model could then be used to determine which errors or
effects were the driving the performance and guide future manufacturing improvements.
2.1.1 Modeling Software
Design of the next generation of light guide concentrators relied heavily on computer
simulations to guide the development from concept through prototype fabrication. The primary
38
software packages used in the design and performance modeling of these light guide
concentrators were SolidWorks® (Dassault Systèmes SolidWorks Corp.) and LightTools® (Synposys
Inc.). These two programs were dynamically linked, which allowed for optical simulations of
systems involving complicated feature geometries. These simulations were able to model the
effects of a wide variety of design options, manufacturing limitations, and system conditions to
refine system concepts toward fabrication.
SolidWorks® is a 3D CAD design and analysis software package that was used to model the
complicated dimple and lenslet geometries required by the light guide systems. Solidworks®
enables the modeling of the complicated designed features and the inclusion of expected
manufacturing defects resulting from the high volume production processes these systems are
designed to cater to. This provides critical information about the sensitivity of the various designs
to manufacturing defects or errors. SolidWorks® also has a Finite Element Analysis (FEA)
simulation package that allows for analysis of both mechanical and thermal strains on the system.
The FEA capability is used to predict thermal and mechanical stability of light guide concentrator
systems when deployed in the field.
LightTools® is a non-sequential ray tracing program that is used to simulate the optical
performance of the light guide concentrators. The dimple and lenslet features are linked from
SolidWorks®, which allowed for optimization of these features within LightTools®. LightTools®
provides an efficient means to trace a large number of rays through complicated systems, and this
allows precise optimization of many parameters that have a complex interdependence. In
addition to accounting for the complicated geometries demanded by these systems, LightTools®
39
can model a wide variety of materials, optical coatings, and provide an approximation of the yearly
energy output of a solar system.
2.1.2 Modeling Methods
These systems were generally modeled by patterning features designed in Solidworks®
using LightTools’® 3D texturing feature. This allowed any feature that was modeled in
SolidWorks® to be patterned in an appropriate geometry, as shown in Figure 2-1. The array of
microfeatures could then be used in ray tracing simulations, and the optical performance could be
modeled. Several of the systems involved dimple features that blended together, and these could
be designed in the same fashion, though extra care must be taken to ensure that the region
joining neighboring features had an appropriate boundary.
Figure 2-1: A lenslet feature is created in Solidworks and then patterned appropriately using LightTools.
Another common modeling practice for these light guide concentrators is to use a thin
section, such that the dimension perpendicular to both the guiding axis and the incident light is
much smaller than would be practical for an actual system. This is simply to reduce the
computation time required for the ray trace, and has been observed to have no noticeable effect
40
on the performance as long as an entire repeating section is used. Rays propagating down the
guide that are incident upon these edges are reflected back into the guide as long as they are
below the critical angle. While rays incident with an angle greater than the critical angle are
unduly ejected from the system, these rays generally do not propagate far enough in the actual
guiding system to account for a significant discrepancy.
2.1.3 General System Parameters
The light guide systems are designed for deployment, and thus several common parameters
are assumed throughout the design process. All high concentration systems require the use of a
solar tracker, and the pointing accuracy of these devices determines the acceptance angle that is
required for the mounted concentrator. There are a variety of precision solar trackers available,
and more advanced systems are being developed. The maximum pointing error can be as small as
half a degree, but in most systems that consider wind loading it is substantially larger than that.
Increasing the acceptance angle decreases the concentration that the system can achieve, so a
tradeoff must be made between concentration and acceptance angle. The dimpled light guide
concentrators are designed to have acceptance angles of ±1.0°, which is achieved by a range of
tracker systems and designed to tolerate some manufacturing error in the concentrator as well as
in the deployed system. In a 2008 study of tracker accuracies, Green Mountain Engineering
measured the fraction of captured energy as a function of acceptance angle. The dependence of
tracker pointing on wind speed can be seen in Figure 2-2.29
41
Figure 2-2: Tracker Accuracy is dependent on wind loading (left) and the acceptance angle determines the fraction of energy available (right). Image reproduced with permission of copyright holder Green Mountain Engineering.
The source used in the ray tracing simulations is designed to mimic the sun both in angular
extend and spectral composition. The sun has an angular extent of ±0.26° as measured from the
surface of the earth. While the systems are designed to optimize performance over the range of
±1.0° to account for manufacturing defects and tracker pointing errors, the optical efficiencies
reported correspond to a ±0.26° source. The spectral power density of the source is critical to the
performance of the PV material the concentrator is coupling the light onto. The intensity and
spectral composition of the sunlight striking the surface of the earth varies according to a wide
array of factors including location, time of day, season, atmospheric conditions, and solar activity
variations. ASTM G173-0330 consists of two standard spectra. The first spectrum shows the direct
incident solar radiation at the average latitude of the contiguous United States at peak irradiance
under normal atmospheric conditions, and the second spectrum shows the radiation adds in the
diffuse radiation contribution to provide the total spectrum incident under optimal conditions at
37° latitude. The ASTM G173-03 spectra are shown in Figure 2-3 The direct spectrum is used as
42
the incident spectrum for design and modeling, as CPV systems cannot capture the diffuse
component.
Figure 2-3: ASTM G173-03 Direct Component. This is a standard spectrum representative of peak irradiance in the United States
The solar spectrum spans a broad band of wavelengths, but not all of the energy that
reaches the earth can be converted into electricity by photovoltaic materials. For characterizing
the optical efficiency, the spectrum is usually truncated at an energy corresponding to the lowest
bandgap material used in the solar cell. For simple silicon cells, this corresponds to a wavelength
of approximately 1100 nm, while for multijunction cells that commonly use germanium as bottom
cell, this can extend to 1770 nm. During the design and characterization of the light guide
concentrators, for this thesis the spectrum is truncated at 1856 nm, which is in the center of an
atmospheric absorption band. The spectrum is truncated on the low wavelength end at 280 nm,
43
as there is a negligible amount of solar radiation incident at wavelengths shorter than this, and
this high energy radiation is frequently damaging to system components.
2.2 Concentrator Optical Design and Performance Modeling
The previous generation of light guide concentrators was used as an initial starting point
for judging performance, and provided a substantial amount of information from which the
system design were altered to improve optical performance. Three novel families of light guide
concentrator were investigated using the knowledge gained from previous generations. The
design of each system used a set of materials and manufacturing limitations that were expected to
be compatible with high volume processes. In addition to the designed optical performance, the
sensitivity to various manufacturing and deployment defects was presented. The presented cases
did not cover all the potential variations of these concentrator families, but were designed to
demonstrate an ideal performance target while maintaining realistic manufacturing cost goals.
2.2.1 Phase Space Performance metric
The primary goal of modeling and design work is to achieve as high a geometric
concentration and optical efficiency while maintaining the ability to manufacture these systems in
a cost effective manner. One of the most useful tools for judging the performance of a
concentrator during the design process is to gauge the performance of a concentrator using
étendue, specifically the simplified phase space. The phase space volume is defined in section
1.3.1. The input to an optical system gives a volume in phase space, and in a lossless system, this
cannot be compressed. This can be expanded or diluted, but cannot be reduced without losing
light. A 4 dimensional space is difficult to use as a design tool, and thus some simplifications and
approximations are used.
44
For light guide systems, the cross-sectional area of the guide is approximately constant as
light travels down the guide. While the tail features of the guide shrink the guiding area slightly in
order to shield a downstream injection facet, this is taken to be a small effect. The angular
spectrum of the guide is also taken to be approximately independent of cross sectional position.
This does not take into account the sections close to an injection facet where a cone of injected
light has not dispersed throughout the cross sectional area. This injected light forms a small
portion of the flux through any surface except the facets farthest from the chip. The phase space
is thus reduced to the two angular components.
The angular spectrum propagating in a light guide system provides a substantial amount of
information on how effectively the system is concentrating light from an étendue standpoint.
How the system is filling the target angular space provides insight on how to improve concentrator
performance and how much further concentration can be done. The extent of the angular
spectrum dictates what loss is necessary for a certain increase in geometric concentration. The
goal for a light guide concentrator is to couple the input angular spectrum (the effective injection
angular spectrum) and couple it to the angular spectrum that can be contained within the guide at
the output face of the concentrator.
The target angular spectrum that can be contained by the guide is determined mostly by the
materials chosen, though the dimple geometry has a small effect. The light is contained by the
interface with the low index layer on one or potentially 2 sides and an air interface on the
remaining sides. The target angular spectrum is thus rectangular with a larger containment angle
in the plane of the guide. The use of the angular spectrum as a diagnostic tool is shown by looking
at the angular spectrum propagating through a dimpled light guide assembled previously and
45
described in detail elsewhere31. The angular spectrum is shown at several different distances from
the injection element in Figure 2-4.
The angular spectrum provides a useful tool for assessing the performance of a light guide
concentrator. The asymmetry of the target angular area can be seen in Figure 2-4, as the vertical
containment comes from the interface with the low index layer. Whether the bottom of the
dimple structure is an air interface or another low index interface, repeated reflections from the
top and bottom surface quickly mirror incident light, and thus the boundary is effectively
symmetric about the guide plane. The asymmetry of containment shows that it is more effective
to spread light in the guide plane (laterally) than vertically if the two can be isolated.
46
Figure 2-4: The angular spectrum a) 1mm, b) 18mm, c) 36mm, and d) 52mm from the injection element. The angular spectrum 1mm from the guide is that injected by the lenslet. The angular spectrum expands as light travels further down the guide, and any light travelling with too large an angle (outside the blue rectangle) is lost from the system.
47
Close to the injection element, the angular spectrum almost exactly mirrors the spectrum
input by the lenslets (only adding facet scattering and defects). In Figure 2-4a, the hexagonal
lenslet aperture is clearly visible. The injection element was not designed to rotate the light 90°,
which can be seen by the angular spectrum being asymmetric about the guide plane before it has
reflected off the top surface. There is also substantial spatial dependence before the injected light
has uniformly spread throughout the guide. In Figure 2-4b, the injected light has spread to fill the
light guide fairly uniformly. The angular spectrum is symmetric about the origin, and several
distinct patches can be seen. Every interaction with the tail of a dimple feature spreads the
angular spectrum by a discreet amount, which leads to these distinct regions. As light travels
down, more of the light is spread into these discreet regions as successive dimples are
encountered. The injected light does not evenly fill the space, as light traveling with a higher
lateral angle will be exposed to a larger cross section of the tail, increasing interaction probability.
It can be seen in Figure 2-4c and d that the angular spectrum has expanded to fill the entire target
area. This is the limit of the concentration that can be achieved without loss, and at these
distances, substantial guiding loss has already occurred.
The next generation of concentrators was designed to improve on some of these
deficiencies. It was critical to evenly fill the target angular area, and to minimize the dilution of
high angles. The dilution of angular space at higher angles was inherent to the dimpled light guide
concept. Higher vertical angles yielded more frequent interactions with the dimple layer and thus
faster increase in angle. Propagating with a high lateral angle increased the effective cross section
of dimples, making interaction with a dimple more likely when light traveled through the dimple
layer.
48
2.2.2 Lenslet Design
First generation light guide prototypes used conic lenslet arrays with flush hexagonal
apertures.32 This aperture shape was used because a hexagon produced the most nearly circular
angular spread while still meeting the requirement that the aperture shape tessellate. The lens
array had no designed discontinuities to make fabrication less challenging and reduce cost. The
next generation of light guide concentrators made use of offset lens arrays which had designed
discontinuities at the edge of the apertures. Offsetting the lenslet aperture substantially
decreased the effective injection numerical aperture, and thus allowed superior guiding while
maintaining high injection efficiency. Optimizing the aperture shape of the lenslets also allowed
manufacturing limitations to be considered and further optimized performance.
The ideal injection system produces an injected ray bundle that centered around the guiding
axis, which requires a 90° deviation of the input light. The exact shape of the ray bundle depends
on the dimple geometry and other system parameters. In light guide systems, light reflects many
times off the top of the guide layer and bottom of the dimple layer. These frequent reflections
cause the angular spectrum to be vertically symmetric, and thus effective vertical angular extent
injected into the guide is determined by the largest angular deviation from the guiding axis.
Having the injected ray bundle centered around a vertical angle of 0° thus produces the minimum
effective injected angular extent, and thus allows the greatest concentration.
While 90° rotation of incident sunlight is optimal for guiding efficiency, this rotation is
dependent on total internal reflection at the facet. The critical angle of a material with n=1.5 at an
air interface is approximately 42°. If the center of the ray bundle is incident at 45° in order to
reflect along the guiding axis, the incident ray bundle can only span 3° without injection losses,
49
which corresponds to an NA of .079. This small lenslet aperture makes a 45° injection facet
impractical. Offsetting the lenslet so that the center of the aperture does not correspond to the
optical axis allows the center of the ray bundle to not be normal to the guide plane. This offset
allows a significant injection numerical aperture while still maintaining TIR for the limiting rays and
injecting the center of the ray bundle in the plane of the guide. Figure 2-5 shows the light incident
on the offset lens array (red rays) and highlights the ray at the center of the aperture (cyan ray)
and the ray incident on the injection facet with the greatest angle which limits TIR (green ray).
Figure 2-5: Offsetting the lenslet aperture causes the central ray to be deviated from normal to the plane (Left). If the central ray is not deflected to travel in the plane of the guide the effective injection angular spread will be much
larger (Upper Right) than if it is deflected directly parallel to the plane of the guide (Lower Right)
50
In addition to offsetting the aperture to minimize the effective injected angular spread,
the shape of the aperture can be altered to further optimize concentrator performance.
Customizing the aperture shape is especially useful for accounting for manufacturing limitations.
The shape of the lenslet aperture determines the shape of the injected ray bundle in angular space.
The various dimple geometries all transform the injected ray bundle differently as it propagates
down the guide, so the optimal shape of the lenslet varies based on the dimple geometry.
Customized aperture shapes also allow manufacturing limitations to be taken into account. As an
example, the offset lenses tend to have large surface sag, which is the displacement along the
optical axis of the surface away from the apex. Several processes are limited in the depth they can
accurately produce optical quality surfaces. This restriction combined with the requirement that
the aperture be able to tessellate leads to an unusual optimal lenslet geometry shown in Figure
2-6.
Figure 2-6: A lenslet geometry designed to cater to a manufacturing process where the sag depth is restricted. Allowing customized aperture shapes allows improved injection and guiding efficiencies.
51
While the aperture geometry varies based on a large set of parameters, the lenslet surface
shape only depends on the guide thickness, material choice, and manufacturing capabilities. In
general, the systems performs best when the lenslet focuses to the smallest possible spot on the
injection facet. The ideal surface shape is a conic focusing to a small spot on the injection facet, as
seen in Figure 2-7. A spherical lenslet produces a large amount of spherical aberration which
results rays from the edge of the aperture having up to 250um error in the focal plane. This
causes both lost light on axis and reduced acceptance angle. Spherical lenslets cause an especially
large performance drop when combined with offset lens apertures due to spherical aberration.
Figure 2-7: Optimal optical performance requires a conic surface shape for the lenslets to achieve a tight focal spot (left). A spherical lenslet surface will leave a large amount of spherical aberration and reduce performance (right).
2.2.3 Dimple Tree Performance and Sensitivity
The “dimple tree” concentrator family is the most directly related to the earlier generation
of concentrators. One performance limitation of the “knife edge” dimple that the first dimpled
light guide concentrators employed is the rounding of the tail that is designed to be a sharp edge. .
This is overcome by blending the tails together so that there is no such sharp angle. Each row of
52
dimples is combined into a set of long strips as shown Figure 2-8: The first generation "knife edge"
dimple has been fused into a long strip to avoid the problems associated with manufacturing the
knife edge geometry.Figure 2-8.
Figure 2-8: The first generation "knife edge" dimple has been fused into a long strip to avoid the problems associated with manufacturing the knife edge geometry.
Like the first generation dimples, the system was designed to increase the angular extent
only within the plane of the guide. This system was designed to increase the vertical angular
extent as little as possible, and thus sidewall drafts were avoided if possible. To avoid the extra
draft face that arises from the side with no injection facets shown in Figure 2-8, the strip is made
wider, and the adjacent row of injection facets is attached to the side of the strip opposite the first
set of facets. The injection facets were also rotated about an axis normal to the guide to ensure a
minimum lateral angular spread was injected. The amount of rotation was determined by the
lenslet geometry to compensate for the initial interaction with the wedged side wall that
approximately half of the injected light interacted with almost immediately. These alterations
yielded the tree like shape shown in Figure 2-9.
53
Figure 2-9: The dimple geometry of the "dimple tree" family of light guide concentrators. The injection facets are highlighted in teal.
This dimple geometry is particularly well suited to a reflective design, as the mostly flat
geometry causes relatively little shadowing. In a reflective design, light going through the flat
region of the dimple structure is still focused onto the injection prism face. The only components
that cause shadowing losses are the actual injection faces, draft regions, and fillets. The
shadowing losses are given by
where X, Y, and Z represent the corresponding dimensions of either the injection facet or lenslet,
θd is the draft angle on the side walls, and rF is the fillet radius. For a system with draft angles of 2°,
a fillet radius of 2μm, 100μm injection facets, and 1 mm lenslets, this corresponds to a shadowing
loss of approximately 2.5%. This is substantially less shadowing than for other dimple geometries,
and therefore this system is a much better candidate for a reflective system.
