Ultrafast all-optical techniques for remote - Lirias - KU Leuven

Ultrafast all-optical techniques for remote - Lirias - KU Leuven

Ultrafast all-optical techniques for remote temperature detection and thermophysical property determination Liwang LIU Supervisor: Prof. Dr. Christ ...

4MB Sizes 0 Downloads 4 Views

Recommend Documents

introduction - Lirias - KU Leuven
Reay, D., Crozier, G., & Clayton, J., '“Strangers in paradise?” Working-class students in elite universities', Socio

Katholieke Universiteit Leuven - Lirias - KU Leuven
from the Contract Net protocol. We will add dynamic task assign- ment, synchronization of abort messages and scope handl

CoNTENTS 3 - Lirias - KU Leuven
but also foreign languages. [email protected] • Hong Kong Baptist U • Hong. Kong. Cabrales-Sanchez, Juan Enrique

Towards an Eco-Marxism - Lirias - KU Leuven
fascinating international debate on the development of an Eco-Marxism. ... convincingly shown a reconstructed Eco-Marxis

less vision, more senses - Lirias - KU Leuven
1 GLASS HOUSE 2001 FOR A BLIND MAN, PENEZIĆ & ROGINA ..... representing space in architectural conventions and space in

'Energizing' Relations: Western European - Lirias - KU Leuven
economic alliances. Mao Zedong decided to make a radical change by steering the Chinese economy towards heavy industry t

Relevance of Strategic Management for Universities - Lirias - KU Leuven
This article tries to identify the crucial dimensions of strategic management in universities. To do so, the author look

Voltage-Gated Sodium Channels Are Targets for - Lirias - KU Leuven
as some spider toxins interact with site 4. These sub-. Voltage-Gated Sodium Channels Are Targets for Toxins from the Ve

please scroll down for article - Lirias - KU Leuven
Jul 1, 2009 - Cambridge: Cambridge University Press. Pettigrew, T.F., Jackson, J.S., Ben Brika, J., Lemaine, G., Meerten

KU Leuven
Jan 1, 2016 - Sourire, Django Reinhardt (jazz), Toots. Thielemans, Stromae. • Sports men and women: Eddy Merckx,. Just

Ultrafast all-optical techniques for remote temperature detection and thermophysical property determination

Liwang LIU

Supervisor: Prof. Dr. Christ Glorieux

Examination Board: Prof. Dr. Mark Van der Auweraer Prof. Dr. Michael Wübbenhorst Dr. Helge Pfeiffer

Dissertation presented in partial

Dr. Ling Wang

fulfillment of the requirements for

Prof. Dr. Stéphane Longuemart

the degree of Doctor in Science:

Prof. Dr. Jean-Pierre Locquet

Physics

November 2015

© 2015 KU Leuven, Science, Physics Uitgegeven in eigen beheer, Liwang Liu, JiangSu Province, China.

Alle rechten voorbehouden. Niets uit deze uitgave mag worden vermenigvuldigd en/of openbaar gemaakt worden door middel van druk, fotokopie, microfilm, elektronisch of op welke andere wijze ook zonder voorafgaandelijke schriftelijke toestemming van de uitgever.

All rights reserved. No part of the publication may be reproduced in any form by print, photoprint, microfilm, electronic or any other means without written permission from the publisher.

Acknowledgements I would like to express my sincerest gratitude to many people who made this thesis possible. First and foremost, I want thank my promotor Prof. Christ Glorieux, for supporting me throughout my PhD study, with his patience and profound knowledge. His generous guidance and constant encouragement helped me in all the time of study and writing of this thesis. His advices on both research as well as on my career have been priceless, I feel so lucky to have such a friendly and supportive supervisor. I would also like to thank my thesis committee: Prof. Stéphane Longuemart, Prof. Mark Van der Auweraer, Prof. Michael Wübbenhorst, Dr. Helge Pfeiffer, Dr. Ling Wang, and the chairman Prof. JeanPierre Locquet, for their thoughtful review, insightful comments and suggestions on the improvement of this thesis. The Laboratory of ATF/ZMB has provided the support and equipment that I needed to produce and complete this thesis project and the CSC has funded my studies. I thank my fellow labmates at ATF/ZMB group for all the help on researches and for the joyful time together. Particularly I am grateful to Dr. Salvador Alvarado, Dr. Jichuan Xiong, Dr. Sebastiaan Creten, Dr. Ling Wang, Kuo Zhong, Dr. Tristan Putzeys, and Angeline Kasina for their help and valuable time on photothermal measurements and coating.

I

Besides, I acknowledge Danielle Verachtert, Werner Neefs, Geert Dierckx, Johan Vandamme, Philip De Greef, and Valentijn Tuts from the department for their administrative assistant and technical support during my PhD. My sincere thanks also go to Dr. Eduard Fron, and Dr. Yuliar Firdaus from the research group of Prof. Mark Van der Auweraer, who gave me access to their laboratory and research facilities to perform fluorescence quantum yield and lifetime measurement, and guide me to analyze the data. I am deeply thankful to my family and friends for their encouragement and trust. Special thanks to my wife Yi for her constant support and sacrifice.

II

Abstract

This PhD-project deals with the development and experimental validation of an all-optical thermometry method that enables to detect temperature remotely and with very high temporal resolution (10 nanoseconds, 100 MHz bandwidth). The technique was used to tackle the challenge to investigate the relaxation behavior of the heat capacity of the glassformer glycerol, and thus supply information that can be compared with the relaxation features of other, classically investigated physical properties of glassformers such as the dielectric constant and the mechanical properties. The developed method is exploiting the temperature dependence and rapid response time of the fluorescence characteristics of a probe material. In the investigated application, a transient temperature jump was induced by irradiating the sample with a nanosecond laser pulse, whose energy was optically absorbed and converted into heat. Also a configuration in which the sample was periodically heated by optical excitation was implemented.

III

Nederlandstalig abstract

Dit doctoraatswerk handelt omtrent de ontwikkeling en experimentele validatie van een volledig optische methode die toelaat om temperatuur en veranderingen van temperatuur op afstand te meten met een zeer hoge tijdsresolutie (10 nanoseconden, 100 MHz bandbreedte). De techniek werd gebruikt om een antwoord te bieden op de uitdaging om de relaxatie van de soortelijke warmte van de glasvormer glycerol in kaart te brengen, en aldus informatie te verschaffen die vergeleken kan worden de relaxatie-kenmerken van andere, klassiek onderzochte fysische parameters van glasvormers zoals de diëlektrische constante en de mechanische eigenschappen.

IV

Abbreviations/Symbols

AOM

Acousto-optic modulator

APD

Avalanche photodiode

C

Specific heat capacity

CuCl2

Copper chloride

CW

Continuous wave

D

Translational coefficient

DM

Dichroic mirror

F

Focal length of a lens

FFT

Fourier transform

FL

Fluorescence

FOM

Figure of merit

FWHM

Full width at half maximum

IP NN

Integrated intensity and peak intensity based neural network

IR

Infrared

ISS

Impulsive stimulated scattering

NA

Numerical aperture

NDT

Nondestructive testing

NF

Notch filter

V

NN

Neural network

OBL

Objective lens

OD

Optical density

OSC

Oscilloscope

PL

Photoluminescence

PMT

Photomultiplier tube

PPE

Photopyroelectric

PS

Polystyrene

PT

Platinum

Q

Pulse energy

RhB

Rhodamine B

RTDs

Resistance temperature detectors

SNR

Signal to noise ratio

SP NN

Spectral shape based neural network

SR101

Sulforhodamine-101

TCSPC

Time-correlated single photon counting

TL

Thermal lens

TG

Transient grating

VFT

Vogel–fulcher- tamman

AT

Amplitude of thermal wave

H(t)

Heaviside step function

Ntr

Number of training examples in a neural network

TNN/TNN

Temperatures predicted by neural networks

Ttrue/Ttrue

True temperatures known from calibration measurement

VI

∆T

Upward temperature jump (step)

a

Beam radius

cL

Longitudinal acoustic

∆t

Actual time delay



Thermal diffusivity



Absorption coefficient



Thermal expansion coefficient



Viscosity



Density

c

Characteristic diffusion time

∆T

Phase of thermal wave

VII

Contents Acknowledgements................................................................................................................................................ I Abstract ................................................................................................................................................................... III Nederlandstalig abstract .................................................................................................................................. IV Abbreviations/Symbols ...................................................................................................................................... V Chapter 1 .................................................................................................................................................................. 1 Introduction ............................................................................................................................................................ 1 1.1

Background and state of the art ................................................................................................. 1

1.2

Motivation and research objectives .......................................................................................... 4

1.3

Outline of the dissertation .......................................................................................................... 5

References.................................................................................................................................................... 8 Chapter 2 ............................................................................................................................................................... 15 Temperature dependent fluorescence ....................................................................................................... 15 2.1

Calibration setup for investigation of temperature dependent fluorescence ...................... 16

2.2

Experimental results and discussion ....................................................................................... 17

2.3

Conclusion .................................................................................................................................. 23

References.................................................................................................................................................. 24

IX

Chapter 3 ............................................................................................................................................................... 29 Temperature retrieval by neural network (NN) recognition ............................................................. 29 3.1

Introduction ............................................................................................................................... 29

3.2

NN implementation of fluorescence-based thermometry ..................................................... 30

3.3

Performance of temperature reconstruction by NN .............................................................. 33

3.4

Conclusion ................................................................................................................................... 38

References.................................................................................................................................................. 40 Chapter 4 ............................................................................................................................................................... 43 Development of ultrafast fluorescence based thermometry .............................................................. 43 4.1

Introduction .............................................................................................................................. 43

4.2

CW probe-pulsed pump implementation .................................................................................... 45

4.3

Pulsed probe-pulsed pump implementation .......................................................................... 52

4.4

Conclusion .................................................................................................................................. 61

References.................................................................................................................................................. 63 Chapter 5 ............................................................................................................................................................... 65 Spatially resolved photothermal fluorescence spectroscopy in frequency domain .................. 65 5.1

Experimental setup ..................................................................................................................... 65

5.2

Results and discussion ................................................................................................................. 71

5.3

Conclusion ................................................................................................................................... 74

References.................................................................................................................................................. 75

X

Chapter 6 ............................................................................................................................................................... 77 Lock-in photothermal photoluminescence spectroscopy in frequency domain ......................... 77 6.1

Introduction ............................................................................................................................... 77

6.2

Temperature dependence of the QD photoluminescence ...................................................... 79

6.3

Lock-in detection of photothermal photoluminescence signals............................................ 81

6.4

Experimental results and discussion ....................................................................................... 86

6.5

Conclusion .................................................................................................................................. 89

References.................................................................................................................................................. 91 Chapter 7 ............................................................................................................................................................... 95 Temperature dependent time-resolved photothermal spectroscopy of glycerol ...................... 95 7.1

Introduction ............................................................................................................................... 95



Relaxation dynamics in supercooling systems ............................................................................ 95



Relaxation of specific heat capacity............................................................................................ 97

7.2

Time-resolved thermal lens spectroscopy of supercooled glycerol...................................... 99



Experimental setup...................................................................................................................... 99



Results and discussion .............................................................................................................. 102

7.3

Photothermal fluorescence spectroscopy of supercooled glycerol..................................... 112



Simulation of C() effect on the impulsive temperature response .......................................... 113



Experiment results and discussion ............................................................................................ 118

7.4

Conclusion ................................................................................................................................ 121

References................................................................................................................................................ 122

XI

Chapter 8 ............................................................................................................................................................. 129 General conclusion and perspective.......................................................................................................... 129 8.1

Conclusion ................................................................................................................................ 129

8.2

Perspective............................................................................................................................... 131

Curriculum Vitae............................................................................................................................................... 133 List of publications ........................................................................................................................................... 134