Z
Y X
54
The model of the reflective system used a basic silver coating for the mirrors. This was
primarily for cost reasons, as this was a readily available coating that is common for high volume
manufacturing purposes and still had high reflectivity. This coating had a reflection loss of
approximately 5%, and when combined with the Fresnel losses entering the guide and the
shadowing loss, the reflective systems had relatively low injection efficiency. The injection
efficiency of these systems was expected to be lower than 90%. This could be improved through
adding an anti-reflection coating to the front surface or using a more reflective coating, such as
enhanced silver. These would improve performance, and their use was strictly a cost tradeoff. An
uncoated front surface and normal silver coating were projected to be the most viable in terms of
cost effectiveness and formed the baseline for the reflective models. The basic design form of the
dimple tree geometry is consistent, but the specific parameters will vary based on the
manufacturing process, materials used, and the concentration required by the system. For each
set of materials and manufacturing limitations, the specific design parameters are reoptimized to
maximize performance. The optimal design assumes a manufacturing process capable of writing
features with 1 µm fillet radii and 1° draft angles. These are characteristic of high end
manufacturing capabilities. For a material choice, the guide layer and the dimple layer are
assumed to be made of a high refractive index, ultra-transmissive glass, NBAK-1. This is a
commercially available glass, but is dense, expensive, and likely not compatible with most
manufacturing processes capable of achieving the previously mentioned feature sizes. A reflector
array with offset rectangular apertures is assumed.
The ideal design is capable of achieving geometric concentrations well over 1000x with
optical efficiencies of above 70%. This design may not be viable for manufacturing, and is
55
extremely unlikely to be commercially viable due to the high cost and density of the glass.
Relaxing some of the design parameters to use less expensive materials and easing the
manufacturing tolerances reduces the concentration that can be achieved (while maintaining a
reasonable optical efficiency, in this case 70%). The number of possible permutations is too large
to model each individually. A sample of the range of performances that can be expected is given
below in Table 2-1, showing the range between the optimal design discussed above and a less
expensive design catering to a much less precise manufacturing process.
Table 2-1: Development of reflective dimple tree light guide variants showing an optimal design and then tracking various concessions for manufacturing purposes. The results of the earlier Generation 2.5 light guide prototypes are
also shown for comparison purposes.33
This represents a small subset of light guide designs, all of which were designed to
maintain optical efficiency of at least 70%. The field of view, defined as the angle where the
optical efficiency reaches 90% of the peak value, was kept to be ±1.0°. There is a tradeoff
56
between the geometric concentration and the optical efficiency determined by the length of the
guide. The geometric concentration increases as the guide gets longer, yet guiding losses and
absorption losses increase and reduce the optical efficiency. For a few of the dimple tree light
guide structures, the tradeoff between optical efficiency and geometric concentration is shown in
Figure 2-10. Changing the material to a lower index glass for the guide layer primarily causes an
injection efficiency decrease, while relaxing the manufacturing tolerances primarily affects the
guiding efficiency.
Figure 2-10: The efficiency fall off for various dimple tree light guide concentrators as the geometric concentration increases.
This family of dimple geometry is designed to spread light laterally in the plane of the
guide. In a system without drafts and fillets, the angular extent in the vertical direction does not
increase as light propagates down the guide. The angular spectrum that light guide concentrators
can contain is approximately rectangular with lateral containment being larger. The air-guide layer
interface has a higher index contrast than the guide layer-low index layer. The angular expansion
57
is still exponential, as the rays propagating with higher lateral angles interact more frequently with
the wedged dimple side walls. Having a small spread in vertical angle ensures fewer interactions
with the dimple layer and improves the concentration that a secondary concentrator at the output
can achieve.
58
Figure 2-11: The angular spread of a reflective dimple tree light guide with nearly ideal manufacturing parameters (1µm fillet radii, 1° draft angle). The angular spectrum expands almost entirely laterally, with the vertical expansion being due to the draft angles and fillets. The red rectangle shows the boundaries for TIR containment in the light guide structure.
59
2.2.4 Two Stepped Guide Performance and Sensitivity
The stepped light guide provides a different means of concentrating light while employing
similar manufacturing technologies. The other dimpled light guide concentrators that have been
previously investigated and the other families of light guides presented in this thesis concentrate
incident light by expanding the angular spectrum with a relatively consistent spatial extent. The
stepped light guide is not designed to expand the angular spectrum as light propagates, but
instead the guide expands in the spatial dimension. The most important consequence of this
difference is that the concentration at the output face of the guide does not increase as the guide
increases in length. The two stepped guide system has a concentration determined by the relative
size of the injection facet and lenslet aperture, not the guide dimensions. This system allows a
feasible chip size and a larger collection aperture while maintaining the approximately flat system
geometry.
The two-stepped light guide design makes use of two different means of patterning the
injection facets. The first of these is horizontal stepping, in which each facet is slightly offset
laterally from the upstream facet. This allows collection of the light from a series of lenslets into
the same guide. The concentration of this system is equal to the lenslet area divided by the
projected injection facet area. A limited number of these horizontal steps can be taken before the
facet would interfere with an injection facet directly upstream. The number of horizontal steps
that can be taken without interfering with a downstream facet is equal to the width of lenslet
aperture divided by the width of the injection facet. A schematic of horizontal stepping illustrating
the maximum number of steps for a few relative lenslet and facet geometries is shown in Figure
2-12.
60
Figure 2-12: A Schematic of horizontal stepping. This represents a top view with the light purple squares representing the lenslet apertures, the dark purple squares representing the injection facets, and the red rectangles representing
the chips. The relative size of the lenslets and injection facets determines both concentration and how many horizontal steps can be taken. Shown are a 36x concentrator with 6 steps (left), a 9x concentrator with 3 steps
(center) and a 4x concentrator with 2 steps (right)
The other important design geometry for this family of light guides is vertical stepping.
This employs the same concept, but instead of the facets being laterally offset from each other (X
offset), the facets are vertically offset (Y offset). This causes the guide to get thicker at each step,
and thus concentration does not increase with the length of the guide. The expanding guide layer
demands an alteration in the lenslet layer to compensate for the increased focal distance. The
lenslet layer can either become thinner by an amount corresponding to the increase thickness in
the guide layer, or the surface shape of the lenslets can change to have a longer focal length as the
guide gets thicker. These two geometries impose different limits on the number of vertical steps
that can be accommodated in a light guide, as if the lenslet layer is getting thinner, eventually the
curved surface of the lenslets intersects the low index layer. If the focal length is changing, it will
become long enough that the acceptance angle becomes inadequate. A schematic of vertical
Z
X
61
stepping in which the lenslet layer is decreasing in thickness to compensate for the thickening
guide is shown in Figure 2-13.
Figure 2-13: A schematic of vertical stepping. The guide layer increases in thickness so that each injection facet will not interfere with light injected upstream. In this schematic, the thickness of the lenslet layer is decreasing to
compensate for the increased guide layer thickness.
These two stepping geometries can be combined to form a two stepped guide that can
achieve the concentrations required for HCPV while maintaining reasonable chip geometries.
These guides employ horizontal stepping (along the X axis) to the maximum number allowed by
the lenslet and injection facet geometry. Upon reaching the maximum number of horizontal steps,
a vertical step is taken (along the Z axis). Combining these two concepts allows a concentration
equal to the area of the lenslet divided by the projected area of the injection facet. For a
reasonable set of parameters, this can be between 100x and 400x at the output face while still
maintaining the required acceptance angle and manufacturing tolerances. The angular spectrum
Z Y
62
at the end of the guide should be similar to that at the lenslet focus, and thus these systems are
ideal for use with secondary concentrators to bring the concentration to HCPV levels of 500x to
1000x. A schematic of a two stepped guide combining vertical and horizontal stepping is shown in
Figure 2-14.
Figure 2-14: A schematic of a two stepped guide. When a horizontal step would interfere with an upstream injection facet, a vertical step is taken.
The stepped light guide has the advantage of high optical efficiency. Because the angular
spectrum is not expanded as light travels down the guide, the only loss mechanisms in the guide
are the initial Fresnel losses during injection and losses inherent to propagating in the guide
material. These material losses are dominated by absorption, but may also include scattering and
inhomogeneity. The downside of this design is that the concentration is not increased as the guide
is made longer, and therefore it is limited by the relative sizes of the lenslets and injection facets.
Without a secondary concentrator, this geometry is typically limited to below 500x. To achieve a
higher concentration while maintaining the acceptance angle of the system, the numerical
aperture of the lenslet must increase, which will increase the surface sag and make fabrication
more difficult.
63
The angular spectrum of a perfect two stepped light guide is the injection spectrum from
the facet made symmetric about the guide plane. If the device has no manufacturing defects, the
angular spectrum at the output resembles the angular spectrum input into the guide. If the guide
layer has drafts on the side walls and/or filleted corners these does not expand the angular
spectrum, but will circularize it, making it more rotationally symmetric about the guiding axis. This
can lead to performance degradation, as the containment of a light guide concentrator is
rectangular in angle space. For an ideal system, a rectangular lenslet aperture is used to inject a
rectangular angular spectrum with the lateral angle having a substantially greater range than the
vertical angle. Significant draft angles or fillets prevent this angular discrimination and reduce
both the potential injection aperture and the concentration achievable by a secondary
concentrator at the output face of the guide.
While this angular circularization from fillets and draft angles is present in all systems, it is
most readily isolated in the two stepped geometry, as ideally there are no other effects on the
angular spectrum. Shallower drafts and larger fillet radii both increase the degree of
circularization for an angular spectrum, and thus the degree of angular discrimination must
consider the magnitude of these manufacturing defects. An example of an input angular spectrum
designed to have a much greater angular extent in the lateral direction than the vertical extent is
shown for an ideal two stepped guide system. Three different levels of manufacturing precision
are compared in Figure 2-15, which highlights the effect of manufacturing imperfections on the
circularization of the angular spectrum.
64
2.2.5 Axial Index Variation Performance and Sensitivity
The last family of concentrators utilizes an index variation along the axis normal to the
guide plane to improve guiding efficiency and concentration potential. The injection region is
Figure 2-15: The effect on the angular spectrum of a two stepped light guide for various levels of manufacturing defects. More precise manufacturing tolerances allow isolation of vertical and horizontal angles.
65
made to be a higher refractive index than the bulk guide layer, which causes rays travelling up
from the injection layer to be redirected more directly down the guide. This can result in some
rays injected with shallow vertical angle to be trapped within the injection layer, but with the
appropriate dimple geometry this does not result in rapid ejection.
The primary loss mechanism for an ideal dimpled light guide concentrator is through
repeated interactions with the dimpled structure. These interactions increase the propagation
angle of the light travelling down the guide, and if the propagation angle falls outside of the
containment region for the light guide, the light will be lost. In most light guides, the ray angle will
increase exponentially, as rays with larger angles interact more frequently with the dimple layer
and are closest to the angular containment boundary. Light injected from the edge of the lenslet
aperture has the greatest propagation angle and is lost from the system after a relatively short
propagation distance down the guide. This is the performance limitation on most light guide
concentrators. Systems with axial index variation along the y-axis are designed to cater to these
rays from the edge of the lenslet aperture by redirecting them more directly down the guiding axis.
This allows these rays to experience fewer dimple interactions as they propagate down the guide,
and thus increases the achievable concentration. Two configurations for achieving this redirection
are to have a mismatched dimple layer with a higher refractive index than the bulk guide and to
have a gradient index (GRIN) region with the highest index edge smoothly varying to the central
portion of the guide. A schematic for this bending is shown in Figure 2-16.
66
Figure 2-16: An axial index variation will reroute a ray injected with a steep angle to travel more directly down the guide. This can be done with a mismatched guide layer or a GRIN in the guide layer.
Ideally, the index change would be a smooth gradient with the higher index at the edge of
the system. Gradient index parts are more difficult and expensive to fabricate, and the same
principle can apply to a system that has a higher index injection layer with a Fresnel interface
between the guide layer and the dimple layer. The GRIN system avoids Fresnel reflections, which
can cause a portion of light to be trapped despite hitting the interface at an angle greater than the
critical angle. This reduces the performance, as a longer distance trapped in the dimple layer
opens a more direct loss mechanism. Repeated Fresnel reflections of rays incident at an angle
greater than the critical angle, the mismatched concentrators show guiding losses at shorter
propagation distances than GRIN concentrators.
67
Axial index variation does not represent an improvement for all design geometries. While
the rays with the steepest vertical injection angle are bent to travel more directly along the
guiding axis, rays with shallow vertical angles cannot escape the higher index injection region and
repeatedly interact with the dimple region. The axial index variation proves advantageous if these
dimple interactions impart some small added vertical angle. This causes the light trapped in the
higher index injection region to gradually increase in vertical angle until it is able to escape the
higher index containment region and propagate down the guide at a shallow angle. This occurs
over a relatively short distance, and thus the input light homogenizes in vertical angle relatively
quickly. If the bypass prisms impart a primarily horizontal angle, the trapped light quickly reaches
the horizontal containment edge. If the vertical angle imparted is too severe, the light is not
necessarily redirected at a shallow angle by the index contrast.
2.2.6 Wedged Stepped Concentrator
One system that has been modeled that works well with axial index variation is the
wedged step dimple geometry. This system uses the concept of horizontal stepping, as described
in the previous section. Instead of vertical steps making the guide thicker when the limit of
horizontal steps is reached, the facets that are directly along the guiding axis from each other are
connected by a small wedge. This system does not change thickness over its length like the two
stepped guide, and thus the concentration will increase with guide length as for other dimpled
light guide concentrators. Every time a ray hits the bottom surface, it is reflected with a small
increase in vertical angle proportional to the wedge in the bottom surface. Repeated interactions
with this wedged surface cause trapped light to slowly approach the critical angle needed to
escape the higher index region. This light eventually exceeds the critical angle by a small amount
68
and propagates down the guide at a shallow vertical angle. This occurs over a much shorter
distance than the angular accumulation of the initial high angle rays, and thus the vertical angular
spectrum isrelatively homogenous after a certain propagation distance. Each successive
interaction with the bottom surface only increases this propagation angle by a small amount,
which allows the light to travel down the guide a relatively long distance before it begins to
experience guiding losses. A model of the wedged stepped concentrator is shown below in Figure
2-17.
Figure 2-17: A model of a wedged stepped concentrator. The injection facets are joined to the facet directly downstream by a slight wedge. This concentrator geometry uses horizontal stepping.
This concentrator design allows for minimal guiding for a substantial propagation distance.
For the gradient system, an ideal concentrator (no draft angles or fillets impost by manufacturing
limitations) does not experience guiding losses until the rays injected at the edge of the aperture
reach the critical angle for containment. The angular spectrum spans only a small extent in the
69
vertical direction after a given propagation distance. All the rays that propagate a given distance
are ejected from the guide over a relatively short range. These guides thus have a fairly sharp
cutoff from having no guiding losses to complete loss over a fairly narrow propagation length
range.
The ideal GRIN concentrator system can contain the entirety of injected light for a distance
of half a meter or more. At propagation lengths this long, absorption in the guide layer becomes a
critical performance driver. For a concentrator that is 600mm long, absorption of 0.2% for 10mm
of travel causes an absorption loss of greater than 6%. This corresponds to a commercially
available “ultratransmissive” glass, and thus the materials available for such a long concentrator
are limited.
The index variation between the injection facet and the bulk guiding layers should be
strong enough to contain almost the entirety of the injected vertical spectrum. If the index
difference is too small, light propagates into the guide layer with vertical angles higher than ideal,
and the light guide loses the uniform vertical angular spectrum for a given propagation distance.
An index difference that is too high does not diminish performance for a GRIN light guide, but
increases the Fresnel reflections of a mismatched design, which degrades performance. If the
index difference is high enough, performance can be improved by increasing the injected vertical
aperture, though this may cause difficulties in the lenslet manufacturing. The index difference for
a given vertical angular extent θV should be approximately ninj[1-cos (θV)], where ninj is the index of
the injection layer. The index change that is required for a lenslet with an injected vertical NA of
0.3 is approximately 0.03. Many gradient index plastics have Δn on the order of 0.1, but is difficult,
though possible, to fabricate glass gradients with Δn of 0.03 using common methods .
70
These index varied designs rely heavily on isolating the vertical and angular components of
the contained light. Manufacturing defects such as drafts and fillets allow light to transfer some
horizontal angle into a vertical angle, and thus possibly avoid the range of vertical angles close to
the critical angle for the injection region. This guide design relies on all of the light that is injected
increasing in vertical angle slowly through interactions with the wedged bottom surface. Near the
critical angle of the high index injection region, the light picks up vertical angle slowly over a
relatively long propagation distance. These manufacturing errors provide an alternate means of
increasing in vertical angle through this crucial range. These defects prevent the containment of
all of the input light, and cause guiding loss to begin after a much shorter propagation distance.