XII

Chapter 1

Introduction

1.1

Background and state of the art

Temperature, one of the fundamental thermodynamic quantities of many-particle systems, plays a key role in the behavior of matter, and its measurement is ubiquitous in scientific research and industrial applications. A variety of techniques [1] have been developed to enable both contact and remote thermometry. Due to the required electrical wiring, contact methods – like thermocouples, thermistors and resistance temperature detectors (RTDs) – are not suited for situations where electromagnetic noise is strong, where the environment is corrosive or where parts are rapidly moving [2, 3]. From the physics point of view, remotely and accurately measuring the temperature in a material on a location of interest, without cross-talk from other variables (chemical composition, physical constitution, and pressure) is an exciting challenge due to the fact that temperature is a statistical quantity that can only be determined by exploiting the temperature dependence of other measurable quantities. Infrared (IR) thermometry, on the basis of Planck radiation, allows for remote temperature measurement to be flexibly performed and it has opened the door for a wide range of possibilities for research and technological applications. However, it requires information of the emissivity of the measured material and cannot probe the temperature of a sample beyond the sample’s surface [4]. Besides, the use of this method is

1

cumbersome in humid conditions due to the strong absorption of the IR light by water vapor in the line of view between the detector and the sample [2, 5]. Together with the accelerating proliferation of modern technology, there is a growing amount and variety of situations that require accurate knowledge of the temperature. The bandwidth of a thermometry technique may not be an important figure of merit (FOM) when studying large systems, in which, due to their large heat capacity, temperature fluctuations happen slowly, so that it is usually sufficient to monitor time-averaged temperature values [6]. However, as the systems of interest for technology and science get smaller, temporal variations become increasingly important and the thermal relaxation time of interest, below which measurements are to be acquired, can be very short. An additional incentive of this thesis for developing wide bandwidth thermometry is the historical interest of the Laboratory for Soft Matter and Biophysics research group [7-13] in the complex relaxation behavior in glass forming liquids: the availability of an all-optical ultrafast thermometry tool would be of particular benefit to enable broadband spectroscopy of the dynamic specific heat capacity. The glass transition has been a subject of intense study for many years, due to the ubiquitous technological applications of glasses and amorphous materials [14] in a variety of fields [15-17], but a full understanding still remains elusive [14, 16, 18]. Upon cooling below the freezing point, a majority of liquids (molecular liquids, polymeric liquids, etc.) form a glass if cooled sufficiently fast to avoid the crystallization [18, 19]. The most apparent feature of the glass transition, as a liquid is supercooled, is the rapid increase of the characteristic relaxation time with decreasing temperature. When a relaxing system is stimulated on a time scale shorter than the relaxation time, then it behaves rigid. When the system is given more time, its molecules get sufficient time to respond cooperatively to stimuli, so that it behaves softly. The relaxation time strongly depends on temperature and has been observed in many different physical quantities that are related to molecular mobility [20, 21], such as the elastic moduli and the dielectric permittivity [22]. The relaxing physical quantity, to which this work is focused, is the specific heat capacity. As it is closely tied with the thermodynamics of the system, the Kauzmann paradox [23] 2

indicates that some sort of transition of specific heat must occur between the liquid and glass [24]. Interestingly, contrary to the dielectric or mechanical susceptibilities [25], the specific heat couples to all the modes in the system. The frequency dependence of the specific heat capacity indicates that some modes need a finite relaxation time to take up or release heat. Specific heat capacity spectroscopy has proven to be a valuable tool for studying the glass transition in various fields [26-34]. In some cases where the important modes do not couple to molecular rotations or density changes, the specific heat might be the only spectroscopic tool available [26]. Up to now, the 3 technique, introduced by [26] Birge and Nagel in 1985 and subsequently used extensively by Nagel and his co-workers [24, 26], has for a long time been the most used method. Presently, the state of the art of the 3 method using multiband heaters, allows one to determine the frequency dependence of the heat capacity over 6 decades [26]. The approach has been complemented by other methods, such as photoacoustic spectroscopy [35, 36] and modulated differential scanning calorimetry [37-39]. A significant breakthrough in the domain of heat capacity spectroscopy of glass forming liquids was achieved by the introduction of photopyroelectric spectroscopy [40, 41], which extended the maximum accessible frequency from 8 kHz [26] to 100 kHz [41]. Since the characteristic frequency in relaxation phenomena evolves quasi-exponentially with (the inverse) temperature, a large bandwidth of experimental methods to study them is crucial to get a complete view on the involved physical processes. One of the questions is whether and to what extent the Arrhenius plots of different relaxing quantities show the same fragility. The fragility of glass formers is a measure for the temperature dependence of the activation energy [42]and thus is one of the quantifiers of the energy landscape that is relevant for the probed modes. An additional stimulus for extending the bandwidth came from the experimental finding [43-45] that the transition cooperativity has an onset in the crossover region, which is important for an understanding of the dynamic glass transition [14, 46, 47]. However, there are only a few glass formers with a crossover frequency in the currently experimentally accessible

3

window [45]. This has been one of the motivations of this work to extend the bandwidth of heat capacity spectroscopy, via an extension of the bandwidth of thermometry.

1.2

Motivation and research objectives

Addressing the points of interest listed in Section 1.1, this PhD-project (i) includes the development of an all-optical thermometry method to enable both remote and ultrafast (broadband) temperature sensing, and to deliver, in the physics laboratory, an experimental proof of concept of the metrology, and thereafter (ii) responds to the limited bandwidth of currently available specific heat capacity spectroscopy approaches, an extension of which would open a new perspective on relaxation spectroscopy in general, and on the physics of wide class of glass forming materials in particular, by complementing classical relaxation techniques such as dielectric and mechanical spectroscopy, and scattering techniques. Being already very well characterized by a plethora of techniques, glycerol [48] is used as material of interest. The chosen method to achieve fast thermometry is by exploiting the temperature dependence and rapid response time of the fluorescence characteristics of a probe material [1, 49] in the context of a photothermal experiment.

In the main configuration of interest, a transient temperature jump at a

detection location is induced by irradiating the sample with nanosecond laser pulse [50-52], whose energy is optically absorbed and converted into heat. In an alternative configuration, periodic temperature oscillations, induced by periodically intensity modulated laser light illumination, are monitored. In summary, the objective of this research is to establish: 1) an all-optical broadband thermometry technique allowing dynamic temperature determination from 0.01 Hz till 100 MHz or more than 9 decades of frequency, 2) a straightforward experimental tool for specific heat capacity spectroscopy of relaxing glass forming materials in a (combined) frequency range from 0.01 Hz till 100 MHz, thus

4

complementing material characterization techniques in frequency ranges or circumstances where other response functions are difficult to determine experimentally, 3) the possibility to compare over a substantial wide frequency range the relaxation of the specific heat (all motional degrees of freedom) with the one of the dielectric permittivity (rotational mobility) and elastic moduli (translational mobility), and hence to address actual questions on the universality of the fragility value between different relaxing quantities, and on the existence of a cusp in the temperature dependence of the relaxation [53].

1.3

Outline of the dissertation

In this thesis, by combining the advantages of the quick response of the fluorescence based thermometry and the stroboscopic approach, the development of an ultrafast all-optical thermometry method, which allows remote temperature detection in a very large temporal bandwidth, from 10 ns to 4 seconds, is reported and demonstrated on the determination of the full temperature evolution induced by a nanosecond laser in a rhodamine B (RhB) dyed glycerol sample. Together with the ultrafast implementation of the fluorescence based thermometry approach, special attention is given to a photothermal application of the fluorescence based thermometry in frequency domain, with possible applications for remote thermal property characterization. Furthermore, the feasibility of using the established ultrafast thermometry to extract the relaxation behavior of the specific heat capacity, thus a broadband specific heat capacity spectroscopy, is explored. In addition, a complementary time-resolved thermal lens [54, 55] spectroscopy approach is implemented to determine the relaxation characteristics of the thermal expansion response to sudden heating, which includes both density and heat capacity relaxation information. The resulting information on the thermal expansion relaxation is compared with literature data on the relaxation of the dielectric permittivity and elastic moduli.

5

By virtue of its high fluorescence quantum yield and substantial temperature dependence of its fluorescence yield [56], RhB has been employed as a marker for thermal sensing in many works [57-59]. Here it is chosen as a dopant in glycerol, to which also CuCl2 is added, in order to increase the IR absorption coefficient of the sample for the wavelength of a 1064 nm pump laser, for later application in the photothermal implementation of the setup. The temperature dependent fluorescence of the sample is investigated in the temperature range between 234 K and 310 K, as discussed in Chapter 2. Chapter 3 addresses the involved inverse problem of temperature retrieval from fluorescence spectrum. A neural network (NN) recognition based calibration approach is developed, and its robustness as a temperature reconstruction tool, which is simultaneously exploiting multiple features of fluorescence spectra as parameters for temperature determination, is verified. A detailed investigation of the application of NN recognition is carried out and different types of NNs, based on the same set of fluorescence spectra recorded at steady state temperatures, by using different compact sets of spectral features for NN inputs are trained and compared. A stroboscopic implementation of fluorescence thermometry, using a pulsed fluorescence evoking probe laser, for remote ultrafast detection of temperature changes resulting from a pulsed laser heating, is described in Chapter 4. In Chapter 5, the position dependence of the calibrated amplitude and phase of photothermally induced temperature oscillations along the axis of a pump laser beam are determined at different modulation frequencies. The spatial and frequency dependence of the extracted temperature signals is fitted by a 1D multi-layer thermal diffusion model.

6

In Chapter 6, besides, a digital lock-in photothermal photoluminescence spectroscopy method is demonstrated on a CdSe/ZnS-polystyrene nanocomposite, spin-coated on a copper substrate, in the framework of possible application of nondestructive (NDT) thermal property determination. The feasibility of two time domain photothermal techniques to extend the frequency range for heat capacity spectroscopy further, till MHz and higher, is explored. In Chapter 7, ultrafast fluorescence thermometry is exploited to monitor the pure temperature response to photothermally induced heating, from which the frequency/time dependent relaxation of specific heat capacity is extracted. In a second approach, the time-resolved thermal lens technique is utilized to determine the relaxation of thermal expansion, from which density and heat capacity relaxation information extracted as complementary. A general conclusion and an outlook of this work are given in Chapter 8.

7

References

1.

Childs, P.R.N., J.R. Greenwood, and C.A. Long, Review of temperature measurement. Review of Scientific Instruments, 2000. 71(8): p. 2959-2978.

2.

Wang, S.P., S. Westcott, and W. Chen, Nanoparticle luminescence thermometry. Journal of Physical Chemistry B, 2002. 106(43): p. 11203-11209.

3.

Fischer, L.H., G.S. Harms, and O.S. Wolfbeis, Upconverting Nanoparticles for Nanoscale Thermometry. Angewandte Chemie-International Edition, 2011. 50(20): p. 4546-4551.

4.

Hartmann, J., P. Voigt, and M. Reichling, Measuring local thermal conductivity in polycrystalline diamond with a high resolution photothermal microscope. Journal of Applied Physics, 1997. 81(7): p. 2966-2972.

5.

Hale, G.M. and M.R. Querry, Optical constants of water in the 200-nm to 200-µm wavelength region. Applied Optics, 1973. 12(3): p. 555-563.

6.

Gasparinetti, S., K.L. Viisanen, O.P. Saira, T. Faivre, M. Arzeo, M. Meschke, and J.P. Pekola, Fast Electron Thermometry for Ultrasensitive Calorimetric Detection. Physical Review Applied, 2015. 3(1).

7.

Glorieux, C., K. Nelson, G. Hinze, and M. Fayer, Thermal, structural, and orientational relaxation of supercooled salol studied by polarization-dependent impulsive stimulated scattering. The Journal of chemical physics, 2002. 116(8): p. 3384-3395.

8.

Leys, J., M. Wübbenhorst, C.P. Menon, R. Rajesh, J. Thoen, C. Glorieux, P. Nockemann, B. Thijs, K. Binnemans, and S. Longuemart, Temperature dependence of the electrical conductivity of imidazolium ionic liquids. The Journal of chemical physics, 2008. 128(6): p. 064509.

8

9.

Bentefour, E.H., C. Glorieux, M. Chirtoc, and J. Thoen, Thermal relaxation of glycerol and propylene glycol studied by photothermal spectroscopy. The Journal of chemical physics, 2004. 120(8): p. 3726-3731.

10.

Sinha, G., C. Glorieux, and J. Thoen, Broadband dielectric spectroscopy study of molecular dynamics in the glass-forming liquid crystal isopentylcyanobiphenyl dispersed with aerosils. Physical Review E, 2004. 69(3): p. 031707.

11.

Glorieux, C., J. Descheemaeker, J. Vandenbroeck, J.-P. Groby, L. Boeckx, P. Khurana, and N. Roozen, Temperature and frequency dependence of the visco-elasticity of a poro-elastic layer. Applied Acoustics, 2014. 83: p. 123-126.

12.

van den Berg, O., W.G. Sengers, W.F. Jager, S.J. Picken, and M. Wübbenhorst, Dielectric and fluorescent probes to investigate glass transition, melt, and crystallization in polyolefins. Macromolecules, 2004. 37(7): p. 2460-2470.

13.

Lupaşcu, V., H. Huth, C. Schick, and M. Wübbenhorst, Specific heat and dielectric relaxations in ultra-thin polystyrene layers. Thermochimica Acta, 2005. 432(2): p. 222-228.

14.

Angell, C.A., K.L. Ngai, G.B. McKenna, P.F. McMillan, and S.W. Martin, Relaxation in glassforming liquids and amorphous solids. Journal of Applied Physics, 2000. 88(6): p. 31133157.

15.

Crowe, J.H., J.F. Carpenter, and L.M. Crowe, The role of vitrification in anhydrobiosis. Annual Review of Physiology, 1998. 60: p. 73-103.

16.

Debenedetti, P.G. and F.H. Stillinger, Supercooled liquids and the glass transition. Nature, 2001. 410(6825): p. 259-267.

17.

Slade, L. and H. Levine, The glassy state phenomenon in food molecules. Glassy State in Foods, 1993: p. 35-101.

18.

Angell, C.A., Formation of Glasses from Liquids and Biopolymers. Science, 1995. 267(5206): p. 1924-1935. 9

19.

Berthier, L. and G. Biroli, Theoretical perspective on the glass transition and amorphous materials. Reviews of Modern Physics, 2011. 83(2): p. 587-645.

20.

Wong, J. and C.A. Angell, Glass: structure by spectroscopy. 1976: M. Dekker New York.

21.

Dixon, P.K., L. Wu, S.R. Nagel, B.D. Williams, and J.P. Carini, Scaling in the Relaxation of Supercooled Liquids. Physical Review Letters, 1990. 65(9): p. 1108-1111.

22.

Yuan, Y.H., L. Zhang, X.L. Wang, X.N. Ying, F. Yan, Y.N. Huang, J.S. Zhu, and Y.N. Wang, Mechanical spectrum study of glass transition by a composite method. Physica B-Condensed Matter, 2009. 404(20): p. 3771-3774.

23.

Kauzmann, W., The Kauzmann Paradox - a Thermodynamic Quandary - a Citation-Classic Commentary on the Nature of the Glassy State and the Behavior of Liquids at Low-Temperatures by Kauzmann,W. Current Contents/Physical Chemical & Earth Sciences, 1993(44): p. 8-8.

24.

Birge, N.O., Specific-Heat Spectroscopy of Glycerol and Propylene-Glycol near the GlassTransition. Physical Review B, 1986. 34(3): p. 1631-1642.

25.

Birge, N.O. and S.R. Nagel, Specific-Heat Spectroscopy of the Glass-Transition. Physical Review Letters, 1985. 54(25): p. 2674-2677.

26.

Birge, N.O., P.K. Dixon, and N. Menon, Specific heat spectroscopy: Origins, status and applications of the 3 omega method. Thermochimica Acta, 1997. 305: p. 51-66.

27.

Bras, A.R., M. Dionisio, H. Huth, C. Schick, and A. Schonhals, Origin of glassy dynamics in a liquid crystal studied by broadband dielectric and specific heat spectroscopy. Physical Review E, 2007. 75(6).

28.

Carpentier, L., O. Bustin, and M. Descamps, Temperature-modulated differential scanning calorimetry as a specific heat spectroscopy. Journal of Physics D-Applied Physics, 2002. 35(4): p. 402-408.

29.

Dixon, P.K., Specific-Heat Spectroscopy and Dielectric Susceptibility Measurements of Salol at the Glass-Transition. Physical Review B, 1990. 42(13): p. 8179-8186. 10

30.

Drulis, M., Low-Temperature Specific-Heat and Ir Spectroscopy of Eud2 and Ybh2. Journal of Alloys and Compounds, 1993. 198(1-2): p. 111-115.

31.

Krause, C., H.J. Yin, C. Cerclier, D. Morineau, A. Wurm, C. Schick, F. Emmerling, and A. Schonhals, Molecular dynamics of a discotic liquid crystal investigated by a combination of dielectric relaxation and specific heat spectroscopy. Soft Matter, 2012. 8(43): p. 11115-11122.

32.

Leyser, H., A. Schulte, W. Doster, and W. Petry, High-Pressure Specific-Heat Spectroscopy at the Glass-Transition in O-Terphenyl. Physical Review E, 1995. 51(6): p. 5899-5904.

33.

Saiter, A., L. Delbreilh, H. Couderc, K. Arabeche, A. Schonhals, and J.M. Saiter, Temperature dependence of the characteristic length scale for glassy dynamics: Combination of dielectric and specific heat spectroscopy. Physical Review E, 2010. 81(4).

34.

Jakobsen, B., N.B. Olsen, and T. Christensen, Frequency-dependent specific heat from thermal effusion in spherical geometry. Physical Review E, 2010. 81(6).

35.

Büchner, B. and P. Korpiun, Frequency-dependent specific heat near the glass transition measured with the photoacoustic effect. Applied Physics B, 1987. 43(1): p. 29-33.

36.

Kojima, S., Study of Phase Transitions by Frequency Dependent Photoacoustic Measurements, in Physical Acoustics, O. Leroy and M. Breazeale, Editors. 1991, Springer US. p. 399-404.

37.

Lu, Z.P., Y. Li, S.C. Ng, and Y.P. Feng, Study of glass transition of metallic glasses by temperature-modulated differential scanning calorimetry (MDSC). Thermochimica Acta, 2000. 357: p. 65-69.

38.

Boller, A., C. Schick, and B. Wunderlich, Modulated differential scanning calorimetry in the glass transition region. Thermochimica Acta, 1995. 266: p. 97-111.

39.

Song, M., A. Hammiche, H.M. Pollock, D.J. Hourston, and M. Reading, Modulated Differential Scanning Calorimetry .1. A Study of the Glass-Transition Behavior of Blends of Poly(Methyl Methacrylate) and Poly(Styrene-Co-Acrylonitrile). Polymer, 1995. 36(17): p. 3313-3316.

11

40.

Bentefour, E.H., C. Glorieux, M. Chirtoc, and J. Thoen, Thermal relaxation of glycerol and propylene glycol studied by photothermal spectroscopy. Journal of Chemical Physics, 2004. 120(8): p. 3726-3731.

41.

Bentefour, E.H., C. Glorieux, M. Chirtoc, and J. Thoen, Broadband photopyroelectric thermal spectroscopy of a supercooled liquid near the glass transition. Journal of applied physics, 2003. 93(12): p. 9610-9614.

42.

Martinez, L.-M. and C. Angell, A thermodynamic connection to the fragility of glass-forming liquids. Nature, 2001. 410(6829): p. 663-667.

43.

Beiner, M., H. Huth, and K. Schroter, Crossover region of dynamic glass transition: general trends and individual aspects. Journal of Non-Crystalline Solids, 2001. 279(2-3): p. 126-135.

44.

Huth, H., M. Beiner, and E. Donth, Temperature dependence of glass-transition cooperativity from heat-capacity spectroscopy: Two post-Adam-Gibbs variants. Physical Review B, 2000. 61(22): p. 15092-15101.

45.

Huth, H., M. Beiner, S. Weyer, M. Merzlyakov, C. Schick, and E. Donth, Glass transition cooperativity from heat capacity spectroscopy - temperature dependence and experimental uncertainties. Thermochimica Acta, 2001. 377(1-2): p. 113-124.

46.

Fujara, F., B. Geil, H. Sillescu, and G. Fleischer, Translational and rotational diffusion in supercooled orthoterphenyl close to the glass transition. Zeitschrift für Physik B Condensed Matter, 1992. 88(2): p. 195-204.

47.

Beiner, M., S. Kahle, E. Hempel, K. Schroter, and E. Donth, Crossover region of dynamic glass transition in poly(n-hexyl methacrylate) by heat capacity spectroscopy. Macromolecules, 1998. 31(25): p. 8973-8980.

48.

Davidson, D. and R. Cole, Dielectric relaxation in glycerol, propylene glycol, and n‐propanol. The Journal of chemical physics, 1951. 19(12): p. 1484-1490.

12

49.

Lou, J.F., T.M. Finegan, P. Mohsen, T.A. Hatton, and P.E. Laibinis, Fluorescence-based thermometry: Principles and applications. Reviews in Analytical Chemistry, 1999. 18(4): p. 235284.

50.

Liu, Y., D. Cheng, G. Sonek, M. Berns, C. Chapman, and B. Tromberg, Evidence for localized cell heating induced by infrared optical tweezers. Biophysical journal, 1995. 68(5): p. 2137-2144.

51.

Huang, X., P.K. Jain, I.H. El-Sayed, and M.A. El-Sayed, Plasmonic photothermal therapy (PPTT) using gold nanoparticles. Lasers in medical science, 2008. 23(3): p. 217-228.

52.

Zondervan, R., F. Kulzer, H. van der Meer, J.A.J.M. Disselhorst, and M. Orrit, Laser-driven microsecond temperature cycles analyzed by fluorescence polarization microscopy. Biophysical Journal, 2006. 90(8): p. 2958-2969.

53.

Yang, Y. and K.A. Nelson, Impulsive stimulated light scattering from glass‐forming liquids. I. Generalized hydrodynamics approach. The Journal of chemical physics, 1995. 103(18): p. 77227731.

54.

Dovichi, N. and J. Harris, Time-resolved thermal lens calorimetry. Analytical Chemistry, 1981. 53(1): p. 106-109.

55.

Snook, R.D. and R.D. Lowe, Thermal lens spectrometry. A review. Analyst, 1995. 120(8): p. 2051-2068.

56.

Kubin, R.F. and A.N. Fletcher, Fluorescence Quantum Yields of Some Rhodamine Dyes. Journal of Luminescence, 1982. 27(4): p. 455-462.

57.

Coppeta, J. and C. Rogers, Dual emission laser induced fluorescence for direct planar scalar behavior measurements. Experiments in Fluids, 1998. 25(1): p. 1-15.

58.

Ross, D., M. Gaitan, and L.E. Locascio, Temperature measurement in microfluidic systems using a temperature-dependent fluorescent dye. Analytical Chemistry, 2001. 73(17): p. 4117-4123.

13

59.

Romano, V., A.D. Zweig, M. Frenz, and H.P. Weber, Time-Resolved Thermal Microscopy with Fluorescent Films. Applied Physics B-Photophysics and Laser Chemistry, 1989. 49(6): p. 527533.

14

Chapter 2

Temperature dependent fluorescence

The temperature-sensitive nature of fluorescence provides a non-contact, all-optical, and emissivityindependent approach to determine the temperature of a system of interest [1]. Fluorescence-based thermometry circumvents many of the limitations encountered in other traditional thermometry techniques [2] that are based on thermoelectric devices, electrical resistance based probes, emitted infrared (IR) radiation, etc. As a consequence, fluorescence-based thermometry is receiving increasing attention in the framework of various kinds of scientific research and industrial processes [3-10] .