The effects on the ideal GRIN design from both absorption and small manufacturing defects are
highlighted in Figure 2-18. This assumes a homogenous glass with a base refractive index of 1.536
with a gradient index of .03 over the outer 200µm of both the top and bottom of the guide layer (Y
extremes of guide layer).
71
Figure 2-18: The modeled optical efficiency for light injected into the guide at given distances. Overall the system was designed to be 600mm long with a geometric concentration of 750x. Absorption and manufacturing
imperfections both reduce performance.
The index variation causes light that has traveled a certain distance down the guide to
have a narrow vertical angular extent. This is mirrored about the guide plane, and thus the
angular spectrum comprises two symmetric bands with vertical angle proportional to the distance
travelled down the guide. This effect allows a more uniform filling of the angle space of the guide,
and thus a higher concentration without guiding losses. The ideal wedged stepped geometry
should not affect the lateral angular spectrum, but manufacturing defects will prevent perfect
angular isolation. This creates a loss mechanism by which light can gain vertical angle other than
repeated interaction with the wedged bottom surface, and thus the bands from the ideal design
will become more diffuse.
72
Figure 2-19: The angular spectrum for a gradient index light guide. Manufacturing errors will circularize the spectrum and reduce guiding efficiency of the device.
73
The system described above assumes a shallow axial gradient of 0.03 with the refractive
index at the edge of the guide layer then in the central region. This poses fabrication challenges,
as producing this gradient may be difficult or too costly. Both the GRIN light guide and the
mismatched index light guide perform favorably to a light guide with the entire guide layer and
dimple layer being a high index homogenous material. The performance of these three systems is
shown in Figure 2-20 with no absorption to highlight the difference between these three
configurations.
Figure 2-20: The modeled effects of both a mismatched guide layer and gradient index guide layer compared to the homogenous design
2.3 Material Durability and Lifetime Modeling
One of the most important challenges facing HCPV systems is finding materials capable of
withstanding many years of deployment in high solar resource areas. Typical areas for initial
74
deployment are in sunny areas with inexpensive land such as the desert in the southwestern
United States. The temperature in these areas can vary 25° C or more over the course of a day. In
addition to the daily thermal cycling that these systems must endure, HCPV concentrates sunlight
several hundred times. If the chip is encapsulated, which most HCPV systems are, the
encapsulating material must be able to endure several hundred suns of radiation flux. This
drastically reduces the number of materials available and will also prohibit materials with a
substantial amount of defects that lead to absorption centers that can damage the surrounding
material. Fresnel systems commonly use plastic-on-glass components in low flux regions (the
Fresnel lens), but high flux areas (the secondary concentrator) must be made of glass.
Concentrations of several hundred suns place strict requirements on any materials that
used in HCPV systems, yet the ultimate viability still requires considerations based on cost. In
addition to the material costs, the weight of the modules that are put onto the trackers affects the
cost of the installation. Glass is commonly used for regions of high flux due to excellent
transmission and durability, yet the cost and weight of glass materials are much higher than
polymer equivalents. Most HCPV systems to date employ a mixture of glass and plastic, which is
less expensive, lighter, and easier to manufacture to precise geometries. Material resistance to
environmental exposure is also critical, as air gaps must be hermetically sealed to prevent
moisture or particulate build-up degrading performance.
Light guide concentrators pose a different set of challenges for material selection. One of
the key advantages of light guide concentrators compared to Fresnel lens systems is the thin
geometry. Light guide concentrators are typically be 3-5 mm thick, while most Fresnel systems are
200mm thick or more. Light guide concentrators do not have a substantial air space, and thus do
75
not require hermetic sealing, and significantly reduce the load requirements on the tracker.
However, due to the long path length in the guiding layer, high transmission is even more of a
performance driver for light guide concentrators. The layered structure also poses potential issues
for delamination under temperature variation. The photostability and temperature stability of
light guide concentrators is modeled to provide a potential life cycle analysis and inform cost
modeling.
2.3.1 Ultraviolet and Infrared Absorption Models
Photostability is critical to the performance of HCPV systems, which must deliver as much
of the useful light to the cell as possible while not being damaged. Any material used for the
guiding layer of light guide systems must be extremely transparent in the visible, which contains
the majority of the sun’s energy. Both ultraviolet and infrared light are potential sources of
failure for long term HCPV deployments. While the near infrared can be used by many potential
cells, IR light from the sun with too long a wavelength is not useful for producing electricity. This
unwanted light can still damage a concentrating module if not adequately accounted for. Both
these spectral bands must be considered in the long term photostability of an HCPV system,
though they will have varying effects and solutions.
The models presented here can be easily adapted for a large range of desired materials,
whether glass or plastic. Most glasses have excellent photostability and are insensitive to both UV
and even moderate IR concentrations, and thus the examples are shown for poly(methyl
methacralate). This is a common optical plastic that is readily available in large quantities and
easily machined. PMMA is representative of many available optical plastics; UV can damage the
polymer chain and the material has substantial absorption in the IR, which can cause thermal
76
damage with sufficient power densities. PMMA has excellent transmission in the visible, which is
one of the reasons it is used so commonly as an optical material. The same modeling techniques
can be applied to glass, which usually is much less susceptible to either UV or IR, and applies for
other plastics, which have different susceptibilities that can be modeled readily.
UV light from the sun makes up only a small fraction of the energy hitting the Earth’s
surface, but the high energy photons are capable of causing damage on the molecular level. Solar
UV poses a risk to the long term durability of many optical materials even without concentration,
and thus must be considered. Plastic materials are made of long chains of individual monomers,
and these chains can be broken by photons with sufficient energy. This chain scission causes the
material chain to break down, and pushes the UV absorption edge farther into the visible. The
material begins to absorb blue photons and thus appears yellow. This reduces the concentrator
optical efficiency and can cause thermal runaway if enough light is absorbed. For PMMA, the
Figure 2-21: The UV transmission characteristics of PMMA compared with the solar spectrum incident on the surface. The chain scission peak will cause damage to the material, but a radical scavenger can be
added to absorb this UV.
77
absorption maximum that causes chain scission is at approximately 300 nm. Light with a
wavelength shorter than approximately 330nm must be removed even for applications with no
concentration. Figure 2-21 shows the UV overlap of the solar spectrum and PMMA absorption.
The chain scission absorption is also shown, and a doped PMMA material that is designed to resist
UV damage are also shown.
There are several possible methods for preventing UV damage from light guide systems.
The cost of implementing these methods determines the practicality to a large extent. The UV is
ideally removed before entering the active regions of the light guide. The cover layer can be made
of a material designed to absorb UV without suffering damage, or it can have a coating applied to
this outer surface to reflect or absorb the UV. Another means of removing UV is to employ a
reflective design. Most metal reflector coatings absorb UV, and thus prevent injection into the
guide. This does require the material endure an unconcentrated level of solar radiation before
reaching the reflectors. Many guide materials can also be treated with dopants that preferentially
absorb the damaging UV and prevent damage to the host material.
Excess heat generation from infrared absorption is another material challenge facing HCPV
systems. While many materials and applications must consider UV exposure as a failure
mechanism, the high concentrations of sunlight make overheating from absorption an additional
problem for HCPV systems. IR photons are too low energy to break most chemical bonds, and
thus small amounts is not hazardous to most materials. A large portion of solar radiation incident
on the Earth is infrared (as much as 50%), and when concentrated several hundred times, this can
lead to thermal runaways. Many PV materials can efficiently convert the near IR into electricity,
and thus materials used in concentrators should not absorb this energy before reaching the chip.
78
For most HCPV applications, IR absorption restricts the use of most materials, notably
plastics, anywhere in the concentrator system that carries concentrated flux. For common Fresnel
concentrator designs, this dictates that the secondary concentrator and any encapsulating
material be made of glass, as a plastic material in that position absorbs enough IR radiation to
deform or even melt. The absorption spectrum of most plastics has a significant overlap with the
solar spectrum in the infrared, and thus these materials cannot be used where there is sufficient IR
flux to deform or damage them. The overlap between the solar spectrum and PMMA absorption
is shown in Figure 2-22.
Figure 2-22: The spectral overlap between solar radiation hitting the Earth's surface and PMMA. PMMA shows excellent transmission in the visible, but significant absorption in the infrared at wavelengths longer than 1100 nm.
In light guide concentrators, there are two key regions where light may be concentrated
enough to result in thermal damage to the materials. The first is the foci of the lenslets, which
have concentrations of several hundred times. The concentrators are designed to have a small
0 500 1000 1500 2000 25000
0.2
0.4
0.6
0.8
1Overlap of Solar Spectrum and PMMA Absorption
Norm
aliz
ed F
raction
Wavelength (nm)
Solar Irradiance
Pmma Absorption
Solar Energy Absorption in PMMA
79
spot that can move some on the injection face to provide a more substantial field of view with a
sharper roll off. The size of the focal spot will depend on the lenslet, but is not diffraction limited,
as there is only one optically active surface. Each individual lenslet collects approximately 1mW of
incident radiation and focuses to a spot of approximately 10 µm diameter. This region of high flux
is ideally on the injection face, and therefore this material must be both able to endure this focal
spot and machinable to the precise forms required for the injection prism geometries. The effect
of this focal region depends on the absorption and thermal properties of the material. For many
materials, the actual power dissipated is low enough that the material does not heat up to a
problematic temperature. In order to model this, the absorption information from LightTools was
used in an FEA thermal simulation in Solidworks. The power absorbed throughout the focus cone
was dissipated into the surrounding air, and the final stable temperature profile was modeled. For
PMMA, this model show the focus of the lenslet heating less than a degree. The simulated
thermal effects of a PMMA guide layer absorbing the focused cone of light is shown in Figure 2-23.
80
Figure 2-23: The thermal effects of a cone of light focusing onto an injection facet in PMMA. The temperature differential is not projected to be more than a degree, even at the focal spot.
In addition to the lenslet focus, light guide concentrators face thermal challenges from IR
absorption in the guide layer close to the chip where the concentration is several hundred times.
The flux through this region must be high, and covers the entire cross section of the guide as
opposed to a small spot at the lenslet focus. This region can see a flux of several hundred suns
before the secondary concentrator. The high concentration region near the chip is problematic for
most HCPV systems, and thus the material options are extremely limited and usually require high
quality glass. The nature of light guide concentrators allows materials to be used that cannot be
employed in other systems.
81
Light guide concentrators allow the use of some materials that are not typically available
for HCPV applications due to the self-filtering of the spectrum that travels down the guide. Any
injected light with a wavelength that is strongly absorbed by the guiding material is quickly
removed from the system and does not propagate farther down the guide. Light that is absorbed
by the material cannot build up to levels that damage the material. This is illustrated by a
simplified model which considers a light guide to be a two dimensional system of arbitrary length
laterally. Light is assumed to be injected evenly along the guide’s length and to travel directly
along the axis of the guide. This ignores the discreet injection facets and the increasing angular
spread but illustrates the effect.
If the flux incident on and injected into the guide is given by Φ a d the ab orpt o of
the mater al g ve by α , the the power ab orbed a d ta ce X from the e d of the gu de
given by
As can be seen from this equation, the power absorbed at any given distance cannot exceed the
incident flux for that wavelength. The power dissipated by the guide cannot exceed one sun
worth of intensity. For spectral components where the guide material has high absorption, the
Figure 2-24: Simplified model of light guide IR absorption. Light is assumed to be injected evenly traveling towards the chip at the left.
82
light injected upstream is removed by the material quickly, and this prevents the buildup of this
energy. This self filtering in light guide concentrators may allow materials such as plastics that
absorb in the IR to be used even in the regions of high flux encapsulating the chip, as the light
guide filters out the regions of the solar spectrum that damage these materials. Alternative
concentration systems do not have this linear property, and thus the infrared that is strongly
absorbed by the guide reaches concentrations similar to the useful light, and damages any plastic
parts carrying high flux.
The light guide concentrator geometry enables the use of a PMMA guide layer for carrying
a high flux and not being damaged. PMMA absorbs fairly strongly in the IR at wavelengths longer
than approximately 1200 nm. Wavelengths longer than this are useful for some cells, especially
triple junction cells, but PMMA cannot efficiently be used to couple these wavelengths in a light
guide concentrator. This highlights the need to optimize the cells and concentrator materials in
parallel, as the spectrum delivered to the cell is dependent on the materials used in the
concentrator, especially in a light guide system where the light is filtered strongly.
For a guide layer of PMMA, infrared light with a wavelength of longer than 1600nm does
not propagate more than a few centimeters before being absorbed almost entirely. Light with
wavelengths between 1200 and 1600 nm travels for a significant distance within the guide, but
this energy is dissipated at approximately the same rate as it is introduced near the end of the
guide. PMMA has excellent transition in the visible, so a minimal amount of this light is absorbed,
even at hundreds of suns worth of concentration. While the optical efficiency for the entire solar
spectrum is not particularly high, most of the light that is lost in a long light guide will be the
infrared, which is less useful to many designs of PV chip. The approximate absorption at the end
83
of the concentrator is compared with the solar spectrum in Figure 2-25 for several different
lengths of concentrator.
Figure 2-25: Heat dissipation by various lengths of PMMA light guide concentrators. Most of the absorption is in the infrared. The absorption approaches the solar spectrum with a characteristic length determined by the material
absorption of the guide layer.
Light guide concentrators have the advantage of limiting the amount of heat that is
created by removing problematic light upstream in the guide. For absorption profiles that are
highly spectrally dependent, such as PMMA, the heat that must be dissipated at the end of the
guide can be fairly similar the heat that is dissipated at regions of much lower concentration. This
enables high flux regions to endure a relatively moderate temperature rise compared to
alternative concentration systems. This screening of absorbed radiation can allow many more
materials to be used as guide layer materials while still undergoing only modest temperature rises
200 400 600 800 1000 1200 1400 1600 1800 2000 22000
0.5
1
1.5Heat Dissipation from Absorption in PMMA Light Guide Concentrator
Spectr
al P
ow
er
Density (
W/m
2*n
m)
Wavelength (nm)
Solar Irradiance
Pmma Absorption
Absorbed at 200mm
Absorbed at 500mm
Absorbed at 800mm
84
despite transmitting hundreds of suns worth of flux. To determine the mechanical effects of this
absorption across the length of a guide, a thermal model of a PMMA guide layer was made to
dissipate the heat predicted through absorption through air convection through the bottom
surface. The temperature increase in this system is shown in Figure 2-26. In this example, the end
of a 600mm long guide increases in temperature by approximately 16 °C above ambient
temperature. This is not near the softening point of PMMA which is 105°C. While this
temperature change may cause some warping, the ambient temperature variation is likely to be
greater than the temperature change due to IR absorption.
Figure 2-26: Thermal model of temperature increase due to infrared absorption of a PMMA two stepped light guide. In this model, the only means of dissipating heat is convection from the bottom surface interacting with 300K air.
85
2.4 Integrated Module Performance Modeling
HCPV systems are comprised of many components, and a critical factor for assessing a
concentrator is how it is expected to perform when assembled into a full HCPV module. Extensive
research and development work is being done on several other components of the HCPV systems
such as high efficiency solar cells, high precision solar trackers, and improved electronics and
storage. While it is expected that these components will improve during any further development
of this concentrator family, currently available technologies are modeled to provide a baseline for
potential yearly energy generation.
For the module output simulations, the silicone on glass design of the ideal modeled
reflective dimple tree system was used operating at 500x concentration. The system was designed
to be 500mm long with a 2mm thick guide layer with a 2x secondary concentrator to achieve the
desired concentration. This design was used due to the faster ray tracing than either of the
stepped systems, as the yearly output requires a substantial amount of computation time. The
concentrator system was assumed to be mounted on a tracker with a tracking accuracy of ±1.0°. A
simple inverter with 95% conversion efficiency was assumed, and no storage was used. This is
consistent with a small pilot plant supplementing the grid, but storage is a critical component if
grid support is not available.
The system is assumed to be installed in Phoenix, Arizona, which is an ideal location for
concentrating photovoltaics. CPV systems require a large amount of direct sunlight and can utilize
very little of the available diffuse light. Locations with frequent cloud cover provide much less
energy for all solar projects, and concentrating systems have performance reduced much more
than flat plate designs where the increased diffuse radiation on cloudy days can slightly offset the
86
loss of direct radiation. A comparison between direct/diffuse breakdown of a characteristic
summer day and a characteristic winter day in Phoenix Arizona and Chicago Illinois is shown in
Figure 2-27.
Figure 2-27: Characteristic summer and winter days in Phoenix (top) and Chicago (bottom). Phoenix has much less cloud cover, and thus much more direct sunlight. High concentration systems are primarily only able to collect direct
sunlight.
87
2.4.1 Cell Models
HCPV systems are designed for use with ultra high efficiency cells that would be cost
prohibitive without the high concentration reducing the cell area required. These are typically
multijunction cells, which have stacked several different semiconductor materials with each
junction efficiently converting a portion of the solar spectrum. Development of these cells has
yielded consistent progress in efficiency and cost, with the efficiency record being broken many
times in recent years. For these models, a cell similar to the triple junction cell produced by Solar
Junction is used. This cell held the efficiency record when it was first produced, and with 41%
conversion efficiency at 500x concentration, represents an extremely efficient cell.