Rhodamine B (RhB) has been employed as a marker for thermal sensing in many works [7, 11, 12], by virtue of its high fluorescence quantum yield and substantial temperature dependence of its fluorescence[13]. In this work, RhB was chosen as a dopant in the polar compound of interest, glycerol, in view of its high solubility in a polar solvent, and its considerable photochemical stability to visible photoexcitation [14]. Semiconducting quantum dots (QDs) which have been used for accurate thermal sensing in the past [15, 16], could also be an interesting candidate, although special surface functionalization with water-soluble ligands [17, 18] is required to make QDs soluble in glycerol, which may cause some other complications. In view of a particular photothermal implementation of the setup, also CuCl2 was added, in order to increase the IR absorption coefficient of the sample for the wavelength of a pump laser (1064 nm). The absorption coefficient of the sample at 1064 nm was measured by optical 15

transmission method and found to be IR=0.50±0.01 Neper.mm-1 at room temperature, compared to 0.01 Neper.m-1 for the sample without CuCl2. In this Chapter, the temperature dependent fluorescence of the sample excited by a CW-532 nm laser is investigated, and the underlying photophysics responsible for the effects of temperature is discussed as well. 2.1

Calibration setup for investigation of temperature dependent fluorescence

liquid N2

PC GPIB

trigger spectrometer

temperature controller

vacuum pump

fiber

excitation laser 532 nm

delay generator

PT1000

_s

thermocouple

cryostat filter sample cell

mirror

lens

Figure 2.1

Calibration setup. The sample cell was placed inside a vacuum optical cryostat. A CW 532-nm laser was used to evoke fluorescence in the sample. The fluorescence emission spectra were recorded by a spectrometer. The sample temperature was measured by a combination of a platinum resistor (PT1000) and a type T thermocouple.

The temperature dependence of the fluorescence spectrum of the sample was investigated by using the setup shown in Figure 2.1. To excite the fluorescence of RhB, a CW 532-nm laser (Samba 100, Cobolt ®) was used and slightly focused to the sample by a lens with focal length of 300 mm. The entrance beam diameter to this lens was 2.5 mm, while the spot diameter at focal point was approximately 128 m via

16

2F·/d with F=300 mm the focal length, nmthe wavelength, and d=2.5 mm the entrance beam diameter. The emitted fluorescence was collected by a lens system with numerical aperture NA= 0.22 and sent via fiber to a USB 4000 spectrometer (Ocean Optics®), where the spectrum was recorded. The sample cell was placed inside an optical cryostat (Optistat-DN-V, Oxford Instruments®), where the temperature could be regulated by an accompanying temperature controller (ITC 503, Oxford Instruments®). In order to determine possible temperature differences between the cold finger and the sample holder (inset, Figure 2.1), a resistive PT1000 thermometer (read by an HP 34401A® multimeter) was glued to the sample holder, monitoring its temperature, while a type T thermocouple, attached nearby the region of interest at the cuvette wall, was used to measure the temperature difference between the temperature in the optical detection region and the sample holder. In this way, information on the temperature in the region of interest in the sample, which was needed for the calibration of fluorescence thermometry, was continuously available. The concentrations of RhB and CuCl2 were 5.4×10-4 molar and 1.4×10-1 molar respectively. The optical power of the probe laser evoking the fluorescence was 3 mW.

2.2

Experimental results and discussion

It is noteworthy that photobleaching, the rapid loss of fluorescence due to the photochemical destruction of fluorophores upon exposure to an excitation light source, of RhB fluorophore has been reported in many works [19-21]. It was thus of importance to check, prior to performing fluorescence-based thermometry experiments, for possible photobleaching of the system. This was carried out by continuously illuminating the sample by probe laser light and monitoring the fluorescence intensity for a long time (60 minutes) at room temperatures.

17

Figure 2.2

Photobleaching rate of the RhB dyed sample at room temperature under continuous illumination of the fluorescence evoking probe laser. The probe laser intensity was 23 W∙cm-2 and the sample was sealed in a cuvette in a vacuum cryostat.

A typical time evolution is presented in Figure 2.2. The integrated fluorescence intensity over time was normalized to an initial value, one minute after switching on the probe laser. In the investigated system, the intensity decay was negligible, 0.5% in 60 minutes. This is because the used laser intensity was moderate (23 W∙cm-2) and the sample was sealed in a cuvette that was placed in a vacuum cryostat (oxygen-free). After about 45 minutes, recovery of the fluorescence signal from the small decrease occurred, probably as a consequence of translational diffusion of the RhB molecules, which lead to refreshing the fluorophores in the illuminated zone. For three-dimensional (3D) isotropic diffusion, the translation time (T) of a molecule can be estimated by

a2 T  6D

Eq. 2.1

with a the radius of the spot (64 m in this work) through which the molecules diffuse. D is the translational coefficient defined by Stoke-Einstein equation: 18

D

k BT 6 R

Eq. 2.2

where kB is the Boltzmann’s constant (1.38×10-23 J∙K-1), T is the absolute temperature (room temperature in this work, 293 K), is the viscosity (1.3 Pa∙s [22]), and R is the hydrodynamic molecular radius (around 0.57 nm for RhB [23] ). By substituting those values into Eq. 2.1 and Eq. 2.2, the translation time is estimated at 39 minutes, close to the 45 minutes duration observed in Figure 2.2. Fluorescence spectra of the sample were collected at 11 different steady state temperatures in the range from 234 K to 311 K. The temperature dependence of the fluorescence spectrum of the sample is summarized in Figure 2.3a. Both the integrated and peak (left axis, inset) intensities evolve in a nonmonotonic way, with a maximum around 288 K, as shown in Figure 2.3b. The observed decrease in florescence intensity, above 288 K, with rising temperature can be attributed to a decrease of the quantum yield of RhB with temperature. This decrease is due to an increase of the rate constants for internal conversion or intersystem crossing, which compete with the fluorescence when the temperature is increased [1]. The opposite evolution below 288 K might be due to thermally activated recovery from the long-lived dark state, the radical anion of RhB [24, 25], as a result of the electron transfer from glycerol to the excited RhB. However, the ratio (solid line) of the integrated intensity to the peak intensity increases as the temperature increases (right axis, inset) due to the emission bands becoming narrower towards low temperatures. Also the shape, width, and maximum position of the spectrum turn out to be strongly temperature dependent.

19

4 3 2

1000

Figure 2.3

580

600 620 640 wavelength (nm)

(b)

200

600

190

5 180

4 3

1 0560

800

170 240

660

ratio (a.u.)

intensity (a.u.)

5

234.4 K 242.1 K 250.0 K 257.9 K 265.6 K 273.2 K 280.7 K 288.2 K 295.7 K 303.0 K 311.2 K

(a)

intensity (a.u.)

6

260 280 300 temperature (K)

320

Fluorescence spectra measured at different temperatures (a) and (b) temperature dependence of the integrated intensity (solid squares), the peak intensity (hollow squares) and the ratio of the integrated intensity to the peak intensity (stars, right axis).

In order to highlight the temperature dependence of the spectral shape, in Figure 2.4a, right axis, the spectra in Figure 2.3a were normalized to their peak value. One normalized absorbance spectrum representative of the dilute solution of this sample was measured at room temperature (293 K) by a UV– VIS spectrometer. The results were shown in the left axis of Figure 2.4a, with the absorption maximum around 563 nm. The emission maximum around 595 nm monotonically shifts to shorter wavelengths with decreasing temperature, as shown in the Figure 2.4b (left axis). The full width at half maximum (FWHM) is broadening (right axis) as the temperature is increasing, which is consistent with the behavior of the ratio of the integrated intensity to the peak intensity depicted in Figure 2.3b. The blue-shift covers 9 nm over a temperature range of nearly 80 K.

This shift occurs because the rate of relaxation of the

fluorophores to a solvent is strongly dependent on temperature because of the solvent viscosity increasing dramatically with decreasing temperature [22], as depicted in Figure 2.5. At high temperatures, the relaxation time is much shorter than the fluorescence decay time. However, at low temperatures the relaxation time can no longer be neglected compared to the fluorescence decay time (due to increased solvent viscosity), and fluorescence from non-relaxed solvent configurations occurs, leading to a blue

20

shift of the emission spectrum. Briefly speaking, the increase of viscosity with cooling leads to increasing

0.8 0.6 530 nm

0.4 0.2 normalized absorbance

0.0 450 540

Figure 2.4

600

620

640

wavelength (nm)

660

1.0 0.8 0.6 0.4 0.2

604

40

(b)

602

38

600

36

598

34

596

32

594

30

592

0.0 680

FWHM (nm)

234.4 K 242.1 K 250.0 K 257.9 K 265.6 K 273.2 K 280.7 K 288.2 K 295.7 K 303.0 K 311.2 K

(a)

563 nm

emission maximum (nm)

1.0

nomalized intensity (a.u.)

normalized absorbance (a.u.)

fluorescence contributions of non-relaxed states and thus blue-shift [26, 27].

240

260 280 300 temperature (K)

320

28

Normalized fluorescence spectra measured at different temperatures (right axis), and normalized absorbance spectrum measured at 293 K (left axis). (b) Redshift of the emission maximum (solid circles, left axis) of the fluorescence spectra, and broadening (hollow circles, right axis) of the FWHM as the temperature increases.

Besides the main peak, the spectra seem to contain an underlying lower and broader peak with a maximum around 640 nm, which becomes more pronounced with decreasing temperature. The emission band, around 640 nm, corresponds with the absorption band around 530 nm, the latter being shown in the left axis of Figure 2.4a. This can indicate a vibrational progression of an involved transition from the zeroth vibrational level of the excited state to the first vibrational level of the ground state (a difference between 595 and 640 nm corresponds to 1182 cm-1, which is close to the distance between the 0-0 and 0-1 transitions in the absorption spectra 1106 cm-1). In this case the more pronounced 640 nm band at low temperatures is merely due to the fact that bands become narrower at low temperatures. However, also effects of dimerization cannot be excluded [28-31], as the concentration of the RhB in the strongly hydrogen bonding solvent is quite high. In this case the emission band around 640 nm can be due to dimers. Decreasing the temperatures will on one hand increase the dimerization and will on the other hand

21

increase the fluorescence quantum yield of the dimers, hence increasing the importance of the 640 nmband at lower temperatures. Also the possible temperature dependent absorbance of the sample at 532 nm could also be a reason. It is at present difficult to determine to which extent each of both scenarios play a role. In order to reduce the complications of the sample system, the doping of RhB and CuCl2 in following photothermal applications were reduced to 2×10-6 molar and 0.1 molar respectively, which on one hand still yields sufficient fluorescence light for detection, and on the other hand allows moderate absorption of the sample to the IR pump laser (1064 nm). 7

10

5

viscosity (Pa s)

10

3

10

1

10

-1

10

Figure 2.5

210 225 240 255 270 285 300 315 temperature (K)

Temperature dependent viscosity of glycerol calculated by Eq. 2.3 with parameters list in Table 2.1 [22] .

log10   log10 0 

22

B T  T0

Eq. 2.3

Temperature range (K)

log10 log10(Pa∙s)

B (K)

T0 (K)

213-283

-7.1

1260

118

283-323

-5.5

780

153

Table 2.1

2.3

VFT (Eq. 2.3) parameters for viscosity of glycerol [17]

Conclusion

An experimental setup, consisting of fluorescence excitation, temperature control, and fluorescence spectrum acquisition, has been established to study the temperature dependent fluorescence emission and as well the photobleaching rate of the system. The temperature dependence of fluorescence of RhB in a mixture of CuCl2 and glycerol in the temperature range between 234 and 311 K was studied. Measuring fluorescent intensity is straightforward, but in practical its use in the determination of temperature can be hampered by the possibility of the measured fluorescence intensity being affected by factors such as concentration, excitation intensity, optical configuration, and detector efficiency. On the contrary, however, the spectral features like FWHM, emission maximum, and shape, are independent of the laser power fluctuation and alignment drifts of the optical elements. They can therefore be used as a reliable source of information for temperature extraction.

23

References 1.

Lou, J.F., T.M. Finegan, P. Mohsen, T.A. Hatton, and P.E. Laibinis, Fluorescence-based thermometry: Principles and applications. Reviews in Analytical Chemistry, 1999. 18(4): p. 235284.

2.

Childs, P.R.N., J.R. Greenwood, and C.A. Long, Review of temperature measurement. Review of Scientific Instruments, 2000. 71(8): p. 2959-2978.

3.

Gao, L., C. Zhang, C.Y. Li, and L.H.V. Wang, Intracellular temperature mapping with fluorescence-assisted photoacoustic-thermometry. Applied Physics Letters, 2013. 102(19).

4.

Okabe, K., N. Inada, C. Gota, Y. Harada, T. Funatsu, and S. Uchiyama, Intracellular temperature mapping with a fluorescent polymeric thermometer and fluorescence lifetime imaging microscopy. Nature Communications, 2012. 3.

5.

Walker, G.W., V.C. Sundar, C.M. Rudzinski, A.W. Wun, M.G. Bawendi, and D.G. Nocera, Quantum-dot optical temperature probes. Applied Physics Letters, 2003. 83(17): p. 3555-3557.

6.

Valerini, D., A. Creti, M. Lomascolo, L. Manna, R. Cingolani, and M. Anni, Temperature dependence of the photoluminescence properties of colloidal CdSe/ZnS core/shell quantum dots embedded in a polystyrene matrix. Physical Review B, 2005. 71(23).

7.

Romano, V., A.D. Zweig, M. Frenz, and H.P. Weber, Time-Resolved Thermal Microscopy with Fluorescent Films. Applied Physics B-Photophysics and Laser Chemistry, 1989. 49(6): p. 527533.

8.

Aigouy, L., G. Tessier, M. Mortier, and B. Charlot, Scanning thermal imaging of microelectronic circuits with a fluorescent nanoprobe. Applied Physics Letters, 2005. 87(18).

24

9.

Maestro, L.M., P. Haro-Gonzalez, J.G. Coello, and D. Jaque, Absorption efficiency of gold nanorods determined by quantum dot fluorescence thermometry. Applied Physics Letters, 2012. 100(20).

10.

Saidi, E., B. Samson, L. Aigouy, S. Volz, P. Low, C. Bergaud, and M. Mortier, Scanning thermal imaging by near-field fluorescence spectroscopy. Nanotechnology, 2009. 20(11).

11.

Coppeta, J. and C. Rogers, Dual emission laser induced fluorescence for direct planar scalar behavior measurements. Experiments in Fluids, 1998. 25(1): p. 1-15.

12.

Ross, D., M. Gaitan, and L.E. Locascio, Temperature measurement in microfluidic systems using a temperature-dependent fluorescent dye. Analytical Chemistry, 2001. 73(17): p. 4117-4123.

13.

Kubin, R.F. and A.N. Fletcher, Fluorescence Quantum Yields of Some Rhodamine Dyes. Journal of Luminescence, 1982. 27(4): p. 455-462.

14.

Watanabe, T., T. Takizawa, and K. Honda, Photocatalysis through excitation of adsorbates. 1. Highly efficient N-deethylation of rhodamine B adsorbed to cadmium sulfide. The Journal of Physical Chemistry, 1977. 81(19): p. 1845-1851.

15.

Maestro, L.M., E.M. Rodríguez, F.S. Rodríguez, M.I.-d. la Cruz, A. Juarranz, R. Naccache, F. Vetrone, D. Jaque, J.A. Capobianco, and J.G. Solé, CdSe quantum dots for two-photon fluorescence thermal imaging. Nano letters, 2010. 10(12): p. 5109-5115.

16.

Li, S., K. Zhang, J.-M. Yang, L. Lin, and H. Yang, Single quantum dots as local temperature markers. Nano letters, 2007. 7(10): p. 3102-3105.

17.

Larson, D.R., W.R. Zipfel, R.M. Williams, S.W. Clark, M.P. Bruchez, F.W. Wise, and W.W. Webb, Water-soluble quantum dots for multiphoton fluorescence imaging in vivo. Science, 2003. 300(5624): p. 1434-1436.

18.

Debruyne, D., O. Deschaume, E. Coutiño-Gonzalez, J.-P. Locquet, J. Hofkens, M.J. Van Bael, and C. Bartic, The pH-dependent photoluminescence of colloidal CdSe/ZnS quantum dots with different organic coatings. Nanotechnology, 2015. 26(25): p. 255703. 25

19.

Low, P., B. Kim, N. Takama, and C. Bergaud, High-spatial-resolution surface-temperature mapping using fluorescent thermometry. Small, 2008. 4(7): p. 908-914.

20.

Natrajan, V.K. and K.T. Christensen, Two-color laser-induced fluorescent thermometry for microfluidic systems. Measurement Science & Technology, 2009. 20(1).

21.

Glawdel, T., Z. Almutairi, S. Wang, and C. Ren, Photobleaching absorbed Rhodamine B to improve temperature measurements in PDMS microchannels. Lab on a Chip, 2009. 9(1): p. 171174.

22.

Schroter, K. and E. Donth, Viscosity and shear response at the dynamic glass transition of glycerol. Journal of Chemical Physics, 2000. 113(20): p. 9101-9108.

23.

Gendron, P.O., F. Avaltroni, and K.J. Wilkinson, Diffusion Coefficients of Several Rhodamine Derivatives as Determined by Pulsed Field Gradient-Nuclear Magnetic Resonance and Fluorescence Correlation Spectroscopy. Journal of Fluorescence, 2008. 18(6): p. 1093-1101.

24.

Zondervan, R., F. Kulzer, S.B. Orlinskii, and M. Orrit, Photoblinking of rhodamine 6G in poly(vinyl alcohol): Radical dark state formed through the triplet. Journal of Physical Chemistry A, 2003. 107(35): p. 6770-6776.

25.

Zondervan, R., F. Kulzer, H. van der Meer, J.A.J.M. Disselhorst, and M. Orrit, Laser-driven microsecond temperature cycles analyzed by fluorescence polarization microscopy. Biophysical Journal, 2006. 90(8): p. 2958-2969.

26.

Birks, J.B., Organic Molecular Photophysics. 1975, New York: J. Wiley.

27.

Lakowicz, J.R., Principles of fluorescence spectroscopy. 3 ed. 2009: Springer.

28.

Ballet, P., M. VanderAuweraer, F.C. DeSchryver, H. Lemmetyinen, and E. Vuorimaa, Global analysis of the fluorescence decays of N,N'-dioctadecyl rhodamine B in Langmuir-Blodgett films of diacylphosphatidic acids. Journal of Physical Chemistry, 1996. 100(32): p. 13701-13715.

26

29.

Pevenage, D., M. Van der Auweraer, and F.C. De Schryver, Influence of the molecular structure on the lateral distribution of xanthene dyes in Langmuir-Blodgett films. Langmuir, 1999. 15(24): p. 8465-8473.

30.

Vanderauweraer, M., B. Verschuere, and F.C. Deschryver, Absortion and Fluorescence Properties of Rhodamine-B Derivatives Forming Langmuir-Blodgett Films. Langmuir, 1988. 4(3): p. 583-588.

31.

Vuorimaa, E., H. Lemmetyinen, M. VanderAuweraer, and F.C. DeSchryver, Photochemical processes in mixed Langmuir-Blodgett films of rhodamine B and anthracene. Thin Solid Films, 1995. 268(1-2): p. 114-120.

27

Chapter 3

Temperature retrieval by neural network (NN) recognition

This Chapter addresses the inverse problem of temperature retrieval from the fluorescence spectrum. A neural network (NN) recognition based calibration approach is developed. Detailed investigation of the application of NN recognition is performed and different types of NNs by using different compact sets of spectral features for NN inputs are trained and compared. 3.1

Introduction

Fluorescence-based thermometry utilizes the temperature dependence of the fluorescence emission of thermographic phosphors (organic dyes, inorganic phosphors and quantum dots [1-15]) to determine the temperature [16]. Fluorescence-based thermometry methods can be classified into several different categories on the basis of the characteristic parameter of fluorescence that is utilized for temperature readout, e.g., intensity, band-shape, spectral position, polarization, lifetime, and bandwidth of fluorescence. Intensity-based methods, involving the measurement of the fluorescence integrated intensity and/or peak intensity, require relatively simple instrumentation and data analysis, and therefore, tend to be the most cost-effective approach to implement [13]. The accuracy of the acquired temperature, however, is hampered by the fluorescence intensity being also affected by mechanical, electrical, and optical instabilities of the measurement arrangement, such as drifts of optical element positions and fluctuations

29

of the power of the probe laser used to evoke the fluorescence [7]. In order to partially bypass such issues, ratiometric/dual-color fluorescence thermometry [11, 17-22] based on the intensity ratio of fluorescence at two selected emission bands and in some cases from two different types of fluorophores, or based on the fluorescence anisotropy, extracted from the fluorescence intensities of two orthogonally polarized components, has been proposed. In this work, neural network (NN) recognition, a powerful tool for solving inverse problems and retrieving information of interest from experimental data in many applications [23-25], has been explored to provide an alternative approach to improve the measurement accuracy of fluorescence-based thermography [26]. This was accomplished by exploiting multiple features of fluorescence spectra, as discussed in the previous Chapter, as parameters for temperature determination, resulting in a robust information extraction algorithm. 3.2

NN implementation of fluorescence-based thermometry

Fluorescence-based thermometry requires a calibration procedure to link the employed spectral information and the corresponding temperature. In literature, this is mostly done by a polynomial fit of the curve of the temperature dependent fluorescence intensity, decay time, or line shift. In this following, we take an alternative approach on the basis of the NN recognition algorithm reported in Ref. [26]. The idea is to exploit the temperature information in the combination of different features of fluorescence spectra simultaneously, rather than being limited to only one correlation. In this way, we aim to improve the temperature retrieval performance. In NN recognition, parameters of an ad hoc-chosen NN function are optimized in order for the function to adequately convert numerical input information to a numerical output, namely the parameter to extract. Provided by enough experimental data, NN is capable of capturing linear and/or non-linear relationship between input and output variable(s) without requiring explicit mathematical representations or a priori models, i.e., here the complicated temperature dependent

30

fluorescence.

The optimization is based on a large set of training examples of NN input values to be

recognized, and their corresponding temperature. In the case of fluorescence-based thermometry, the numerical fluorescence information serves as NN input, while the NN output yields the sample temperature to be recognized. The search for the optimum set of NN parameters can be done iteratively by minimizing Ntr

  Ti NN [ FL]  Ti ture 

2

2

Eq. 3.1

i 1

i.e., the sum over N tr examples of the square differences between the predicted temperature values, T NN (on the basis of the corresponding input fluorescence data, FL) and the respective true temperature values,

T true (known from the calibration). Figure 3.1 shows the one-hidden-layer NN architecture used in this work. It consists of N ne hyperbolic tangent input neurons and one linear output neuron, as defined in Eq. 3.2. For a given number of NN inputs ( N in ) and a set number of neurons, the NN is fully determined by the weights of the hidden neurons ( v j ,k ) and output neurons ( wk ). v0,k and w0 represent the bias nodes, in the hidden layer and output layer, respectively. In this work, the number of hidden neurons, N ne , was fixed at 2. N ne Nin   Ti NN [ FL]  w0   wk  tanh v0,k   v j ,k  FL( j )  k 1 j 1  

31

Eq. 3.2

Figure 3.1 Data flow of a NN architecture with one hidden layer, used in the current fluorescencebased thermometry implementation.

In our implementation, a large set (Ntr=8250) of experimentally obtained pairs of spectra and temperatures were used as examples to train and test the NN. The spectra were recorded while stabilizing the temperature at 11 values between 234 K and 311 K, in a temperature scan, by means of the calibration setup (Figure 2.1) described in Chapter 2. The temperature scan was performed stepwise to insure thermal equilibrium, i.e. equality between the temperature in the probed region and the thermometer location, 750 pairs of spectra and temperatures were recorded at each target temperature. 80% of the examples ( N tr =6600) were randomly taken for training the NN and the remaining 20% constituted the test set used for validating the NN. This procedure, which involved randomly shuffling the training and test examples, was repeated several times for cross-validation and to avoid the possibility of over-training. Only when the different choices of training and test set produced similar reconstruction performance, the calibration was considered as trustworthy. The NN yielded from one combination of training and test examples was selected and evaluated.