This cell is a stacked triple junction cell with an Indium Gallium Phosphide top cell (InGaP)
for converting the high energy visible light, an indium gallium arsenide (InGaAs) middle cell for
converting the low energy visible and near IR, and a germanium bottom cell (Ge) for converting
the lower energy IR radiation. Light incident on the top cell with energy below the bandgap of the
top cell travels through the top cell to be absorbed by the other junctions, while light that is
sufficiently high energy is converted more efficiently. A schematic of this cell is shown in Figure
2-28, with the spectral regions converted by each layer highlighted in the appropriate colors.
88
Figure 2-28: Schematic for a triple junction solar cell. Each successive layer efficiently absorbs and converts the appropriate spectral region while allowing lower energy light to pass through to the cells underneath. Image courtesy
of Solar Junction
An important consideration for any multijunction solar cell is how the system is wired.
Each junction can be wired separately (parallel) or having only the electrodes at the top and
bottom of the stack (series). Wiring in series increases the output voltage and does not require
the additional complexity introduced by adding electrodes between the various layers. However,
wiring the cell in series requires that the cells be current matched, or the cell producing the least
current limits the other cells. While careful tuning of junction bandgaps can match the solar
spectrum, the incident spectrum varies substantially over the course of the day. When the sun is
lower in the sky near sunrise or sunset, there is substantially more effective atmospheric
insulation, and thus more high energy light is scattered, and the sky appears to have a redder color.
This can cause the top cell to be current limiting, and reduce system performance over the course
of the day.
The cell is modeled in LightTools using a User Defined Merit Function (UDMF) developed
by the University of Purdue34 to model multijunction solar cells. This function takes into account
parameters such as the power and spectrum of light incident on the cells and other parameters
89
such as junction temperature to model the performance of each junction individually. This UDMF
calculates the efficiency of each junction and the efficiency if the cells were all wired in series. The
values reported here are for a system where each junction is wired individually and the efficiencies
are added together, as the series values require careful optimization of the band gaps.
2.4.2 Yearly Energy Output Simulations
The output of a full concentrator module was modeled using the LightTools Solar
Simulator utility. This had information about the insulation at various locations every day in 2005.
This utility separated the diffuse and direct radiation into two separate sources, the direct having a
±0.26° angular spread while the diffuse radiation was assumed to come from a hemisphere with
uniform intensity. The electrical output resulting from both diffuse and direct illumination was
Figure 2-29: Averaged monthly output from the system with insulation data taken in Phoenix in 2005. There is substantial variability between the months, as this particular year had a unique distribution of
weather in addition to the annual seasonal variability.
90
calculated. As expected, the concentrating system had an efficiency of less than 2% for diffuse
light. The averaged output for several months of the year is shown in Figure 2-29. This was
particular to the year 2005, but provided a representative example of the electrical power
produced by this system.
The yearly performance of the system was summed over the entire year to show the
power this system would optimally produce annually. The total incident power on the 1m2 input
aperture of the system was 3004 kWh, of which 2466kWh was direct sunlight. The system
converted 711kWh of this into electricity, which is a conversion efficiency of 28.8%. The efficiency
varied by less than 0.3% month to month, which is expected for a tracked system with
independently wired junctions. To compare to flat panel systems, the total incident light must be
accounted for, and in this case the efficiency dropped to 23.6%, which was still higher conversion
efficiency than available flat panel systems.
Three new types of concentrator have been designed and the performance for a variety of
fabrication parameters has been modeled. In addition to the optical performance of these
systems, thermal models provide critical information for determining the lifetime of these
concentrators when deployed. Finally, an approximation of the performance of these
concentrators when assembled into full modules has been made, which provides information
about potential system output when deployed.
91
3 Fabrication and Testing
The first dimpled light guide prototypes produced at University of Rochester were part of
a joint development effort that ended in 2009. Using the information gained from constructing
these prototypes35, design work continued with the aim of increasing optical performance and
manufacturability. The initial concepts and modeling work described in the previous chapter
represented the potential new generation of light guide concentrators, but a new manufacturing
partner would require a whole new process development.
The University of Rochester entered into a sponsored research agreement with Rambus to
produce the next generation of concentrator prototypes. The goal of the project was to produce
functional concentrators with three different structures: wedged stepped, two stepped, and
reflective tree. While Rambus had substantial expertise in manufacturing custom micro-optic
systems, these structures would require substantial process development.
3.1 Overview
At the outset of the project, the relative difficulty of producing each of the three systems
was unclear, so work was done in parallel on all three designs to maximize the chance of
producing a successful system. One critical challenge was determining how accurately the dimple
and lenslet structures could be produced. In addition to the microstructure geometries, the
component materials required some development work, which would also occur largely in parallel.
For all the designs, a low index adhesive capable of curing the acrylic lenslets to the glass or acrylic
guide layer was necessary. The system with a glass guide layer required an index matched
material that could be cast onto the glass to form the dimple layer. The models of these systems
are shown below in Figure 3-1.
92
Figure 3-1: The wedged stepped concentrator model and illustrative schematic of the wedged stepped concentrator (a), the two-stepped concentrator model and schematic (b), and the reflective concentrator model and dimple
schematic (c).
For the first set of prototypes produced, the size of the entrance aperture must trade off
demonstrating scale and manufacturing plausibility. While the systems were designed to be
approximately half a meter along the guiding direction, this was impractical for a proof of concept
system. The size of the guide plane perpendicular to the guiding direction was inconsequential if
the side walls were polished and at least one full repeat unit was formed, as these could be tiled
indefinitely. This project was designed to demonstrate the ability to manufacture these new
geometries, and thus the final concentration was not as critical. The system was thus designed to
maintain approximately the same geometric concentration as the original prototypes (~60x) and
93
the goal for the guide plane was set to be 120mm by 21mm for an approximately 2mm thick
guiding layer.
It was anticipated that the two most challenging parts of this system were to be accurately
producing the lenslet and dimple microstructures. The University of Rochester was responsible for
most of the metrology of these components and to incorporate the results into modeling software
to guide further development work. This chapter details the results of prototyping the lenslet
arrays, the dimple arrays, and then the efforts to combine these two critical components with the
rest of the system components into functional concentrator modules.
3.2 Lenslet Arrays
Manufacturing of conic microlens arrays with an offset rectangular aperture was believed
to be a new technology as of the writing of this thesis. In order to yield an offset aperture with a
high surface quality and shape, each transverse row of lenslets (perpendicular to guiding direction)
was cut as an individual blade as shown in Figure 3-2. This allowed for a steep discontinuity at the
edge of each aperture. An entire repeat unit of the wedged system (15 blades with 15 lenses
each) was produced by this method and eventually had the potential for tiling into a larger pattern.
Due to financial and time constraints, the larger pattern was not pursued. Each prototype instead
had several repeat units of lens arrays individually aligned.
94
Figure 3-2: Definition of angles and transverse direction. The blades of lenslets were combined together to form a master.
In order to function properly in the concentrator system, these components had to meet a
number of critical specifications. This required measurement of the lenslet surface shape, the
lenslet surface microroughness, the thickness of the lenslet layer, and the exact spacing between
the focus of each lenslet. Testing these properties required a variety of techniques and equipment.
The surface quality was tested over a small region near the apex of each lens using a ZYGO Nuview
white light interferometer. The thickness of each lens array was checked both using a visual
microscope and calipers. To test the surface shape and focal spacing of the lenslet array, a
customized Twyman-Green interferometer was built.
3.2.1 Lenslet Interferometer
One of the most important qualities in a lenslet array was the precise location of the focal
spots. Matching the spacing of the lenslet focal spots to the injection facets of the dimple array
was critical to optical performance. It was important to know precisely the location of the focal
spots for a lenslet array, both to determine any systematic spacing error and the random variance
of the focus location for each lenslet. If the spacing between the foci and injection facets was not
consistent, the field of view would depend on the injection location. A field of view that varied
95
over the input aperture would cause the integrated guide field of view to lose its sharp cutoff. A
mismatch of greater than 75µm over the input aperture would reduce maximum on-axis
performance.
Another critical parameter for diagnosing lenslet performance was measuring the surface
error of the lenslets. This provided information of the blurring of the focal spot which would
reduce the field of view of the system, or, if there was a sufficient error, degrade the on-axis
performance. This was also useful for diagnosing manufacturing errors such as scratches that
would either scatter incident light or fail to couple a certain area of the input aperture to the
injection facet.
In order to get precise measurements of both the focus location and the surface profile of
each lenslet, a unique Twyman-Green interferometer has been built. The schematic for the
system is shown below in Figure 3-3, and is designed to measure a single lenslet at a time. The
mirror in the reference arm is mounted on a PZT to enable phase shifting. The system uses N+1
bucket phase stepping algorithm to recover the phase difference between the two arms36. The
PUMA algorithm37 is used to unwrap the phase profile. If the back surface of the lens array is
considered to have negligible error, this gives an accurate surface map of the system.
96
Figure 3-3: Schematic of custom Twyman-Green interferometer used to characterize lenslet arrays.
A Graphic User Interface (GUI) was created to allow the automated inspection of a full
array of microlenses. The lenslet array was mounted on a set of precision motorized stages to
allow full control of motion in all three dimensions. The program was designed to go to the
approximate coordinates of each successive lenslet in an array, find the location of the optimal
focus, and then take an interferogram to determine the surface profile of the lenslet at the
optimum focus. This allowed the precise determination of the focal point lattice of the lenslet
array and the shape error of each lenslet.
The CCD was manually focused into a lenslet in one corner of the array, and this stage
position provided a reference coordinate. The zero coordinate was found by moving the stage to
reduce the fringe density. An error in any of the three coordinates produced a characteristic
97
effect on the phase pattern. An error in the x or y coordinate caused a linear tilt in the phase over
the lenslet aperture, while an error in the z position gave a parabolic error around the apex of the
lenslet. The coefficients of both of these errors were proportional the position error in the
appropriate coordinate, with the conversion factor dependent on the specific lenslet geometry.
In order to calibrate the error induced by a position shift for a specific lens array, the
system was moved a controlled distance in each direction. After moving a controlled distance
from the approximate zero, the phase error was calculated and fit to the appropriate function (a
line or parabola). The system was then moved the same distance from approximate zero in the
opposite direction to compensate for any error in the initial reference coordinate. By fitting this
difference with a known position change, the ratio between position movement and the
magnitude of the induced phase error could be determined for all three directions.
Once the relation between the induced phase error and position error was calibrated for the
specific lenslet, the system began measurement of any number of desired lenslets in the array.
Either a list of approximate lenslet coordinates could be entered or was generated from the ideal
lenslet spacing and pattern. The system travelled to the approximate coordinate of each target
lenslet and took a measurement of the phase error. This information was then converted into a
position error, and the system calculated and travelled to the position of the optimum focus for
that lenslet. In addition to taking the interferogram at the optimum focus, the precise coordinates
of the stages were then recorded.
The recorded coordinates of optimal focus were then mathematically rotated about the
initial reference point in order to remove as much tilt as possible from the system. The set of
coordinates was first rotated about the Z axis so that the best fit of the first row of lenslets was the
98
X-Z plane. The coordinates were then rotated about the Y axis so that the first row of lenslets was
best fit by the X-Y plane, and then rotated about the X axis so that the first column of lenslets were
best fit by the X-Y plane. These rotations removed any residual tilt or clocking introduced by the
mounting of the lenslet array. These processed coordinates were then compared to the optimum
lenslet positions. Each coordinate error was recorded, and the total spatial error of the optimum
focus was calculated and displayed graphically.
In addition to calculating the error in the optimum focus for any given pattern of lenslets,
this system was designed to measure the surface error at the optimum focus of each lenslet. By
measuring the phase error at each position in the lenslet aperture, an optical path difference could
be determined. The laser used in the interferometer was a HeNe laser(MFC), and the refractive
index of PMMA was well known at the HeNe wavelength of 632.8nm. The optical path difference
could be converted into a sag difference (departure from the apex along optical axis), which
provided a surface figure error. A customizable aperture of each surface profile was then set, and
the RMS surface error of each lenslet within this aperture was calculated.
In order to ensure that the lenslet interferometer was running properly, a previously
characterized lens array was measured. The reference array was a set of conic lenses with
hexagonal apertures that was used for a previous dimpled light guide design. The system was
known to have a slight pitch mismatch in one direction, determined by an alignment variation
across the aperture when used with a prior dimple array having known spacing. Figure 3-4 shows
the output of the GUI measuring the Y error, perpendicular to the direction of the pitch mismatch.
There is no discernible trend, and the RMS surface error of the first lenslet is under ±2µm.
99
Figure 3-4: The output of a sample patch of the reference lenslet array. The Y Error is selected and shows no notable systematic errors and a fairly small random variation. The phase profile is presented on the right, from which an RMS
error can be calculated.
The lenslet array under test was known to have a systematic spacing error in one direction.
This was originally found by measuring the variation of the field of view of an assembled light
guide constructed using a lens array made from the same process38. This was confirmed and
quantified more precisely using the lenslet interferometer. The systematic error in the X direction
can be seen in Figure 3-5. In order to verify that the systematic spacing error was actually in the
lens array and not an artifact of the interferometer, the lens array was rotated 90° and measured
again to make sure that the error then showed as a Y spacing mismatch. The results of the original
measurement of the spacing mismatch in X and the subsequent spacing mismatch of the rotated
sample confirmed this, as shown in Figure 3-5.
100
Figure 3-5: The systematic spacing error can be seen in the reference lenslet array. The original measurement showing an X spacing mismatch is shown on the right, and the left shows a consistent spacing mismatch when the
sample was rotated 90° and remeasured.
The lenslet interferometer verified the spacing mismatch of the reference lens array, and
this capability could easily be extended to any desired lenslet shape and position distribution. This
system was only set up for use with refractive lenslet arrays. The reflective design was tested by
using a part that was uncoated. This reflective system did not have the ideal surface shape for
creating a point focus when used as a refractive system, but the expected error was subtracted
from the final profile to determine the optimum focus and record the corrected surface profile.
3.2.2 Refractive Lenslet Arrays for Stepped Systems
Both the two stepped design and the wedged stepped light guide design were designed
to make use of the same lenslet array in order to save time and cost. The final component
specification was to use a 1mm by 1.5mm rectangular aperture (1mm along the guiding axis)
with a 100µm step offsetting each successive row of lenslets. The optical axis of each lenslet
was offset along the guiding axis by 0.35mm. The radius of curvature was designed to be
Error (mm) Error (μm)
101
1.11mm with a conic constant of -0.45. This produced a lenslet with maximum sag of 0.80 mm.
The semi- aperture of this lenslet is larger than the radius, and therefore a spherical surface
cannot be used with this curvature. The system was designed to be made of PMMA and have a
thickness of 1.35mm from the apex of the lenslets to the back surface. A model showing a
small section of the designed lens array is shown below in Figure 3-6.
Figure 3-6: A small patch of the refractive stepped lens array. The lenses have a rectangular aperture and the optical axis is offset from the center of the lens aperture. In this picture, the light will be guided toward the left.
For the optimal two stepped design, the lenslet focus was stepped vertically to make sure
the focus of the lenslet was coupled to the injection facets of the thickening guide. These lenslet
arrays presented a large manufacturing challenge, and it was decided to use the same lens array
for both stepped prototypes. The lens array for the two stepped prototype would be cast on a
wedged piece of acrylic, and thus the height of the focal spots would vary smoothly over the
prototype. This would result in the lenslets for the two stepped system focusing slightly off the
dimple facet. This defocus would be greatest near the vertical steps in the guide and would be at
most half of a vertical step (50µm for this prototype). While this was not expected to decrease on
axis performance (barring other errors or misalignments), this would reduce the expected field of
view substantially.
Light Guiding Direction
102
Fabricating these lenslets required a substantial amount of manufacturing development,
and several new approaches were tried in order to achieve the desired surface shape and finish
necessary. In order to verify that the surface quality was adequate, the master of these lenslets
was measured in a Zygo white light interferometer. The lenslet has a relatively large curvature,
and thus the system was only able to measure a small region near the apex. The interferometer
was able to measure approximately 200 µm from the apex, which covered about 14% of the
entrance aperture of each lenslet. This showed an excellent surface with the exception of a small
region at the apex of the lenslet. This was an unavoidable defect given the manufacturing process
used. A vertical slice slightly off center (to avoid fitting errors caused by the imperfection at the
apex) was fitted to provide an approximate curvature and information about the surface quality,
as shown in Figure 3-7.
Figure 3-7: White light interferometer measurement of the refractive stepped master. The red vertical line on the surface map is fitted to both a variable radius and to the designed radius. While there is a slight discrepancy, this only
represents the 15% of the lenslet aperture near the apex. The microroughness of these surfaces is approximately 10nm.
The lenslets were produced in small repeat units (15x15 lenslets) which were intended to
be copied and patterned into a larger part. Due to time and financial constraints, it was decided to
103
individually align a series of smaller repeat units that had one row of lenses removed in order to
enable manual alignment. Each of these repeat units covered approximately 12% of the input
aperture of the prototype. While this would substantially reduce full aperture efficiency, these
concentrators would still be able to prove out the experimental manufacturing process and design.
When these repeat units were made, they were inspected under a visual microscope which
showed substantial damage to the outermost rows of lenslets. This was an expected result the
method employed to remove these lenslet arrays from the mold. In addition to damage localized
to the lenslets on the perimeter, the parts showed a bow which was visible.
The bowed lenslet arrays posed a potential problem, as the lens array in the designed
system had a flat back surface, and the apices were designed to be a constant height. A variation
in height across the lenslet array would cause a subset of the lenslets to be defocused. Though
this bow posed a problem if the part were used in a concentrator, the bow could be removed by
using a strong low index adhesive to keep the bottom of the lenslet layer securely attached to the
guide layer. Despite the visible bowing in the fabricated lenslets, the thickness between the
bottom surface and the apex was consistent. The parts produced were found to be thicker than
they were designed. It proved extremely difficult to control the thickness of the part without
damaging the lenslet surface, and thus the guide layers were made to be thinner to compensate
for this error.
When measured on the lenslet interferometer, the bow in the part was expected to
dominate the focal spot distribution. Assuming the part was all of equal thickness but bowed, the
focal spots would be shifted slightly in Z position and slightly shifted along the plane of optimal
focus, as shown in the exaggerated schematic in Figure 3-8. The visible bow of the parts was
104
consistent with the bow direction shown in Figure 3-8. In addition to moving the focal spot, the
bow would cause some lenslets to operate off-axis, which would induce aberrations, most notably
a substantial amount of coma. This would increase the spot size and cause the RMS to suffer
based on the magnitude of the induced aberrations.
Figure 3-8: An exaggerated schematic of the bowing of the lenslet array. This is the direction of bow as can be seen visually. This will cause a focal shift in the Z direction as seen by the green arrow and will reduce the X spacing shown
by the black arrows.
A lenslet array from the same batch as those integrated into the concentrator prototypes
was measured in the lenslet interferometer. The perimeter of the lenslet array was not usable,
and thus the inner region of the part was measured. The central 8x7 group of lenslets was
measured, as this was the portion that was used for characterization of the prototype. The RMS of
the measured lenslets showed an RMS of less than 1µm. The one systematic error that was visible
was a consistent defect in one bottom corner of each lenslet. This defect was consistent with a
stress induced by the removal from the mold. While this defect was consistent, it was small
105
enough that was is not expected to cause a focus issue. The surface profile of a representative
lenslet and the calculated RMS surface value at the best focus was measured and shown in Figure
3-9.
Figure 3-9: The surface profile of a representative refractive lenslet and the calculated RMS surface error of each measured lenslet at the best focus. Both color scales are in mm.
Error (μm)
Error (mmx10-4)
106
The RMS for the refractive lenslet array was good for an experimental manufacturing
process. The focal spot distribution was expected to be dominated by the bow in the lenslets. As
expected, both the X error and the Z error were dominated by this defect as shown below in
Figure 3-10. The bow was clearly visible, and each row was fit to a parabolic shape to quantify the
distortion of the lenslet array.
107
Figure 3-10: The focus error in both Z (top) and X (bottom) directions. The Z error shows a low central region consistent with a bowed array, while the X error shows a consistent spacing mismatch being too close together. Both
of these profiles are dominated by the bow in the lenslet array as expected.
While these two profiles were dominated by the warping of the part, the Y coordinate
error was dominated by a misalignment between two rows. This imperfection could also be seen
visually, as there was some form of defect between two of the lenslet rows, as shown in Figure
Error (mm)
Error (mm)
108
3-11. This was consistent with the manufacturing process and may have been caused by an
imperfection on one of the blades corresponding to a row of lenslets that had some defect. This
error manifested in the other error coordinates, as the discontinuity between the third and fourth
rows was apparent.
Figure 3-11: The Y coordinate error was dominated by an imperfection between the third and fourth row. This is indicative of an imperfection preventing the two rows during assembly.
The Z error was used to approximate the magnitude of the bow by fitting the X coordinate
to the measured Z error. The measured depth of the focal spots and their approximate X
coordinate was fitted as shown in Figure 3-12. A parabola was determined to be an accurate fit
for the warping of the parts. This provided an approximation for quantitatively modeling the
bowing of the lenslets. In addition to this method, the corners were measured with a visual
Error (μm)
109
microscope with a precise adjustable focus. Both these methods provided an approximate bow of
240um at the corners of the lenslet array.
Figure 3-12: The fitting of the bow of the refractive lenslet. This fit should yield a Z error of approximately 240µm at the corners of the lenslet array.
While the warping dominated the lenslet interferometer measurements, it could be
negated by using a strong adhesive for the low index layer. Further work was not done to remove
the bow, as this should only have been present in the small units used in the system. For a much
larger system in which these units would be patterned, the surface was expected to be much
flatter. The primary goal of this development work was to demonstrate capability of producing a
high quality conic lenslet with an offset rectangular aperture, and this was accomplished. These
systems were adequate for characterization in an assembled prototype for demonstration
purposes.
3.2.3 Reflective Lenslet Arrays for Dimple Tree Systems
The reflective lenslet array provided a different set of challenges than the refractive
system. The sag of the surface was substantially less than for the refractive system, but the
aperture was designed to be offset in both directions, and the acrylic surface required a reflective
110
coating. While offsetting the lenslet aperture in the transverse direction provided slight
performance improvement in simulations, this would have required an entirely new
manufacturing process. The offset of the lenslet array in the transverse dimension was thus taken
out so that a similar method could be used to manufacture these lenslet arrays as was used to
fabricate the refractive lenslet arrays. This reduced the performance of the 600mm modeled
system by approximately 3%, but for the current prototype size was not expected to have any
effect on the optical efficiency.
The final designed system also had a rectangular aperture with an aperture of
1.35 mm in the transverse direction and 0.90 mm in the guiding direction. The reflectors had a
parabolic shape and a radius of curvature of 5mm. This resulted in a surface sag of 0.16mm
and a 3um maximum departure from a spherical surface. The final designed thickness of the
reflector arrays was designed to be 0.45mm from apex to the top of the reflector layer. The
alternating transverse shift between two rows of lenslets was designed to be 0.26mm. For
similar reasons to the refractive lens arrays, the reflector arrays were produced as small
patches measuring 14x21 mm which would be attached to the guide layer individually. The
system was originally designed to have a simple silver coating, but this was later changed to an
aluminum coating for cost reasons. This does did affect the optical design, though the
reflectivity and optical efficiency were reduced by approximately 3%.
These lens arrays were produced in a similar manner to the refractive systems, with each
row of lenses being cut individually and then manually aligned to form a master. Before using the
master to produce daughter copies, the surface roughness was checked using the white light
interferometer. The curvature of these parts was substantially lower than for the refractive
111
system, so the data were limited by the field of view of the interferometer. The system was
designed to measure the surface profile of each lenslet with a sample shown below in Figure 3-13.
There was still a small defect at the apex of the lens, but this was less pronounced than for the
refractive lenslets.
Figure 3-13: The surface map near the apex of a reflective lenslet master (right) and the profile of a slice through the apex. The surface roughness is quite low, and the fitting of the curvature shows a slightly weaker curve than
designed.
Daughter copies were pulled from the master and then an aluminum reflective coating
was applied. The lenslet interferometer was not set up to measure reflective microlens arrays,
and thus one of the parts was tested before coating on the lenslet interferometer. The surface
was designed to produce a tight focus when used as a reflector, and thus significant aberrations
were expected when this system was used as a refractive lenslet array, specifically approximately
3 waves of spherical aberration. The expected performance for a perfect system was no longer a
flat phase map when measured through the lenslet interferometer, but the expected phase profile
of a perfectly manufactured system was subtracted in order to determine the optimal focus and in
the resultant phase map that was used to calculate a surface map of each lenslet. The RMS
X (μm)
112
surface profile of the reflective lenslet array was measured, and the surface map of each lenslet
was calculated as seen below in Figure 3-14.
Figure 3-14: The RMS of the reflective lenslet array when used as a refractive system. The error is dominated by the fourth row. It is believed that a scratch along the lenslet array caused a failure of the phase unwrapping algorithm.
The RMS profile of the lenslet array was dominated by the fourth row. The RMS profile of
this row was dominated by what appeared to be a discontinuity along the X-direction. Further
inspection of this row of lenslets showed a scratch or other horizontal defect. This caused a failure
in the phase unwrapping algorithm, which then yielded a discontinuity between the two sections
of the phase profile. The “optimal” focus was then shifted in Y, which produced this RMS profile.
Ignoring the fourth row of lenslets, the RMS profile of the reflective lenslet arrays showed surface
error of less than 100nm RMS, as shown in Figure 3-15.
Error (μmx10) Error (μmx10)
113
Figure 3-15: The RMS profile of a lenslet not in the fourth row. Substantial tooling marks can be seen in the bottom section of this figure, but the profile is very good. There are highly visible tooling marks on the bottom section of the
lenslet, which is consistent with the two regions separated by the discontinuity in the fourth row.
The Y-error measurement from the lenslet interferometer was also dominated by the
scratch on the fourth row, as this caused the phase unwrapping algorithm to yield two distinct
sections with different surface heights. This was fitted as a substantial tilt error, and the optimal
focus was calculated to be substantially off in Y position, as seen in Figure 3-16. While the Y error
was dominated by this scratch, the X error was again dominated by a bow in the system, though
this time in the other direction. This was likely due to the way the part was mounted in the
interferometer, as it was held by tension on the outside perimeter. These parts were thin and
flexible, so a distortion from mounting was plausible.
Error (μmx10-1)
114
Figure 3-16: X (top) and Y (bottom) position errors of the focal spot for the reflective lenslet array. The Y error is dominated by a scratch on the fourth row, while the X error is dominated by a bowing in the system.
The reflective system used in transmission had a much longer focal length than the parts
for the stepped systems. This caused substantial problems with the ability to measure the Z
coordinate of the optimal focus, as this parameter was not as sensitive as for the system designed
as a refractor. The error in the Z coordinate shift was often larger than the calculated shift, and
Error (mm)
Error (mm)
115
thus this measurement was not useful. The Z coordinate shift was thus substantial enough that
the depth of focus of the imaging system began to become a substantial source of error. The Z
position of the optimal focus was thus not able to be reliably obtained using the lenslet
interferometer.
3.3 Dimple Arrays
The dimple geometries presented in this thesis had never been fabricated before, and thus
the prototyping of these microstructures required extensive manufacturing development work. In
order to refine the manufacturing process, small patterns (approximately 10cm2) of the
appropriate geometry were cut into a nickel coated master, and acrylic copies were then made
from these master tools. This work was done by Rambus based on the designs from this thesis.
These copies were characterized in order to refine the machining parameters. Once the small
patch cuts were satisfactory, a longer master cut of the full proposed area was done. This
required a longer cut time (up to 8 days) than had been previously machined, and thus further
development work was required in order to speed up the pattern cutting.
For each dimple geometry, the initial goal was to yield copies of the dimple array in a
variety of different materials in order to test the effect of varying guide materials. Each system
was to be prototyped as a monolithic acrylic piece, with an acrylic guide layer and dimple layer,
and also a polymer dimple layer cast onto a glass guide layer. Light guide systems had not been
produced with an entirely acrylic guide layer, and it was unknown how well the material would
withstand the high concentration light. The glass guide system required development of an
appropriate polymer that had the correct index of refraction and was still compatible with the
processes used to machine the microstructures and adhere to the glass. Both of these approaches
116
were pursued in parallel in order to minimize risks, and both material choices could use the same
dimple geometry master and the appropriate lenslet arrays.
3.3.1 Wedged Stepped Dimple Arrays
The wedged stepped dimple geometry required a surface that was fully covered by the
dimple features. The channels required a sharp transition, as either a large fillet or substantial
draft angle would ruin the angular isolation of this system. The fabrication method used to create
the previous generation of prototypes yielded a fillet radius on corners of approximately 10µm
and a minimum draft angle of approximately 10°. The side wall between adjacent channels in the
designed system approximately 8µm, and thus this would be completely rounded if fabricated by
the same process used for the previous generation. The first cuts of a small patch pattern showed
that the process used was capable of much steeper and sharper side walls. The cutting
parameters clearly still required tuning, especially on the actual injection facet, but the machining
method showed promise for producing the dimple geometry much more faithfully to the design
than for previous fabrication methods. SEM images of these early cuts showed substantial error in
the injection facets, as seen in Figure 3-17.
117
Figure 3-17: The first attempt at an injection facet (left) showed an obvious ripple. After refining manufacturing parameters, the injection features were improved significantly (right).
While the SEM images provided an accurate measurement of the dimple geometry, the
surface roughness was also important to system performance. This was measured using a Zygo
NewView white light interferometer. The system was limited to a field of view of a few hundred
microns, and thus measuring the entire part was impractical. Scans were done of several regions
of the system to measure the surface quality of both the bypass elements and the injection facet.
Surface measurements of the final patch cut were taken with the white light interferometer, as
seen below. The actual measurements required a schematic to show which region of the dimple
structure was shown in the accompanying measurement. The orientation of the schematics was
given a color coded schematic to show consistent height differences (high regions and low regions),
as shown in Figure 3-18.
118
Figure 3-18: Schematic showing dimple layer for the purposes of orienting white light interferometer measurements. The colors of the channel are consistent with respect to height differences.
The surface map is measured for an acrylic dimple layer pulled from the master to
determine measure the surface quality at several regions around the part. A sample of a white
light interferometer measurement of an acrylic part in a region where there is no facet proximate
is shown below in Figure 3-19. In addition to the surface map, one dimensional slices of the data
are presented to highlight the microroughness of the system. The red and blue lines on the
surface map correspond to the vertical and horizontal 1-D scans respectively and are shown in the
corresponding colors.
119
Figure 3-19: White light interferometer measurements of a central region of an acrylic wedged stepped dimple layer. The surface roughness of each channel in this region is approximately ±5nm.
Once the patch cut was satisfactory, the next step was to master an entire repeat unit of
the system. The goal of the system was to be 120mm long and 21mm wide. Even with machining
parameters optimized for speed, cutting a part of this size required enough time that tool wear
and machine failure posed a substantial risk. In order to minimize tool wear and reduce cutting
time, the first full master was cut into aluminum. The softer material produced less tool wear, and
this was expected to increase the chances of the part cutting successfully. However, the softer
material did not produce the geometry faithfully, and this substrate change resulted in a
substantial reduction in the part quality.
X(μm)
120
Due to the extensive degradation caused by cutting the master out of aluminum, another
master was cut out of a nickel substrate. This was more demanding on the tool due to the
increased material hardness and also caused the cut to take approximately 3 days longer. The full
part was successfully cut as shown in the SEM image of the final part in Figure 3-20.
Figure 3-20: An SEM image of the final wedged stepped dimple geometry. The feature replication is quite good, though there is a considerable amount of debris that can be seen on the part.
The geometry of the final replicated wedged stepped part matched the modeled geometry. The
fillet radii were less than 2µm and there were no noticeable draft angles (<1°) from the replication
process. There was a substantial amount of debris on these parts, though some of this debris
resulted from handling after production and was not actually a result of the master cutting process.
121
The image shown in Figure 3-20 was from an acrylic copy pulled from the final wedged stepped
master.
Once the master had been made, two different processes were used to produce the two
different dimple and guide layer combinations. The polymer on glass system required a fairly
expensive glass substrate for this prototype, and thus a destructive SEM test was not performed
on this part. Both the polymer on glass and the monolithic acrylic system were tested on the
white light interferometer to characterize surface quality. Surface quality measurements of the
bypass prisms of the acrylic dimple system were taken as shown in Figure 3-21.
122
Figure 3-21: White Light Interferometer measurements of polymer on glass wedged stepped dimple geometry. Both the middle region and the high region are very smooth surfaces. The casting on glass process introduces some roll off
near steam height transitions, as can be seen in the bottom figure near the abrupt height transition.
While the primary defect in the bypass prism appeared to be debris that was either
collected on the master or after the copies were produced, there was substantial variation in the
X(μm)
X(μm)
123
surface quality of the injection facets. Over 20 injection facets were measured using the white
light interferometer, and every facet observed had obvious defects. Many regions of the
measured facets were high quality surfaces that would produce relatively little scattering. The
facets in the polymer on glass sample were highly variable, shown below in Figure 3-22. One
presented facet showed a substantial amount of debris (Figure 3-22a) on the facet, while the other
had a large smooth region but a missing corner (Figure 3-22b). Variations in the injection facets
caused substantial variations in the injection efficiency and prevented uniform device
performance.
124
Figure 3-22: White light measurement of facets of wedged stepped dimple array cast onto glass. Substantial variations are observed from facet to facet, but there are several regions that will have a small amount of scattering.
The roughness on the injection facet caused some of the light to be scattered from the
surface. Most of the light that was scattered from the injection surface was ejected almost
immediately. If injected with a substantial angle relative to the guiding axis, light initially contained
a)
b)
X(μm)
X(μm)
125
within the guide was coupled out of the light guide after a relatively short propagation distance.