32

3.3

Performance of temperature reconstruction by NN

In the following, the temperature extraction performance of three types of NNs, which differ in the composition of the used set of fluorescence spectral information from the emission band between 575 nm to 675 nm, is evaluated. The first type of NN (from here on referred to IP NN or Integrated and Peak NN) is based on the integrated intensity and peak intensity of the emission band only (input neurons N in =2), which is representative of the conventional calibration procedure of polynomial fitting of the fluorescence integrated/peak intensity. The second type (from here on referred to as SP NN or shape NN) makes use of spectral shape associated information, which contains the emission maximum, FWHM, and values of the normalized spectra at 60 wavelengths between 575 nm and 675 nm (input neurons N in =62). In the third type of NN (multi-band NN), the fluorescence emission spectrum from 575 nm to 675 nm is split into five bands with a bandwidth of 20 nm, and the integrated intensities of the each of the five bands were employed as the NN input (input neurons N in =5). The reconstruction performance of each NN is depicted by box-whisker-plots of the reconstruction errors (TNN-Ttrue) at each calibration temperature, as shown in Figure 3.2a, b, and c respectively for IP NN, SP NN, and multi-band NN. Training examples are presented in the top figure of each plot and test examples in the bottom. In our box-whisker-plots, the top/bottom dots represent the maximum/minimum or range of reconstruction error, the diamond represents the mean value of reconstruction errors for the examples locating at the same target temperature, and the standard deviation (1σ) determines the height of the box. The horizontal line inside the box represents the median reconstruction error, verifying that the reconstructed data is not skewed. However, for the different types of NN illustrated in Figure 3.2, the reconstruction performance at different calibration temperatures is not identical. This might be a consequence result of some coincidental measurement fluctuations occurred during the calibration measurement.

33

It should be mentioned that, every change of samples, e.g. concentration, or optical conditions, e.g. optical filters, excitation laser, should be followed by a recalibration procedure. Figure 3.2 show that the spectral shape-based NN (SP NN) and multi-band NN offer a similar reconstruction performance, and that both of them perform better than the intensity based NN (IP NN). This is due to the spectral shape being substantially temperature dependent, while essentially not affected by mechanical instabilities of the measurement arrangement. Compared to photodetectors used in intensity based fluorescence thermometry, the bandwidth of spectral shape based thermometry is limited by the spectrometer. However, the similar reconstruction performance yielded by the multi-band NN suggests a pragmatic compromise between bandwidth and accuracy, by using (i) a series of optical bandpass filters and fast photodetectors for simultaneously acquiring the respective spectral band intensities, or (ii) a tunable filter that is scanned between the spectral bands, in combination with one fast photodetector.

34

TNN-Ttrue (K)

(a)

3 2 1 0 -1 -2 -3 3 2 1 0 -1 -2 -3

234.4 242.0 250.0 257.9 265.6 273.2 280.7 288.2 295.7 303.0 311.2

temperature (K) 1

(b) TNN-Ttrue (K)

0

-1 1

0

-1

234.4 242.0 250.0 257.9 265.6 273.2 280.7 288.2 295.7 303.0 311.2

temperature (K) 1

(c) TNN-Ttrue (K)

0

-1 1

0

-1

234.4 242.0 250.0 257.9 265.6 273.2 280.7 288.2 295.7 303.0 311.2

temperature (K) Figure 3.2

Box-whisker-plots of the temperature reconstruction error on the basis of IP NN (a), SP NN (b), and multiband NN (c) respectively, for training examples (top) and test examples (bottom) at different calibration temperatures. 35

Provided the constitution of the sample is not changed and data are acquired and pre-processed in the same way, the test errors in Figure 3.2 are representative for the performance of the trained NN that can be expected for new spectra. Thus the mean value of σ at different temperatures for the test examples (0.54 K, 0.27 K, and 0.25 K respectively for IP NN, SP NN, and multi-band NN) can be used to describe the overall reconstruction accuracy in the calibration range. Taking into account the integration time (4 ms) of the spectrometer and the number (5) of spectra that were averaged before being fed to the NN, the averaging σ value of test examples corresponds with 76 mK∙Hz−1/2, 38 mK∙Hz−1/2, and 35 mK∙Hz−1/2, respectively. By representing the reconstruction performance in this way, one can easily estimate the reconstruction error for a given bandwidth, which is an important advantage, in particular for the dynamic temperature measurement that this work will address. In order to further evaluate the reconstruction performance of the proposed method, we have also implemented the setup in Figure 3.3 to monitor dynamic temperature changes. An infrared pump laser (Vector, Coherent®) was used to photothermally induce a temperature change, which was enhanced by absorption of the sample to the infrared light by the added CuCl2. The power of the pump laser light, containing collinear infrared (wavelength 1064 nm) and green (532 nm) beams, was modulated by using an IntraAction® acousto-optic modulator (AOM). The first-order diffracted beam exiting the AOM was split into a 1064 nm pump beam and a 532 nm reference beam. The infrared component was illuminating and heating the sample and the transmitted light through the sample was blocked by a shortpass filter. The 532 nm reference beam was sent to the fiber detector entrance as well, so that the corresponding spectral peak could be used to monitor the pump laser modulation via the spectrometer. The power of the pump laser was sinusoidally modulated at 0.1 Hz between 42 mW and 123 mW. The laser power was measured by a PM100D, Thorlabs® power meter placed just in front of the optical window of the cryostat. The pump laser beam, with diameter 3 mm, was aligned to overlap with the green probe laser in the bulk sample cuvette (optical path 1.0 mm, with a total surface 45(L)×12.5(W) mm2). 36

While

photothermally exciting the sample, the spectrometer was continuously collecting spectra, which contained both a fluorescence and reference contribution, at a sampling frequency of 50 Hz.

liquid N2

delay generator

PC

spectrometer PT1000

pump laser

ITC503

vacuum pump

excitation laser

GPIB trigger

fiber thermocouple

cryostat SP

L1

function generator

BP

AOM

M1 M5 M4

Figure 3.3

L3

L2

reference beam

M2 M3

Experimental scheme for the determination of dynamic temperature changes by using fluorescence-based thermometry.

An intensity modulated pump laser was used to

photothermally alter the temperature of the sample.

The left axis in Figure 3.4 shows the temperature evolutions reconstructed by using IP NN (red), SP NN (blue) and the multi-band NN (green). The black curve depicts, with respect to the right axis is, for reference, the pump intensity, in arbitrary units. The reconstructed temperature evolutions by different types of NN are overlapping. However, the multi-band NN and SP NN yield a lower noise level compared to the IP NN, confirming the interpretation of the temperature reconstruction error in Figure 3.2a, b and c. Note that it was not possible to validate the found temperature evolutions by thermocouple measurements. The response time of thermocouples is too slow to be synchronous with the solvent temperature, and the temperature registered by a thermocouple in the pump laser beam would be affected by additional heating due to direct illumination. 37

pump intensity (a.u.)

3

TNN (K)

314

312

2

310

1

308

0

10

20

30

40

time (seconds) Figure 3.4

Reconstructed photothermally induced temperature evolutions (left axis), under sinusoidal modulation (pump intensity, right axis), by IP NN (red), SP NN (blue), and multi-band NN (green).

3.4

Conclusion

The application of NN recognition in fluorescence-based thermometry was demonstrated on an RhB dyed mixture of CuCl2 and glycerol. The approach exploits the advantages of NN recognition, taking into account temperature dependent spectral features in an optimum way. Three types of NN were presented: an intensity based NN representing the conventional fluorescence intensity based thermometry, a spectral shape based NN utilizing the combination of spectral shape associated fluorescence features (emission maximum, FWHM, and a selection values of the normalized spectrum), and a multi-band based NN where the emission spectra is split into 5 bands and the integrated intensities of each band are used. The spectral shape based NN shows higher reconstruction accuracy (38 mK∙Hz−1/2) than the intensity based NN (76 mK∙Hz−1/2). The similar reconstruction accuracy of the multi-band NN to the spectral shape

38

based NN, 35 mK∙Hz−1/2 provides a compromise between the high accuracy and low bandwidth of temperature measurement, since the spectral shape based thermometry requires a longer data acquisition time. The reconstruction performance was further verified by measuring the photothermally induced temperature changes. The performance of the proposed thermometry method to collect data for determining the frequency dependence of photothermal signals for thermal property characterization will be discussed in Chapter 5.

39

References

1.

Aigouy, L., G. Tessier, M. Mortier, and B. Charlot, Scanning thermal imaging of microelectronic circuits with a fluorescent nanoprobe. Applied Physics Letters, 2005. 87(18).

2.

Lojpur, V., M.G. Nikolic, D. Jovanovic, M. Medic, Z. Antic, and M.D. Dramicanin, Luminescence thermometry with Zn2SiO4:Mn2+ powder. Applied Physics Letters, 2013. 103(14).

3.

Maestro, L.M., P. Haro-Gonzalez, J.G. Coello, and D. Jaque, Absorption efficiency of gold nanorods determined by quantum dot fluorescence thermometry. Applied Physics Letters, 2012. 100(20).

4.

Maestro, L.M., E.M. Rodriguez, F.S. Rodriguez, M.C.I. la Cruz, A. Juarranz, R. Naccache, F. Vetrone, D. Jaque, J.A. Capobianco, and J.G. Sole, CdSe Quantum Dots for Two-Photon Fluorescence Thermal Imaging. Nano Letters, 2010. 10(12): p. 5109-5115.

5.

Borisov, S.M. and O.S. Wolfbeis, Temperature-sensitive europium(III) probes and their use for simultaneous luminescent sensing of temperature and oxygen. Analytical Chemistry, 2006. 78(14): p. 5094-5101.

6.

Ross, D., M. Gaitan, and L.E. Locascio, Temperature measurement in microfluidic systems using a temperature-dependent fluorescent dye. Analytical Chemistry, 2001. 73(17): p. 4117-4123.

7.

Feng, J., K.J. Tian, D.H. Hu, S.Q. Wang, S.Y. Li, Y. Zeng, Y. Li, and G.Q. Yang, A Triarylboron-Based Fluorescent Thermometer: Sensitive Over a Wide Temperature Range. Angewandte Chemie-International Edition, 2011. 50(35): p. 8072-8076.

8.

Fischer, L.H., G.S. Harms, and O.S. Wolfbeis, Upconverting Nanoparticles for Nanoscale Thermometry. Angewandte Chemie-International Edition, 2011. 50(20): p. 4546-4551.

40

9.

Romano, V., A.D. Zweig, M. Frenz, and H.P. Weber, Time-Resolved Thermal Microscopy with Fluorescent Films. Applied Physics B-Photophysics and Laser Chemistry, 1989. 49(6): p. 527533.

10.

Walker, G.W., V.C. Sundar, C.M. Rudzinski, A.W. Wun, M.G. Bawendi, and D.G. Nocera, Quantum-dot optical temperature probes. Applied Physics Letters, 2003. 83(17): p. 3555-3557.

11.

Wang, S.P., S. Westcott, and W. Chen, Nanoparticle luminescence thermometry. Journal of Physical Chemistry B, 2002. 106(43): p. 11203-11209.

12.

Kucsko, G., P.C. Maurer, N.Y. Yao, M. Kubo, H.J. Noh, P.K. Lo, H. Park, and M.D. Lukin, Nanometre-scale thermometry in a living cell. Nature, 2013. 500(7460): p. 54-U71.

13.

Allison, S.W. and G.T. Gillies, Remote thermometry with thermographic phosphors: Instrumentation and applications. Review of Scientific Instruments, 1997. 68(7): p. 2615-2650.

14.

Grattan, K.T.V., R.K. Selli, and A.W. Palmer, Ruby Decay-Time Fluorescence Thermometer in a Fiber-Optic Configuration. Review of Scientific Instruments, 1988. 59(8): p. 1328-1335.

15.

Saidi, E., B. Samson, L. Aigouy, S. Volz, P. Low, C. Bergaud, and M. Mortier, Scanning thermal imaging by near-field fluorescence spectroscopy. Nanotechnology, 2009. 20(11).

16.

Lou, J.F., T.M. Finegan, P. Mohsen, T.A. Hatton, and P.E. Laibinis, Fluorescence-based thermometry: Principles and applications. Reviews in Analytical Chemistry, 1999. 18(4): p. 235284.

17.

Lavieille, P., F. Lemoine, G. Lavergne, and M. Lebouche, Evaporating and combusting droplet temperature measurements using two-color laser-induced fluorescence. Experiments in Fluids, 2001. 31(1): p. 45-55.

18.