The relationship between the RMS surface roughness and the scattering from that surface was
given approximately by39
This equation showed that light was scattered more strongly at lower wavelengths, higher
angles of incidence, and for rougher surfaces. Light incident on the injection facets hit the surface
at a relatively shallow angle of approximately 41° at the center of the ray bundle. Light incident at
500nm was scattered from the injection facet as shown in Figure 3-23. While the light scattered
was not all lost immediately, most of the scattered light was coupled directly out of the guide or
after a short propagation distance.
Figure 3-23: Scattering fraction from injection facet surface is a function of RMS surface roughness.
126
The injection facets shown in Figure 3-22 did not have a uniform roughness, and thus it
was difficult to make simple predictions for the effect of scattering. There were regions of these
facets that have roughness of hundreds of nanometers, while other small regions had RMS errors
of less than 10nm. For this level of facet variability, the location of the lenslet focus on the facet
largely determined the scattering loss, which ranged between a few percent up to approximately
50 percent.
While the RMS roughness provides a common standard for most surfaces, the injection
facets will be most dramatically affect the optical efficiency for high spatial frequencies. These will
scatter light more than the slower rippling that results from the replication process or other non-
uniformities of the dimple arrays. The spatial frequency components of the injection facets are
best characterized by a 2D power spectral density (PSD), which separates the low frequency
variations due to imperfect replication from the high frequency roughness that will strongly
scatter light. The angular spread of the scattered light will depend on the PSD of the facet, as
higher spatial frequencies will cause more of the light to be scattered into angles that the guide
cannot contain.
3.3.2 Two Stepped Dimple Arrays
The two stepped dimple arrays were made using a similar process to the wedged stepped
dimple array. Instead of offsetting the channels in height, the system was designed to cut each
channel to the same height. This yielded the large flat regions between the vertical steps. As with
the wedged stepped design, the initial cutting parameters were refined by cutting small patches
before cutting a full sized master. The small patch cut was characterized using the SEM, with the
results shown in Figure 3-24. The system had the desired large flat regions, though small channels
127
were still visible. There was a substantial amount of debris on the part, but the geometry was
consistent the design.
Figure 3-24: A small patch cut of the two stepped dimple geometry. Channels are still visible, but the flat regions are relatively consistent.
The two stepped design was different from the other geometries designed in that the
guide layer changed thickness along the guiding axis of the part. This was expected to be a
substantial obstacle to fabricating this design. In order to use the machining methods developed
for the other designs, the two stepped dimple geometry would be cut into a flat part at a slight
angle, and then the dimples would be cast onto a wedged guide substrate, either acrylic or glass.
This required a deeper cut than for the other dimple geometries, as the cut depth had to account
for both the facet height and the height of a vertical step.
128
The white light interferometer images of this system showed well defined channels with
small height variations. This was expected, as a similar method was used as was used in the
wedged stepped geometry. The channels did not line up perfectly, but the small sharp boundary
was not expected to substantially degrade performance. Other than these small steps between
channels, the flat regions were smooth, having less than 20nm peak to valley roughness. In
addition to measuring the surface roughness of the flat region, several of the facets were also
measured, a sample of which are shown in Figure 3-25.
129
Figure 3-25: White light interferometer measurements of the two stepped patch cut. The flat region has clearly defined channels separated by a fraction of a micron while the injection facet cut shows signs of tool wear.
X(μm)
X(μm)
130
After cutting the small patch of the two stepped guide, a full size master was cut. This
system was cut into a nickel coated piece of aluminum. Upon inspection of the final master, it was
obvious that the system had substantial defects. The increased depth of cut caused the cutting
tool to penetrate the nickel coating and rip the aluminum underneath. This caused substantial
gouges in the dimple layer, with an extremely rough surface underneath covering a substantial
portion of the guide layer, as shown in Figure 3-26.
Figure 3-26: The master cut for the two stepped dimple geometry showed substantial damage. The deeper cut appears to have ripped away the base substrate through the nickel coating, and caused deep gouges over a
substantial portion of the dimple layer.
This master cut unfortunately would provide little information if assembled into a full light
guide concentrator. A large percentage of the bottom surface would scatter almost all of the light
that hits it, and thus injected light was expected to be coupled out of the system after a
propagation distance of only a few millimeters. The reason this cut failed was known, and it was
expected that a second cut would be able to fix this problem. However, due to time and cost
constraints, this dimple geometry was abandoned due to the failure of this master cut.
131
3.3.3 Dimple Tree Arrays
The dimple tree array for the reflective system could not be fabricated by the same
method as the stepped dimple arrays and required more development work to accurately produce.
As with the other dimple geometries, the first step was to produce a small patch cut in order to
refine the machining parameters. These initial cuts produced the correct geometry with sharp
side walls, which allowed the angular isolation critical to the system. The small patch cut was
measured using the SEM, as shown below in Figure 3-27.
Figure 3-27: An SEM image of the dimple tree patch cut. While there is substantial debris and visible marks ear the facets, the geometry is roughly correct.
132
The white light interferometer measurements showed substantial surface character up to
500nm peak to valley variance, especially near the facets. Due to the cutting design, the flat
region near the dimple exhibited a substantial amount of tool ringing along the guiding axis. This
could be seen from the white light interferometer measurements shown below in Figure 3-28. In
a small section of the cut, the scattering from this ringing was easily visible with the naked eye.
When measured, the magnitude of this error approximately doubled in this section. In addition to
the ringing along the guiding axis, the system also had a noticeable height step along the
transverse direction when approaching the injection feature. This was not expected to cause as
much degradation in performance, as these height transitions were sharp and run parallel to the
guide axis and thus did not cause substantial deviation of the angular spectrum.
133
Figure 3-28: The first cuts of the reflective tree design showed substantial ringing along the guiding axis and a sharp height transition along the transverse axis. There was a substantial portion of the cut in which these errors were
much more pronounced and scattering from this section of the part was easily visible with the naked eye (bottom).
X(μm)
X(μm)
134
After some refinement of the machining parameters based on these parts, a full size
master was cut of the dimple tree system. In this system, the ringing of the dimple layer near the
injection facets was substantially dampened as shown in Figure 3-29. The magnitude was not
substantially lower than for previous cuts. The reduced frequency and distance covered by the
ringing reduced the effect of this error on system performance. In addition to minimizing the
ringing near the injection facets, the height transition near the facets was significantly reduced by
an order of magnitude.
Figure 3-29: The final dimple tree cut showed substantial dampening of the ringing near the facets. While the magnitude is not substantially lower for the master, the distance covered by the ringing will be substantially reduced.
The low frequency also will reduce the scattering effect.
The injection facets of the final dimple tree parts were measured on an acrylic copy pulled
from the master. The roughness of these facets was measured using the white light
X(μm)
135
interferometer, as presented in Figure 3-30. The injection facets showed substantial roughness up
to 400nm peak to valley, especially along the guiding dimension. This was thought to be due to
tool wear. This scattered a substantial portion of the light from the lenslets. These defects on the
injection facets were much more consistent from facet to facet than for the wedged stepped
prototypes, and thus the injection efficiency of the reflective tree guide was expected to be
reduced by up to 40%.
Figure 3-30: The injection facet of the reflective tree guide system showed substantial roughness, mostly along the guiding direction (right to left in this figure). The marks were consistent from facet to facet.
The acrylic dimple tree guide was successfully reproduced from the master, but casting
the tree design on the glass substrate proved difficult. The polymer on glass guide layer showed
substantial damage to the dimple structure. Accurately reproducing the polymer on glass dimple
layer would require substantially more process development of the casting process. Due to time
X(μm)
136
and financial constraints, the polymer on glass prototype of the reflective tree design was
abandoned.
3.4 Assembled Concentrators
The lenslet layers, guide layers, and dimple layers described above were assembled into
full concentrator prototypes and attached to solar cells in order to characterize the final
concentrator performance. The initial goal was to produce six functional concentrators: a polymer
on glass and monolithic acrylic version of all three dimple geometries. Three concentrator
modules were assembled. The final three modules that were assembled were both guide
substrates of the wedged stepped concentrator and the monolithic acrylic reflective dimple tree
concentrator.
Each of these three concentrators was mounted in a custom plastic housing and the
output face of the concentrator was coupled to a silicon solar cell to test optical performance. The
three modules were measured using a variety of techniques in order to ensure that modeled
performance adequately matched the measured performance of the final system. The three
concentrator modules are shown in Figure 3-31.
137
Figure 3-31: The three final concentrator modules that were fabricated and characterized for this thesis. The acrylic wedged stepped guide (left), acrylic reflective tree guide (center), and polymer on glass wedged stepped guide (right).
3.4.1 Concentrator Module Testing Methods
The final modules were tested using the University of Rochester Solar simulator. This
system was designed to match both the angular extent of the sun and the spectral content at
Earth’s surface. The system was made up of a 1kW Xenon lamp which was aimed at a collector
lens that focused the energy onto a hexagonal mixing rod. This homogenized the angular
spectrum of at the output face of the rod. The output face of the mixing rod was placed at the
focus of a large off axis parabola, which then produced a collimated beam approximately 20cm in
diameter. Spectral filters or neutral density filters could be placed between the output face of the
138
mixing rod and the off axis parabola in order to better match the solar spectrum. A schematic of
the simulator used is shown in Figure 3-32.
Figure 3-32: A schematic of the University of Rochester Solar Simulator. This produces an approximately uniform 20cm diameter beam that matches both the spectral content and angular extent of the sun.
Each concentrator module was attached to a 3mm x 50mm silicon solar cell, which
allowed measurement of the optical efficiency of the concentrator. Before mounting to the
concentrator, each solar cell was characterized so the output could be interpreted when a
concentrator was attached to the cell. The cells were characterized by comparison of the IV
curves to the dual diode model40 . This model considers imperfections in the solar cell as ideal
electronic components, which allows a relative quality of the various solar cells to be determined.
The simplified circuit diagram for this model of solar cell is shown in Figure 3-33. The short circuit
current of the three best cells available was measured when illuminated by the slit aperture to
provide a reference point, and then these cells were mounted in the concentrator modules.
139
Figure 3-33: The dual diode model of a solar cell. Image from pveducation.org41
The IV curve of a solar cell measures the current flowing across the cell as a function of an
externally applied voltage. When no light is applied to the solar cell, the solar cell produces no
current unless there is an external voltage applied. When exposed to light, the solar cell should
produce a current in the absence of an external voltage. This is called the short circuit current and
should be directly proportional to the amount of light incident on the solar cell. The external
voltage that must be applied in order to prevent all current flow is referred to as the open circuit
voltage. The maximum power point is the largest product of the current and the voltage along the
IV curve. The fill factor is a measure of this maximum power point divided by the product of the
open circuit voltage and the short circuit current. A fill factor approaching 1 is a quantitative
method of determining the quality of a solar cell. The measured dark IV curves for the solar cells
mounted to the concentrators are shown in Figure 3-34.
140
Figure 3-34: Dark IV curves for the three silicon solar cells mounted to the three light guide concentrators. The cells were chosen as the closest to ideal fits from an available batch of silicon cells.
The current-voltage relationship for a silicon solar cell with no light incident on it is given
by
where RSe is the parasitic series resistance, RSh is the shunt resistance, I0R is the recombination
saturation current, and I0D is the diffusion saturation current.42 The series and shunt resistances
of the solar cells were approximated from the high voltage current profile (>5V) and the low
voltage profile (<1.5V) respectively, as these two effects were dominant in these regions, with the
141
results shown in Table 3-1. The concentrator LG003 had the best performance, and thus was
paired with the wedged stepped on glass concentrator. This was expected to be the best
performing concentrator, and thus pairing it with the best performing cell would provide the
highest efficiency system. The other two concentrators were mounted to the second and third
best performing cells, which exhibited similar performance to each other.
Table 3-1: The shunt and parasitic series resistances were calculated for the three solar cells integrated into the concentrator modules. Ideally a solar cell will have a high shunt and low series resistance value, so LG003 was the
best of this cell batch.
The full concentrator assemblies were then mounted in a setup designed to allow precise
automated control of the angular pointing and one degree of translational motion. This stage
setup is shown in Figure 3-36. A small aperture was placed in the beam to illuminate a small and
controlled section of the concentrator. Moving the concentrator so that a different portion of the
entrance aperture was illuminated allowed the performance of the guide to be measured as a
function of the distance the light had to propagate from injection before reaching the solar cell, as
seen in Figure 3-35.
Figure 3-35: Scanning the illuminated slit allows determination of how far injected light has propagated within the concentrator before reaching the solar cell at the output face.
43
LG002 LG003 LG005
Series Resistance (Ω) 0.0249 0.0165 0.02
Shunt Resistance (Ω) 388 813 446
142
Figure 3-36: The apparatus for measuring optical efficiency at different sections of the concentrator input aperture. Two rotation axes and a translation axes are automated.
The solar simulator was designed to match the spectral content and angular extent of the
sun. While this allowed for the most accurate approximation of concentrator performance when
used as a solar collector, some additional information could be gained by using a source with a
smaller angular extent. A removable mirror was put into the system to allow the use of a HeNe
laser which was collimated by the off axis parabola. This produced a narrow angular spectrum,
allowing tighter focal spots, and the single wavelength simplified absorption calculations.
143
3.4.2 Acrylic Wedged Stepped Concentrator
After a substantial amount of process development, it was determined that the most
practical way to produce the repeat unit of wedged stepped lenslets was to cast the shapes onto
an available acrylic substrate. This produced a back surface that was of sufficient quality and a
consistent thickness of the lenslet parts, though these parts had a noticeable bow as described in
section 3.2.2. This limited the potential thicknesses of the lenslet arrays depending on the
availability of certain thicknesses of acrylic substrate. The lenslet arrays used in this prototype
were thus approximately 1.6 mm thick (compared to the design thickness of 1.35mm), and thus to
keep the focus of the lenslet arrays on the injection facets, the guide layer had to be reduced in
thickness to approximately 1.80 mm (from a design thickness of 2.1mm.
The monolithic acrylic guide layers could be produced at a controllable thickness, though
the final part thickness was not precise. The parts produced in this way also had a substantial
amount of wedge, which was believed to be due to a mechanical misalignment. Parts were
measured to have up to 200µm in thickness variation along the guiding direction and up to 60µm
variation along the transverse axis. This thickness variation caused the lenslet focal point to vary
over the extent of the guide.
The guide was assembled by hand using an LED light board. An LED bar was placed near
the output face of the guide layer, and each lens array was aligned manually to maximize the light
coming directly up out of the lenslet array. Once aligned, the low index adhesive was cured using
a UV source, fixing the lenslet array in place so the next lenslet array could be aligned. The
completed guide was then mounted in the plastic holder and attached to a solar cell. The optical
efficiency of the guide was measured using a slit aperture at each lenslet array along the guide.
144
Each point was scanned in both rotation angles in order to determine the optical efficiency at the
optimum pointing for each lens array. These results give the optical efficiency falloff for
illumination at increasing distance from the cell, as shown in Figure 3-37.
Figure 3-37: The optical efficiency falloff of the acrylic wedged stepped prototype. A small slit aperture illuminated a small portion of each lenslet array and found the optimum pointing to determine the efficiency falloff.
The optical efficiency was strongly dependent on several parameters, and several
fabrication errors prevented this system from performing as designed. The field of view of the
designed system was approximately ±1.0° square with little variation within the acceptance angle
and a sharp transition as the tight focal spot no longer hits the injection facet. The field of view
measured at the lenslet array closest to the solar cell and lenslet array farthest from the solar cell
were substantially different, as shown in Figure 3-38.
145
Figure 3-38: The field of view of the acrylic wedged stepped concentrator at the front lenslet array (15 mm from cell) and at the rear lenslet array (90mm from cell). The profile is fairly broad and the center moves substantially from
lenslet to lenslet.
The field of view was substantially wider than the designed system and had a shape that
resembles a Gaussian more than a square. This was due to a focal length error. These parts were
of approximately matched thicknesses, but the lenslets varied substantially in thickness and the
guide layer was not exactly matched to compensate. In an otherwise perfect system, this amount
of broadening in the field of view would correspond to a defocus of approximately 190 µm. This
reduced the on-axis efficiency by approximately 11% and caused the field of view to be broadened
into a Gaussian shape of approximately this shape. The broadening of the field of view may also
have been due to a misalignment of the lenslet array such that different lenses had different
optimal pointing angles.
Another error that was known to be reducing the injection efficiency of the guide was the
injection facet roughness. The final master for the wedged stepped concentrator was cut with a
significantly worn tool, and thus there was substantial roughness on the injection facets. There
was variability between a few nanometers of RMS roughness and up to 50nm RMS roughness with
146
no discernible pattern. In addition, each facet had substantial regions that were much better than
other regions on the same facet. Some of this was due to dust or debris, which prevented light
from being injected into the guide, and some of this variation was due to uneven roughness
patches.
The acrylic guide had numerous loss mechanisms which caused it to operate below the
ideal modeled performance. The different loss mechanisms are summarized in Table 3-2. The
modeled results agreed with the concentrator measurements. The facet scattering varied
substantially from facet to facet, and thus an approximate average of the measured facet
roughness was modeled. Some of the scattered light was contained within the concentrator for a
short distance. A substantial portion of this light was propagating at larger angles than from the
designed injection, and thus coupled out after a short propagation distance before reaching the
cell.
147
Table 3-2: The various sources of loss for the acrylic wedged stepped concentrator. The facet scattering was the
dominant loss mechanism, though the defocus caused substantial losses.
From Table 3-2, the dominant loss mechanisms are from the focal effort and the injection
facet scattering. These are targets for improvement of any future iterations of this system. The
injection facet roughness is thought to be largely due to tool wear, and if remastered with an
undamaged tool, it is expected that future iterations of the dimple layer would have much
smoother injection facets. The losses due to the lenslets covering only a portion of the input
aperture are thought to be remedied using known manufacturing processes, and thus of relatively
little concern, despite causing the concentrator to perform poorly as a whole unit.
3.4.3 Polymer on Glass Wedged Stepped Concentrator
The polymer on glass system was expected to have the best optical performance of any
concentrator produced. The glass substrate provided a higher guide index for containing light, a
Loss Mechanism Optical Loss Variance Modelled Efficiency Note
Injection Loss
-Entrance to Guide Layer 6% ±.5% (96±0.5)%Dominant injection loss source in
designed system
-Focal Error 11% ±3% (85±3)%Combination of thickness mismatch
and alignment error
-Facet Scattering 20% -15% to +40% (62 +9/-24)%Symmetric roughness variation
produces asymmetric scattering
Guiding Loss
-Defect Scattering 4% -1% to +5% (59 +9/-23)%Density of debris and scratches on
concentrator
-Scatterd Light Loss 3% ±1% (57 +8/-21)%Light scattered from facet will be
contained for a short distance, then
-Absorption Loss 5% ±1% (54 +8/-20)% Modeled light absorption
Concentrator System
-Sparse Lenslet Aperture 30% -0% to 20% (37 +6/-16)%The lenslets did not cover the whole
guide, and perimiter lenslets low
quality -Field of View Misalignment 52% ±5% (19 +3/-9)%
The lenslet arrays were aligned to
slightly different optimal pointing
Total 81% 3/-9% (19 +3/-9)%
148
lower absorption and scattering, and a flat, high quality surface to attach the other layers to and
provide mechanical stability. As with the acrylic wedged stepped guide, the lenslets could only be
fabricated with thicknesses determined by the available sizes of acrylic sheet to cast them on.
Thus the lenslets that were produced were approximately 1.70 mm thick, which was
approximately 0.35 mm thicker than the design. The glass guide substrate was cut to be 1.70mm
thick to compensate (designed to be 2.0mm), as the dimple layer of initial cast parts was
measured to have 0.05 mm thicker base than in the design specification.
The concentrator was assembled by hand using an LED light board. Light going back
through the concentrator was coupled out the lenslets, and if the system was well aligned, should
be roughly collimated normal to the guide plane. A fused silica plate was used as a weight to hold
down the lenslet array and minimize the bow in the system during alignment. When each lenslet
array was aligned, a UV lamp was focused onto the lenslet array in order to cure the low index
adhesive and securely attach the lenslet to the guide substrate. Once all the lenslet arrays were
attached, the concentrator was secured in a plastic housing, and a silicon solar cell was attached to
the output face.
The concentrator module was mounted in the testing apparatus, and the field of view
taken for a slit aperture at each of the attached lenslet arrays. The optical efficiency at the
optimum pointing angle for varying distances from the solar cell is shown in Figure 3-39.
149
Figure 3-39: The optical efficiency falloff for the polymer on glass wedged stepped guide.
While this was slightly more efficient than the monolithic acrylic version, the efficiency
was much lower than the designed system. The relation between efficiency and distance from the
cell varied with little visible trend over the 6 measured lenslet arrays, and was dominated by a
random fluctuation. From the earlier component inspection, it was determined that this was
consistent with variations in the injection efficiency.
The field of view was taken at each lenslet in order to diagnose errors such as focal length
or misalignments. The field of view resembled the designed system much more closely than for
the monolithic acrylic prototype, as seen in Figure 3-40. The field of view moved substantially
between the various lenslet arrays due to slight misalignments, but consistently appeared to be
approximately ±1°. The field of view had a square shape with sharp transitions and substantial
150
uniformity in the middle region. This eliminated a thickness mismatch and clocking mismatch as
performance drivers.
Figure 3-40: The normalized field of view from a slit aperture of the polymer on glass wedged stepped guide. The field of view much more closely resembled the design specifications than for the other prototypes.
The low index adhesive used to secure the lenslet arrays was designed to adhere to acrylic
materials. A quick test showed that it also adhered strongly to the glass substrate. While this
bond was initially quite strong, it proved to be temporary. After the polymer on glass prototype
was assembled, the bond strength between the low index adhesive and the glass substrate
weakened. After a few weeks, the low index layer with the attached lenslets fell off the glass
substrate in a ribbon. The concentrator was reassembled, but it was noticed that within an hour
or so of assembly, the stress that caused the lenslet arrays to bow had overcome the adhesive
strength. The lenslets were thus bowed, and the edges of the lenslet array had delaminated
completely.
The bow of the lenslet had been measured and fitted as described in section 3.2.2. The
effects of a bowed lenslet array were modeled in LightTools in order to approximate the effect this
151
defect would have on system performance. The model used is shown in Figure 3-41. The system
was modeled as though the lenslet had fully returned to its original shape, and the low index
adhesive was not distorting the bow at all. By varying the thickness of the low index layer, the
approximate amount of delamination could be modeled to determine the effect on the optical
efficiency.
Figure 3-41: The model of the bowed lenslet array. The bowing of the lenslet array was approximated from the unattached lenslet array measured in the lenslet interferometer. The amount of delamination was controlled by
changing the thickness of the low index layer.
The edge of the lenslet array was observed to have delaminated, which would add two
significant Fresnel losses and cause the bow in the system to have substantially more optical
power than if immersed. The design specification for the low index layer thickness was 25µm, but
this would have only immersed a small fraction of the bowed lenslet array. The size of the
observed immersed region was consistent with a low index layer reaching up to 100µm. This
bowing defect caused dramatic losses outside of the immersed region (>30%), but within the
152
observed immersed region, only minor losses were predicted (<10%). The modeled loss at each
lenslet in the array is shown in Figure 3-42 for both the modeled 25µm thickness and the observed
100µm thick low index layers.
Figure 3-42: The optical efficiency of each lenslet in the array for a system with the observed bowing defect and no others. The observed immersed region covers slightly more than half the aperture.
The modeling of the bowed lenslet array suggested that there was a substantial region
near the apex of the bow that should be minimally affected by this error. With the thicknesses of
the lenslet layer and guide layer approximately correct and relatively little clocking observed in the
parts, it was expected that the low injection efficiency was primarily due to facet scattering. The
observed facets were rough enough to account for such a low measured efficiency, but were
largely random and highly variable in roughness. If random facet scattering was indeed the
primary loss driver in the immersed regions, using a much smaller aperture was expected to be
able to find one of the smoother facets in the immersed region of the lenslet layer that had high
injection efficiency.
The slit aperture was reduced to be 3mm x 1mm, which spanned two adjacent lenslets.
The front lenslet was scanned in order to find a “sweet spot” with smooth facets. Such an area
was located near the center of the lenslet array closest to the chip. This spot had an injection
153
efficiency of 89.6% at optimal pointing. The efficiency 2 mm to either side of this spot were found
to be much lower than for this “sweet spot” as shown in Figure 3-43. This indicated a large
efficiency variation within one lenslet array, which would indicate a fairly random process was
driving performance. This was consistent with the observed facet roughness variability.
Figure 3-43: The optical efficiency of the glass wedged stepped design near the sweet spot. The variation in efficiency within the immersed region of a lens array is caused by a substantial randomness in the injection facet roughness.
This sweet spot was further investigated using a collimated HeNe laser instead of the
Xenon source. This allowed a more precise measure of the acceptance angle, and also allowed a
more thorough mapping of the efficiency variation over the two facets of this region. The field of
view was taken for with much smaller steps. The HeNe laser had a much smaller angular spread,
and thus formed a much smaller spot on the injection facet which allowed for much more precise
characterization of the facet variability, as seen in Figure 3-44. The field of view had sharp
transitions and clearly resembles the ±1.0° square. Both images in Figure 3-44 were produced
154
from the same data but with different color scales to highlight the variance on the facet.. There
was still variation of approximately ±5% even within this region, which highlighted facet roughness
as a critical performance driver in these concentrator systems.
Figure 3-44: Detailed field of view of sweet spot measured with collumated HeNe beam. The field of view very closely resembled the designed system, and there is relatively little variation within the injection facet.
The above methods accurately explained the discrepancy between the injection efficiency
of the designed system and the measured system. The large variability of the injection efficiency
obscured the guiding efficiency. This made it difficult to evaluate the performance of the system,
specifically the bypass features. In order to characterize the guiding efficiency of the system, the
injection losses were mitigated as much as possible. The regions of the guide covered by the
lenslet arrays provided many other sources of error, and the best way to roughly characterize the
guiding efficiency was to focus the HeNe beam using an external lens through the regions of the
guide layer that did not have lenslets attached at the front and back of the guide. The
performance of this system gave information about the guiding efficiency of the concentrator.
155
A large number of injection facets were scanned at both the front and back of the guide,
as it was determined that injection efficiency would vary substantially from facet to facet. By
comparing the best facet measured at the back of the guide (farthest from the cell) to the best
facet measured at the front of the guide (closest to the cell), a rough approximation of the guiding
efficiency was established. The measurements shown in Figure 3-45 were normalized to the peak
reading for the facet close to the chip that produced the highest reading. This facet was only 2
mm from the chip, and thus there are not expected to be any guiding losses. The best measured
facet near the back of the guide was 116 mm away from the cell, and thus experienced the guiding
loss from travelling approximately the whole length of the system. Adjacent facets to each
showed the injection efficiency variation, and the substantial uncertainty on what portion of this
was actually guiding loss.
156
Figure 3-45: Laser Scans at the front and back of the glass wedged stepped system. The guiding loss from the best facet near the back of the guide to the best facet found near the front of the guide was 22.7%.
This measurement gave an approximate guiding loss for light injected at the back of the
concentrator of 22.7%. This was approximate, as the best facet was chosen from a set of 12
measured facets near the rear of the guide and 8 facets near the front of the guide. With the
known variation of injection facet roughness, the injection efficiency variation gives this an
uncertainty of approximately ±3%. This measurement was taken after the third reconstruction of
this concentrator, and thus was expected to show substantial wear and dust collection despite
being cleaned during each reconstruction. The adhesion between the dimple layer and the glass
157
substrate was also discovered to be temporary, as substantial portions of the dimple layer had
begun to delaminate. Due to these defects at the time the measurement was made, this
measurement was not expected to show the optimal guiding efficiency for a clean and undamaged
system, but provided an approximation.
Table 3-3: Summary of loss mechanisms for glass wedged stepped concentrator.
The loss mechanisms for the glass wedged stepped concentrator are presented in Table
3-3. The injection loss due to facet roughness is still the primary performance driver for this
system. The guiding efficiency is consistent with the experiments, and should improve for a
cleaner system that has not required repeated reassembly. A “sweet spot” corresponding to a
facet with low surface roughness was found, and measured to have excellent injection efficiency
and a field of view that very closely resembled the designed system.
Loss Mechanism Optical Loss Variance Modeled Efficiency Notes
Injection Loss
-Entrance to Guide Layer 4% ±0.5% (96±0.05)% Fresnel Losses
-Focal Length Error 0% 0% (96±0.05)%The thickness stack in this system we
well matched
-Facet Scattering 20% -15% to +40% (77 +12/-31)%
Random variation in roughness.
Symmetric roughness tolerance yield
asymmetric scattering tolerance.
Guiding Loss
-Defect Scattering 4% -1% to +5% (74 +11/-30)%Density of debris and scratches on
concentrator
-Scattered Light Loss 3% ±1% (72 +11/-29)%Light scattered from facet and
contained will be lost quickly
-Absorption Loss 5% ±1% (68 +11/-28)% Modeled light absorption
Concentrator System
-Sparce Lenslet Aperture 30% -0% to 20% (47 +7/-24)%The lenslets did not cover the whole
face of the guide
-Lenslet Bow 9% ±3% (43+7/-22)%The low index adhesive did not
prevent the lenslet from bowing
-Field of View Misalignment 36% ±5% (28+4/-14)%
The lenslets were aligned by hand,
and thus had different optimum
pointing angles.
Total 72% (14/-4)% (28+4/-14)%
158
3.4.4 Reflective Dimple Tree Concentrator
The reflective tree concentrator was very different from the other two concentrators
fabricated, as this system had a different dimple geometry, lenslet geometry, and a reflective
coating on the lenslets. This system was assembled from the small reflector arrays and monolithic
acrylic guide layer by hand using a method similar to how the other concentrators were assembled.
Due to the reflective coating, the system had to be assembled upside-down with respect to the
other systems. The thinner lenslet array and large flat guide layer allowed lenslet arrays to be
pressed flat when cured.
As with the other concentrators, the efficiency at each lenslet array was measured, and a
field of view was taken to find the optimum pointing angle. These data for the reflective tree
concentrator are shown in Figure 3-46.
159
Figure 3-46: Efficiency falloff for reflective tree concentrator system, taken at optimum pointing angle.
The efficiency for the reflective tree concentrator was much lower than the design
specification, which predicted optical efficiencies above 85%. The field of view of the system at
the front lenslet array and the back lenslet array were measured and presented inFigure 3-47.
This field of view was extremely broad, which was indicative of a large focal error or clocking
misalignment. In order to yield a field of view this broad, the thickness mismatch would have to
be approximately 400um, which was inconsistent with caliper thickness measurements.
160
Figure 3-47: Normalized field of view of acrylic reflective tree concentrator for the lenslet array closest to the cell (left)
and farthest from the cell (right).
In order to determine what was attributable to a focal length mismatch and what was due
to a clocking misalignment, a much smaller aperture was used. This aperture would ensure that
only a few lenslets were illuminated at once, and thus if the field of view was still broad, a focal
length problem was causing the blurred field of view. If the field of view moved systematically as
the illuminated spot moved across the aperture, this was indicative of a clocking misalignment.
The results of this small slit scan showed a substantial defocus problem, but also some clocking
misalignment.
161
Figure 3-48: The small slit scan of the field of view for the front lenslet of the reflective tree concentrator. The field of view is substantially blurred, indicating a focal length error, but the center moves almost a full degree and
systematically, indicating a clocking error.
In order to produce a this variation in the field of view, the lenslet array must be clocked
relative to the concentrator by approximately 0.7°, which was well within the range of a manual
162
alignment. This clocking error was more difficult to eliminate than in the refractive systems due to
the focal length error. During assembly, the LEDs at the output face caused the lenslets to light up
when properly aligned, but with this focal length error, it was more difficult to determine precisely
when the lenslets were aligned. The field of view for this small spot was still blurred enough that a
focal error of approximately 170 µm was estimated. In addition to the issues with focal length and
misalignments, the reflective tree concentrator had the roughest injection facets, which caused
further injection loss. The nature of the reflective system also decreased potential injection
efficiency, as shadowing loss and reflection loss must also be included.
Table 3-4: Summary of loss mechanisms for the reflective tree concentrator prototype.
Loss Mechanism Optical Loss Variance Modeled Efficiency Notes
Injection Loss
-Entrance to Guide Layer 11% ±2% (89±2)%Fresnel losses, shadowing losses, and
reflection loss
-Focal Length Error 15% 3% (76±3)%The thickness stack in this system we
well matched
-Clocking Misalignment 5% 3% (72±4)%System misalignment due to manual
assembly
-Facet Scattering 31% -15% to +40% (48 +11/-29)%
Random variation in roughness.
Symmetric roughness tolerance yield
asymmetric scattering tolerance.
Guiding Loss
-Defect Scattering 4% -1% to +5% (46 +11/-28)%Density of debris and scratches on
concentrator
-Scattered Light Loss 3% ±1% (45 +10/-27)%Light scattered from facet and
contained will be lost quickly
-Absorption Loss 5% ±1% (43 +10/-26)% Modeled light absorption
Concentrator System
-Sparce Lenslet Aperture 30% -0% to 20% (30 +7/-23)%The lenslets did not cover the whole
face of the guide
-Field of View Misalignment 39% ±5% (18+4/-14)%
The lenslets were aligned by hand,
and thus had different optimum
pointing angles.
Total 88% (14/-4)% (18+4/-14)%
163
A summary of the modeled defects and resulting efficiency losses for the reflective tree
concentrator is presented in Table 3-4. While the scattering from the injection facet was still the
dominant loss mechanism, other injection losses such as the focal length error and the clocking
misalignment caused substantially reduced performance.