Kim, H.J., K.D. Kihm, and J.S. Allen, Examination of ratiometric laser induced fluorescence thermometry for microscale spatial measurement resolution. International Journal of Heat and Mass Transfer, 2003. 46(21): p. 3967-3974.

41

19.

Barilero, T., T. Le Saux, C. Gosse, and L. Jullien, Fluorescent Thermometers for Dual-EmissionWavelength Measurements: Molecular Engineering and Application to Thermal Imaging in a Microsystem. Analytical Chemistry, 2009. 81(19): p. 7988-8000.

20.

Sakakibara, J. and R.J. Adrian, Whole field measurement of temperature in water using two-color laser induced fluorescence (vol 26, pg 7, 1999). Experiments in Fluids, 1999. 27(1): p. U1-U1.

21.

Natrajan, V.K. and K.T. Christensen, Two-color laser-induced fluorescent thermometry for microfluidic systems. Measurement Science & Technology, 2009. 20(1).

22.

Dramićanin, M., Ž. Antić, S. Ćulubrk, S. Ahrenkiel, and J. Nedeljković, Self-referenced luminescence thermometry with Sm3+ doped TiO2 nanoparticles. Nanotechnology, 2014. 25(48): p. 485501.

23.

Glorieux, C., W.M. Gao, S.E. Kruger, K. Van de Rostyne, W. Lauriks, and J. Thoen, Surface acoustic wave depth profiling of elastically inhomogeneous materials. Journal of Applied Physics, 2000. 88(7): p. 4394-4400.

24.

Glorieux, C. and J. Thoen, Thermal depth profile reconstruction by neural network recognition of the photothermal frequency spectrum. Journal of Applied Physics, 1996. 80(11): p. 6510-6515.

25.

Krasnopolsky, V.M. and H. Schiller, Some neural network applications in environmental sciences. Part I: forward and inverse problems in geophysical remote measurements. Neural Networks, 2003. 16(3-4): p. 321-334.

26.

Liu, L., S. Creten, Y. Firdaus, J.J.A.F. Cuautle, M. Kouyaté, M. Van der Auweraer, and C. Glorieux, Fluorescence spectra shape based dynamic thermometry. Applied Physics Letters, 2014. 104(3): p. 031902.

42

Chapter 4

Development of ultrafast fluorescence based thermometry

This Chapter details an ultrafast temperature detection approach by using fluorescence based thermometry. In a proof-of-concept experiment, high-frequency (fast) heating of a volume of a semitransparent sample is achieved by dumping part of the optical energy in a laser pulse (nanoseconds). Due to the partially conversion of the absorbed laser radiation into heat, a transient temperature jump is generated in the sample [1-3]. With the aim to monitor the photothermally induced temperature evolution, two different experimental implementations of fluorescence based thermometry were examined and compared. One (i) is based on straightforward monitoring the evolution of the fluorescence intensity, excited by a 532 nm-CW laser, after a pump pulse, while the other one (ii) makes use of stroboscopic illumination of the sample by a fluorescence evoking probe laser pulse, synchronized with the pump laser pulse. 4.1

Introduction

The bandwidth of a thermometry technique determines the accessible time resolution of the employed approach. It may not be an important figure of merit (FOM) when studying large systems, since relative fluctuations in large systems are most often small and since it is usually considered as sufficient to monitor time-averaged temperatures. However, in small systems, temporal variations become increasingly

43

important and it can be of vital importance to measure the effective temperature over time scales shorter than the relevant thermal relaxation time of the measured system, defined by [4]

 thermal 

r2

Eq. 4.1



with r the characteristic radius of a spherical system and  the thermal diffusivity of the system. The thermal time constant strongly depends on the system size and its thermal diffusivity. This is illustrated in Figure 4.1, where the radius dependence of the thermal time constant is shown for different values of the thermal diffusivity, 10-8 m2 ∙s-1 (squares), 10-6 m2 ∙s-1 (circles), and 10-4 m2 ∙ s-1 (triangles), which are

thermal time constant (seconds)

representative for a typical liquid, polymer, and metal system respectively.

10 10

0

-8

2

-6

2

-4

2

=10 m /s =10 m /s

-2

=10 m /s 10

-4

10

-6

10

-8

-10

10 -10 10

Figure 4.1

10

-9

10

-8

10

-7

10

-6

radius (m)

10

-5

10

-4

10

-3

System size dependence of thermal time constant plotted at different thermal diffusivities.

44

4.2

CW probe-pulsed pump implementation

Figure 4.2

Experimental setup for monitoring the evolution of pulsed laser induced temperature changes. M, mirror; DM, dichroic mirror; L, lens; BP, bandpass filter; NF, notch filter; APD, avalanche photodiode; and OSC, oscilloscope.

In this section we verify the feasibility of detecting the temperature decay in glycerol shortly after individual pump laser pulses by monitoring the fluorescence intensity evolution. Figure 4.2 shows the layout of the experimental setup. An pulsed infrared ND: YAG laser (Lab-130-10, Quanta-Ray®), was used to excite the photothermal response of the sample, while a CW green laser (532 nm, Samba 100, Cobolt®) was used to evoke fluorescence in the sample, a RhB (2×10-6 molar) and CuCl2 (0.1 molar) dyed glycerol solution. At this concentration, the absorption coefficient of the sample at 1064 nm was measured by optical transmission and found to be IR=0.38 ±0.02 Neper.mm-1 at room temperature. The power of the probe laser was about 10 mW and the pulse energy of the pump laser, measured by a laser pulse energy meter (NOVA II, OPHIR®), was 65 mJ. The probe and pump laser beams, with diameters 0.8 mm and 3 mm, respectively, were aligned to overlap in the bulk of the sample. The sample was sealed in a cuvette (optical path length 10.0 mm, total surface 45(L)×12.5(W) mm2), which was embedded in a

45

copper sample holder that was mounted on the cold finger of an optical cryostat. The temperature control of the sample cell was accomplished by using the same approach as in Chapter 2, except for the PT1000 thermometer, read by an HP 34401A® multimeter, which here was directly attached to a wall of the cuvette while no thermocouple was present. On the right hand side of the setup, the path of the green excitation laser was imaged (perpendicular to the incident light) by a lens (F=100 mm, D=50.8 mm) with numerical aperture NA= 0.25, onto the entrance of the spectrometer fiber, and sent through a fiber to a spectrometer (USB 4000, Ocean Optics®), where the fluorescence spectrum was recorded. Both the object distance and image distance were 200 mm (2F configuration) in order to have an equal-sized image. Symmetrically, on the left side, a similar optical configuration was implemented, except for the fluorescence light not being detected by a spectrometer, but by an avalanche photodiode (APD130A, Thorlabs®), of which the voltage signal was monitored by a fast oscilloscope (LC564A, Lecroy®), thus monitoring the fast fluorescence intensity evolution induced by the pump pulses. During the measurements, the spectrometer was free-running at its maximum sampling rate of about 115 Hz, while the pump laser was controlled by a homemade Labview program to send a pulse every 7 seconds, thus allowing sufficient time for laser induced temperature change to be totally vanished, restoring the sample’s original temperature before the arrival of the next pump laser pulse.

46

1

TNN-Ttrue (K)

0

-1 1 0 -1

234.6 K 243.6 K 252.6 K 261.9 K 27.8 K 279.8 K 288.8 K 297.6 K

temperature (K) Figure 4.3

Reconstruction performance of the recalibrated SP NN represented by box-whisker-plots of the temperature reconstruction error on the basis of the SP NN, for training examples (top) and test examples (bottom) at different calibration temperatures.

For this experiment, a beforehand recalibration procedure was required since, compared to the configuration in Chapter 2, both the sample composition and the optical configuration were altered. In the pulsed configuration, the fluorescence light was collected with the collecting fiber positioned sideways and perpendicular with respect to the probe beam, while in the Chapter 2 it was collected along the path of the transmitted probe laser beam. Spectral shape associated information, containing the emission maximum, FWHM, and values of the normalized spectra, was employed to train the spectral shape based NN, of which the reconstruction performance is presented in Figure 4.3. An overall reconstruction accuracy of 51 mK∙Hz−1/2 was estimated from the mean value of σ at different temperatures for the test examples, 360 mK, by taking into account the effective bandwidth, 50 Hz, of the spectrum acquisition.

47

reconstructed temperature (K)

248.5 248.0 247.5 247.0 0

Figure 4.4

2

4

6 8 10 12 time (seconds)

14

16

Reconstruction of time-resolved temperature evolution after a single laser pulse, and best fit (red) by a single exponential decay.

Figure 4.4 shows the reconstructed temperature evolution after two consecutive pump pulses with 7 seconds in between. The sample temperature was initially at 247 K. The pump laser pulse induced temperature jump of about 1 K in Figure 4.4 is consistent with an estimated value of 1.2 K, as calculated by:

T 

Q  C a 2

Eq. 4.2

with Q=65 mJ the pulse energy, 0.38 mm-1the optical absorption coefficient, a=1.5 mm the beam radius, 1260 kg∙m3the density of glycerol, C=2400 J∙Kg-1∙K-1 the specific heat capacity of glycerol. The decay of the temperature after the initial sudden rise is due to the supplied heat flowing away from the pump laser light illuminated area towards the surrounding fluid, and from there to the sample holder. The decay time c of approximately 1.4 seconds, which was obtained by ad hoc exponential fitting (cf. Figure 4.4), can be expected to be of the order of the thermal diffusion time needed for heat to diffuse from the middle of the pump beam to the surrounding liquid. Using the thermal diffusivity of glycerol

48

=9.5×10-8 m2 ∙s-1, a characteristic diffusion distance of sqrt(4c ∙)=0.7 mm was obtained, which is

FL amplitude(volts)

reasonable, given the actual pump radius of a=1.5 mm.

FL intensity (a.u.)

Time (seconds) 1.8 1.6 1.4 1.2 1.0 230

240

250

260 270

280

290

300

temperature (K) Figure 4.5

Fluorescence intensity evolution recorded by an avalanche photodiode within a time window of 500 s (top) for a sample temperature of 247 K, and a DC calibration curve of fluorescence intensity versus temperature (bottom).

In parallel with the continuous data acquisition of the spectrometer, also the fluorescence intensity evolution monitored by the avalanche photodiode was recorded through the fast oscilloscope, as shown in Figure 4.5 (top). This allowed to investigate the feasibility of extracting the fast temperature evolution

49

from the fluorescence intensity evolution.

One challenge in this experimental configuration is the

intractable optical pickup from the pump laser, as a result of high conversion gain/sensitivity (2.5×106 Volts/Watt), essential for weak fluorescence light detection, of the employed avalanche detector. This is indicated by the saturation of the signal between 20 ns to 100 ns, in spite of the use of a high-quality notch filter and a bandpass filter respectively with optical density OD=6, 4 at 1064 nm. In addition, this approach faces another challenge in the time range between 100 ns and 2 s, as a consequence of both the abnormal response of the APD after saturation, and, partially, of the thermal expansion of the liquid. The issue of optical pickup can expected to be solvable by implementing multiple notch filters, albeit for a substantial extra cost. During this research, also a commercial photomultiplier tube (PMT, H10720 hamamatsu®) was tried out to monitor the weak fluorescence light. However, the observed signal contained a lot of spikes around a mean value, superimposed at random times, which resulted in a poor effective signal to noise ratio (SNR). Such spikes are known to represent one or several photons captured by the detector, since PMTs are capable of detecting single photons. Also contributions from the amplified dark current noise cannot be excluded in view of the high gain (106) of PMTs. All together, the APD gave a better performance. The DC calibration experiment, shown in Figure 4.5 (bottom), suggests that, for a concerned sample temperature of 247 K, the response to a temperature jump should be an increase in intensity. This is however not confirmed: the observed fluorescence intensity drops from 164 mV at t=10 ns (before the pump pulse) to 155 mV at t=2 s (after the pump pulse). Hence, in this case, adequate reconstruction of temperature evolution from the recorded fluorescence intensity change by simply using polynomial fitting of a temperature-fluorescence intensity curve, as is usually proposed in literature, is not possible in our case. The discrepancy is most probably linked to a substantial thermal expansion effect in the signal, which general leads to a reduced intensity as a result of thermal lens effects that induce probe beam

50

divergence and therefore intensity loss. This observation affirms the substantial advantage of temperature retrieval approaches that make use of spectral shape instead of intensity.