Three concentrator prototypes were fabricated and characterized. While none of the
concentrators performed as well as the ideal modeled system, the driving errors for each system
were determined, and the relative effect of these imperfections was determined. It was
determined that many of these defects could be substantially reduced in future prototyping
generations. The most important fabrication errors were highlighted, and the potential for
improvements by fixing these errors was shown.
164
4 Conclusion and Future Work
The three concentrator prototypes produced in this project all performed substantially
below the modeled efficiency of each design. The discrepancies between the laboratory
measurements and the modeled performance were determined such that the updated models
matched the performance of the prototypes. This modeling work allowed determination of the
relative importance of the various manufacturing errors and how the concentrator system could
be most effectively improved for future generations of prototypes.
4.1 Improving Concentrator Prototypes
The prototypes described in Chapter 3 were designed as a proof of process for the new
manufacturing methods used throughout this project. These processes produced both dimple and
lenslet geometries that had never been fabricated before. The tolerances and error drivers of
these processes was unknown. Testing of the components and final prototypes showed the
limitations of the manufacturing processes used. The error drivers of the concentrator designs
made using these new processes provided a pathway to produce a concentrator having
dramatically improved optical performance.
In this project, there were several errors that were accepted due to the exploratory nature
of this process that are projected to be corrected in future generations of light guide prototype.
The original goal of the project was to produce six functional prototypes, yet only three were
successfully fabricated. Both two stepped prototypes were abandoned due to a failure while
producing the dimple master. The cause of this cut failure was believed to be known, and thus a
second cut would be expected to produce a master of comparable quality to the other designs.
The polymer on glass tree design was abandoned due to difficulty casting the dimple features onto
165
the glass substrate, but this was again thought to be a matter of refining the process over a period
of time.
For the concentrators that were produced, the two driving errors were related to the
fragmenting of the lenslet array and the roughness of the dimple array. The lenslet array was
produced in small repeating units with the intention of tiling them into a larger unit spanning the
whole prototype. Tiling these repeat units to form a much larger lenslet array required a process
that was known but time consuming and fairly expensive. The demand for alignment of the
lenslet pieces placed a tight tolerance on the tiling process, yet it was expected that future
generations of lenslet parts would be able to span the entire prototype using methods that had
already been developed. This would almost certainly improve the bowing of the lenslet parts, as
the ratio of thickness to area would be greatly reduced. Producing a lenslet array with an
acceptable back surface quality and a precise thickness proved difficult, and future development
work would likely be needed to produce copies of these tiled lenslet arrays with acceptable back
surface quality and a precisely controlled thickness.
The dimple layer masters were produced using experimental machining processes. It was
determined that parts could be made with extremely small fillets (<2µm) and steep draft walls
(<2°). While the process showed the capability of producing surfaces with microroughness of less
than 5nm, many faces showed substantial scratches or other machining artifacts. A substantial
portion of this could be attributed to tool wear. The tools used to cut these masters were used for
several iterations of process development, and many of the parts produced near the end of this
project showed consistent character due to tool wear, as seen in Figure 4-1. This was especially
prevalent on the cuts of the injection facets, where the roughness was measured to be the largest
166
in these parts, and where roughness had the largest scattering effect. It was expected that a new
cutting tool combined with further iterations of machining parameters could dramatically reduce
the scattering of the part both from the injection facet and bypass prisms.
Figure 4-1: A representative injection facet of the two stepped light guide system. The pattern of scratches was
consistent over all facets measured on this part and is thought to be due to tool wear.
The prototypes described in Chapter 3 demonstrate the fabrication process and provide a
path towards a full scale prototype. In addition to the expected improvements described above, it
is important that future work be done to scale the prototypes to a size where they may be cost
effective at concentrations of several hundred times. These systems are designed to tile in order
to form larger concentrators, and thus this work is expected to use better known manufacturing
techniques and require less experimental process development. This work should yield full size
concentrator panels, which must then be integrated into HCPV modules, which requires additional
process development work.
167
4.2 Material Research
The primary goal of this project was to demonstrate the process for producing the
complicated geometries demanded by both the dimple and lenslet layers. The optimization of the
materials these systems were fabricated out of required future development work. Both the
dimple layer of the polymer on glass prototypes and the low index adhesive used for all the
fabricate prototypes were experimental materials. The dimple layer was designed to have to have
the correct refractive index and allow for accurate replication of the dimple geometries, but the
absorption of this layer was not expected to survive highly concentrated radiation. Finding a
material that could accurately reproduce the mastered geometries, possess the correct index of
refraction, bind to the glass substrate, and capable of withstanding high concentrations of solar
radiation required substantial future development work.
The low index adhesive was another experimental material, and while it created a strong
bond to both acrylic and glass parts, it was determined that this bond was only temporary. This
bond must last for many years in harsh environmental conditions, and finding a material that can
survive the expected module life time deployment that still has a low refractive index and strong
adhesion to both lenslet and guide substrates is critical. This requires further materials research,
and the characterization of the lifetime of the bond between guide layer and lenslet layer.
4.2.1 Gradient Index Material Research
Fabricating a gradient index guide layer that can be produced economically in high volumes
is difficult. Conventional ion diffusion processes that raise the index of refraction tend to use
prohibitively expensive reagents such as silver. This added cost has prevented these diffusions
from being commercialized in significant capacity. The most common exchanges involving alkaline
168
ion substitutions lower the index by diffusing sodium ions into glass and replacing lithium ions.
While the Na+ for Li+ exchange is well understood and has been commercialized, the reverse
exchange (replacing Na+ with Li+) has been shown to cause extensive devitrification and cracking.
Gradient Lens Corporation has explored a solution to this problem. The new process
allowed the index to be raised while maintaining a high quality optical surface. The clarity of the
diffused GRIN part with and without this new process is compared to the undiffused part in Figure
4-2.
Figure 4-2: A comparison of the clarity of the undiffused glass (right), the diffused GRIN sample pretreated with K+
ions(center), and the diffused GRIN sample without pretreatment (left). Photo by G. Schmidt
169
The process is designed to balance the stress on the glass during diffusion. This can be
seen through the edge effects where the surface is distorted. A white light interferometer is used
to measure the surface profile at the edge of a 2mm thick slab of glass before and after addition of
the gradient. The undiffused piece shows an edge roll off of approximately 500 nm over a distance
of approximately 100µm. When the gradient is introduced, the surface height profile undergoes a
substantial change but does not deviate more than a micron from the central flat region. The
edge profiles of these three samples are shown in Figure 4-3.
Figure 4-3: Surface heights at the edge of 2mm thick slab of glass that has not been diffused (left), pretreated for 6
hours (center), and pretreated for 16 hours (right). The 2D plots show the vertical and horizontal slices of the data
represented by the corresponding line color.
170
The refractive index profile of the diffused glass sample can be measured by slicing a
section and using an interference microscope to determine the gradient profile. By measuring the
phase difference in the interferometer, the profile can be reconstructed. The index change is
roughly 0.07, which is lower than the optimal Δn of 0.1 or greater. The fringe pattern observed on
the microscope interferometer and the derived index profile are shown in Figure 4-4.
Figure 4-4: Fringe pattern and calculated refractive index profile for diffused GRIN sample.
Producing a glass sample with a stronger gradient economically while keeping the optical
quality is another challenge. The absorption and scattering that result from this gradient must
also be investigated. The GRIN light guide has the highest potential for optical performance, but
requires a substantial amount more development work in order to fabricate.
4.3 Lifetime Durability and Performance Degradation
The lifetime of the concentrator system and the performance degradation when deployed
are critical parameters for determining the viability of a CPV system. There are several critical
failure mechanisms that must be address. The most critical factors for the concentrator of a CPV
171
system are photostability, temperature stability, and mechanical stability. In addition, the optical
performance degradation that results from deployment must be characterized to inform the cost
modeling.
While models have been developed to provide information about the long term
photostability of these concentrator systems, these were not able to be tested in conditions that
mimicked actual deployment. The materials used to fabricate the dimple layer of the polymer on
glass design and the two monolithic acrylic guide layers both were designed to test the machining
capabilities, and the materials were not designed to withstand actual deployment. The
simulations presented in Chapter 2 were for pure PMMA. Neither the experimental polymer used
in the polymer on glass nor the acrylic used for the monolithic parts had an absorption profile that
was expected to withstand high solar flux. The acrylic system showed non-negligible absorption in
the visible due to impurities (the acrylic used was not pure PMMA) and due to incomplete
polymerization.
CPV systems are most likely to be deployed in a desert environment due the large amount
of direct solar radiation these areas receive and the low cost of land. These areas experience large
temperature fluctuations which can be more than 20°C between day and night. The system must
be able to survive the thermal stresses resulting from the different materials of the layers without
delaminating or becoming internally misaligned. In addition, mechanical factors must be
considered such as wind loading and abrasion. These properties require further modeling work
and experimental investigation.
I
Bibliography 1 US Energy Information Administration, Annual Energy Outlook 2011, June 2011 2 “The Drivers of the Levelized Cost of Electricity for Utility-Scale Photovoltaics” Sunpower Corp., 8/14/2008
3 “Annual Energy Outlook 2012: Full Report Reference Case.” AEO2012 Reference Case . June
2012,Washington, DC 4US Energy Information Administration, Annual Energy Outlook 2012, June 2012, DOE/EIA-
0383(2012) 5 Acciona Solar Inc. Image available from http://www.acciona.com/ 6 Enerdata, from EPIA, Observer, SEIA, Downloaded from http://www.enerdata.net/enerdatauk/press-and-
publication/energy-features/solar-photovoltaic-booming-market.php. 3/2013 7 Enerdata, from EPIA, Observer, SEIA, Downloaded from http://www.enerdata.net/enerdatauk/press-and-
publication/energy-features/solar-photovoltaic-booming-market.php. 3/2013 8 “DOE Pursues SunShot Initiative to Achieve Cost Competitive Solar Energy by 2020” United States
Department of Energy Press Release. 2/11/2011. Available at http://energy.gov/articles/doe-pursues-
sunshot-initiative-achieve-cost-competitive-solar-energy-2020 9 Suntech Power Holdings Co., Ltd. Available at http://www.suntech-power.com/en/technology/technology
10 “First Solar Sets CdTe Module Efficiency World Record, Launches Series 3 Black Modul.” FirstSolar Inc.
Press Release 4/9/2013. Available at http://investor.firstsolar.com/releasedetail.cfm?ReleaseID=755244 11
http://investor.firstsolar.com/releasedetail.cfm?releaseid=755244 12
Zytech Low Concentration PV module, www.zytechsolar.com. 13
Hatwaambo, S. “Performance Analysis of Low Concentrating PV-CPC Systems with Structured Reflectors.”
Solar Power, Prof. Radu Rugescu (Ed.). ISBN: 978-953-51-0014-0. InTech. Acailable from:
http://www.intechopen.com/books/solar-power/performance-of-low-concentration-pv-cpc-system-using-
structured-reflectors 14
SunPower Corparation, 2012. “Sunpower C7 Tracker; Exceptional Efficiency and Performance. Available
from us.sunpowercorp.com 15
EMS Transmission 06/10/2010, United Stated Department of the Interior; Bureau of Land Management.
http://www.blm.gov 16
Indra Develops an Innovative High Precision Solar Tracker for Improving the Performance of Photovoltaic
Panels. January 21,2013 Press Release. Indra Corp. 2013. http://www.indracompany.com 17
Green, Martin A., et al. "Solar cell efficiency tables (version 39)." Progress in photovoltaics: research and
applications 20.1 (2011): 12-20. 18
“Sharp Develops Concentrator Solar Cell with World's Highest Conversion Efficiency of 44.4%
Sets New Record with Concentrator Triple-Junction Compound Solar Cell.” Sharp Corporation Press Release
6/14/2013. Available at http://sharp-world.com/corporate/news/130614.html 19
Lee, Kyusang, et al. "Multiple growths of epitaxial lift-off solar cells from a single InP substrate." Applied
Physics Letters 97.10 (2010): 101107-101107. 20
Smith, S.. Shiao, MJ. “Solar PV Balance of System (BOS) Markets: Technologies, Costs and Leading
II
Companies, 2013-2016”. GTM Research Available at
http://www.greentechmedia.com/research/report/solar-pv-bos-2013 21
Notton, Gilles, V. Lazarov, and L. Stoyanov. "Optimal sizing of a grid-connected PV system for various PV
module technologies and inclinations, inverter efficiency characteristics and locations." Renewable
Energy 35.2 (2010): 541-554. 22
Miller, Nicholas, et al. "Utility scale battery energy storage systems." Power and Energy Society General
Meeting, 2010 IEEE. IEEE, 2010. 23
“Younicos and Vattenfall balance short-term power grid fluctuations with large-scale battery in Berlin /
First commercial use in the primary operating reserve market.” Younicos Inc. Press Release 2/2012
http://www.younicos.com/en/media_library/press_releases/011_1MWPCM-Vattenfall.html 24
“ISOFOTON has developed one of the first HCPV plants in Italy” Isofoton Inc. Press Release 1/2013.
Available at http://www.isofoton.com/us/node/2333
25 xxii
Goetzberger, A. "Fluorescent solar energy concentrators: Principle and present state of development."
High-Efficient Low-Cost Photovoltaics. Springer Berlin Heidelberg, 2009. 159-176. 26
Barber, Greg D., et al. "Utilization of Direct and Diffuse Sunlight in a Dye-
Photovoltaic Hybrid Concentrator System." The Journal of Physical Chemistry Letters 2.6 (2011): 581-585 27
Morgan, John Paul. "Light-guide solar panel and method of fabrication thereof." U.S. Patent No.
7,873,257. 18 Jan. 2011. 28
Karp, Jason H., Eric J. Tremblay, and Joseph E. Ford. "Planar micro-optic solar concentrator." Opt.
Express 18.2 (2010): 1122-1133. 29
Stafford, B., Davis, M., Chambers, J., Martinez, M., Sanchez, D. “Solar Tracker Accuracy: Field Experience,
Analysis, And Correlation with Meteorological Conditions”. Green Mountain Engineering.
http://www.greenmountainengineering.com. 30
NREL: Renewable Resource Data Center - Solar Resource Data. 2010 December 29, 2009 [cited
2012 January 15]; Available from: http://www.nrel.gov/rredc/solar_data.html. 31
Unger, B. “Dimpled Planar Lightguide Solar Concentrators.” PhD Dissertation, University of Rochester
2010 32
Unger, B. “Dimpled Planar Lightguide Solar Concentrators.” PhD Dissertation, University of Rochester
2010 33
Unger, B. “Dimpled Planar Lightguide Solar Concentrators.” PhD Dissertation, University of Rochester
2010 34
Wilcox, J., Haas, A., Gray, J., Schwartz, R. “UDMF rev. 9 User Guide“. Purdue University, 2009 35
Unger, B. “Dimpled Planar Lightguide Solar Concentrators.” PhD Dissertation, University of Rochester
2010 36
Surrel, Yves. "Phase stepping: a new self-calibrating algorithm." Applied optics32.19/1 (1993). 37
Bioucas-Dias, José M., and Gonçalo Valadão. "Phase unwrapping via graph cuts." Image Processing, IEEE
Transactions on 16.3 (2007): 698-709. 38
Unger, B. “Dimpled Planar Lightguide Solar Concentrators.” PhD Dissertation, University of Rochester
2010 39
Bennett & Porteus, “Relation Between Surface Roughness and Specular Reflection at Normal
Incidence,”JOSA 51, 123 (1961)
III
40
Martil, I. Gonzalez Diaz, G., “Determination of the dark and illuminated characteristic parameters of a
solar cell from I-V characteristics.” Eur.J.Phys 83 (1992). 41
Pveducation.org. “Double-Diode Model”. http://pveducation.org/pvcdrom/characterisation/double-
diode-model (2013). 42
Chan, Daniel SH, and Jacob CH Phang. "Analytical methods for the extraction of solar-cell single-and
double-diode model parameters from IV characteristics."Electron Devices, IEEE Transactions on 34.2 (1987):
286-293. 43
Unger, B. “Dimpled Planar Lightguide Solar Concentrators.” PhD Dissertation, University of Rochester 2010
Appendix A: Microstructure Dimensions
Reflective Dimple Tree Geometry
A
Fillets were set to be 2μm andsidewall drafts were set at 2°. Theseare consistent with measurements
made on fabricated parts.
Reflective Tree Mirror Pair
Reflective surface is parabolic withradius of curvature of 5.0 mm.
Wedged Stepped Dimple Geometry
Fillets were set to be 2μm and sidewall drafts were setat 2°. These are consistent with measurements made
on fabricated parts.
Stepped System Refractive Lenslet Geometry
Lenslet has conic surface with a radius of curvature of 1.11mmand a conic constant of -0.45 for PMMA lenslet in air.
Stepped System Refractive Lenslet Geometry
Fillets were set to be 2μm and sidewall drafts were set at 2°.These are consistent with measurements made on fabricated
parts.
Appendix B: Layer Dimensions and Materials
Reflective Tree System
Wedged Stepped System
Two-Stepped System
Material Index of Refraction
n vPMMA 1.491 53
GLC Glass 1.536 59
Material Transmission