Figure 4.6

Reconstruction of temperature evolution induced by repetitive laser pulses at several initial temperatures. The fluorescence spectra from which the sample temperature evolution was reconstructed were collected at 8 Hz sampling rate. Due to the high repetition rate of the laser pulses (10 Hz), the used time scale of several tens of seconds does not allow to resolve the individual temperature increase steps.

We have also verified the feasibility of the spectral shape based approach to serve for remote temperature logging in this experimental configuration. The gradual temperature rise resulting from repeated pulsed laser heating, on a long time scale, up to 100 seconds, is depicted in Figure 4.6. In this experiment, the pump laser was operating at 10-Hz repetition rate while the spectrometer was continuously collecting spectrum. In principle, the sampling frequency of the used spectrometer can be up to 115 Hz in freerunning mode. However, since we were monitoring the temperature evolution on a long time scale, and in 51

order to keep the amount of stored data within reasonable limits, a function generator was employed to externally trigger the spectrometer to record spectra at a relatively low frequency around 8 Hz. From the acquired sequence of fluorescence spectra, the temperature evolution of interest was off-line reconstructed by using the trained SP NN depicted in Figure 4.3. The reconstructed temperature evolutions at different initial temperatures are all characterized by a rapid initial increase due to accumulation in the probed sample region of the repeated heat flux inputs supplied by the optically absorbed laser pulses. The temperature evolutions tend gradually to a stationary value, which is determined by the balance between laser heat supply on one hand, and heat losses by conduction and convection out of the illuminated region towards cooler parts of the sample and the sample holder otherwise. 4.3

Pulsed probe-pulsed pump implementation

With respect to applications requiring the monitoring of faster temperature evolutions, the main limitations of the proposed fluorescence-based thermometry approach are (i) the finite fluorescence time (ii) the 115 Hz sampling rate of the spectrometer in free-running mode, and, for frequency domain applications, (iii) the roughly 1/-decrease with frequency of the temperature oscillation magnitude and signal to noise ratio, (iv) the fluorescence intensity fluctuation arising from the thermal expansion induced by a pump laser. The first limitation is not very stringent, since the 2.5 ns fluorescence time, measured at 280 K by the time-correlated single photon counting (TCSPC) technique [5], of RhB in our sample allows in principle for a very large temperature data acquisition bandwidth up to 500 MHz.

We have

circumvented the second and third limitation by employing a stroboscopic approach, which makes the effective time resolution independent of the sampling rate of the spectrometer, and which exploits the possibility of using a pulsed laser approach in order to supply all the optical energy to heat the region of interest within the short time span of interest.

The fourth limitation is tackled by taking advantage of

neural network recognition of multiple fluorescence spectra features to determine the temperature while without involving the fluorescence intensity. 52

Figure 4.7

Experimental setup for determining the pulsed laser induced temperature evolution in a stroboscopic implementation. M, mirror; DM, dichroic mirror; L, lens; BP, bandpass filter; PD, photodiode; and OSC, oscilloscope.

Figure 4.7 shows the experimental implementation of the stroboscopic ultrafast fluorescence thermometry approach that we have used for detecting photothermally generated temperature jumps in a RhB (2×10-6 molar) and CuCl2 (0.1 molar) dyed glycerol solution. During the measurement the ND:YAG pump laser was working in single shot mode. The flash lamp was continuously firing at 10 Hz, but the trigger for the Q-switch that initiated a laser pulse was only issued about every 4 seconds (0.25 Hz). The TTL signal that was synchronous with the 10 Hz lamp firing, was sent to a digital delay generator (DG535, Stanford Research®), in order to produce, with controllable delay, a TTL signal that was used to trigger the 1 ns probe laser (532 nm PNG-002025-140, JDS Uniphase®) pulse that temporarily evoked the fluorescence of the sample. Since the Q-switch buildup time (after trigger) of the probe laser pulse (140 s) was shorter than the one of the pump laser pulse (170 s), this timing scheme allowed to time the probe pulse from 30 s before, till any delay after the pump pulse. A detailed timing scheme is presented in Figure 4.8.

53

Figure 4.8

Timing scheme of the stroboscopic implementation of fluorescence shape based thermometry

The pump beam and probe beam were aligned coaxially inside the sample and the diameter of the pump beam was around 3 mm, substantially larger than the one of the probe beam (0.8 mm), in order to approximate a situation of uniform laser heating and temperature field on the time scale of interest. The region trespassed by the probe laser was imaged by a lens (F=100 mm, D=50.8 mm) with numerical aperture NA= 0.25 onto the entrance of the fiber of the spectrometer, which was also triggered by the 10 Hz TTL reference signal. The intensity of probe laser light transmitted through the sample was monitored, together with the fluorescence spectrum, by the spectrometer, with the goal of indicating, synchronous with the acquisition of the fluorescence spectra, whether a pump laser pulse had been fired or not. This approach exploited the thermal lens effect evoked by the pump laser heating on the probe laser beam divergence, which was reflected in the collected probe laser light as an intensity drop. With the used pump and probe laser types, we had to deal with jitter on the Q-switch buildup time of the respective laser pulses after the electronic trigger, as shown in Figure 4.9. The jitter on the probe (top) and pump pulse (bottom) timing was respectively about 300 ns and 7 ns. This resulted in an initial 54

uncertainty of about 307 ns on the pump-probe delay time. In order to bypass this issue and reduce the effective uncertainty on the value of the time delay of the probe laser pulse after (or, if negative, before) the pump laser pulse, a fast oscilloscope (LC564A, Lecroy®) and two photodetectors (DET364A, Thorlabs®) were used to monitor extracted parts of the respective laser beams and accurately record the actual time delay (t) between the pump and probe pulse, as presented in Figure 4.10. By virtue of a synchronized acquisition of spectra and oscilloscope signals, each spectrum could be associated with the accurate effective delay time extracted from the oscilloscope signal traces.

amplitude (volts)

8

300 ns

6 4 2 0 -7

amplitude (volts)

5.0x10

-6

1.0x10

-6

1.5x10

-6

2.0x10

-6

2.5x10

-6

3.0x10

-6

-6

3.5x10

4.0x10

7 ns

2

1

0 -7

-7

-7

-7

-7

-7

-7

-7

4.2x10 4.3x10 4.4x10 4.5x10 4.6x10 4.7x10 4.8x10 4.9x10 5.0x10

time (seconds) Figure 4.9

Jitter of the probe laser (top) and the pump laser (bottom) timing.

55

-7

effective probe-pump delay (seconds)

10

-2

10

-3

10

-4

10

-5

10

-6

3.0x10

-7

0.0 -3.0x10

-7

0

50

100

150

200

250

cycle number Figure 4.10

Jitter induced timing noise (dots) and the re-sorted effective delay (solid). A total number of 250 probe-pump delays were arranged between -350 ns and 10 ms in logarithmic steps.

At each measurement cycle, as shown in Figure 4.8, a fixed amount of pump laser pulse energy was deposited (during the 10 ns pulse duration) to the sample, produced an upward temperature step (T). As mentioned in the previous Section, the probed temperature gradually decayed back to its value before the pump pulse before the next pump pulse was fired. The spectrometer was continuously recording spectra at 10 Hz during the entire measurement, Hence, given the pump laser Q-switch repetition rate of 0.25 Hz (one pump pulse every t0= 4 seconds, or equivalently, every M=40 reference periods), M spectra were recorded for each cycle n at time delay values ti=tn+100 ms*i (i=0, 1, 2, …, M-1). For measuring N cycles, a set of N*M spectra was collected at the following times after the beginning of the cycle.

56

t1  100ms  i ... tn  100ms  i

Eq. 4.3

... t N  100ms  i

i  0,1,2... M 1

with tn chosen so as to achieve probe-pump time delays varying between -300 ns and +10 ms (see further in this Section). 1

TNN-Ttrue (K)

0

-1 1 0 -1

207.9 K 215.1 K

224.4 K

233.5 K 243.3 K

252.3 K 261. 4 K

temperature (K) Figure 4.11

Re-calibrated spectral shape based NN for this experimental configuration.

The temperatures for the collection times were reconstructed from the respective spectra by using the spectral shape neural network depicted in Figure 4.11. The first column of the reconstructed temperature matrix (for i=0), T(t1, t2,… , tn,…, tN ), represents a fast temperature evolution between t1 and tN. Temperatures from the 2nd column to the last column at each row of the matrix represent the slow temperature decay, approximately from 100 ms to 4 seconds, of each cycle, in steps of 100 ms. Due to the temperature rise evoked by each pump laser pulse disappearing completely before illumination by the 57

respective next pump pulse, the data acquisition scheme allows to average the time evolution of the temperature after pump laser excitation by combining data from multiple cycles T(i x 100 ms). By merging and sorting temperatures reconstructed from all matrix columns, the full temperature evolution, T(t1, t2,…, , tn,…, tN, 100 ms, 200 ms,…, 100 ms*i,…, t0), induced by a nanosecond laser pulse could be obtained.

Figure 4.12

(a) Reconstructed temperature evolution for probe-pump delay times between -350 ns and +4 seconds (red). 200 early times between 0 ns and +10 ms (tN) are logarithmically spaced. 50 times average of longer delay times were linearly spaced between 100 ms and 4 seconds. The corresponding reference temperature (see text) is shown in blue. (b) the subtracted temperature evolution.

Figure 4.12 shows the result of a measurement sequence performed according to the above algorithm, in RhB and CuCl2 dyed glycerol at 240 K. The effective delay time was chosen between -350 ns (t1) and +10 ms (tN), in logarithmic steps with 250 values in total. The integration time of the spectrometer was

58

set at 40 ms, larger than tN =10 ms, so that the spectrometer integration period included the periods during which the (delayed) probe pulse was evoking the sample to fluoresce. The red curve in Figure 4.12a shows the full reconstructed temperature evolution, which consists of a series of values at the mentioned 250 short probe-pump time delays between t1 (-350 ns) and tN (40 ms), and 50 averages of temperatures reconstructed from spectra acquired at long probe-pump delay times between 100 ms to 4 seconds in steps of 100 ms. The total acquisition time of the data was 250x4 seconds or about 17 minutes. Adequate averaging of the 250 stroboscopically obtained values for long delay times requires the repeatability of the temperature evolution after the pump pulse during the 17 minutes acquisition period. Figure 4.12a also shows the time evolution of a reference temperature (blue), which was taken every 4 seconds for a probe-pump-delay of –100 ms, thus long after the previous pump laser pulse, after the disappearance of the laser induced temperature rise, and shortly before and not affected by the pump laser pulse of interest. The purpose of collecting the reference temperature evolution was to monitor the long term evolution of the sample temperature during the 17 minutes that were needed to collect the data needed to reconstruct the time dependence of the pump laser pulse induced temperature transient. By subtracting the reference temperature trace from the transient temperature trace of interest, fluctuations of the sample temperatures could be removed from the data, as illustrated by the result in Figure 4.12b. The temperature response to a pump laser pulse is characterized by a sudden jump that happens within about 10 ns, i.e. more or less the duration of the pump and probe laser pulse. After about 100 ms, the temperature starts to decay as a consequence of heat diffusing out of the excitation area to the cooler surroundings.

59

Figure 4.13

Noise reduced signal by averaging multiple measurement of 20 times.

A noise-reduced signal (black), obtained by averaging 20 measurements, is presented in Figure 4.13. The error bars indicate the standard deviation of the measurements. The typical noise on the (not averaged) temperatures acquired for the 250 short time delays lies 200 mK rms. This noise figure can be compared with the 38 mK rms noise in the frequency domain data for 1 Hz bandwidth, by taking into account the number of probe laser photons used to evoke fluorescence in both cases. In the pulsed case, the number of photons per pulse was Nph=25J/(h)=6.7×1013, with h Planck’s constant and  the optical frequency of the pulsed probe laser. Translating to an equivalent frequency domain experiment with P probe=3 mW as probe laser power, this would correspond to an equivalent acquisition time acq,FD=Nphh/Pprobe,, and an acquisition bandwidth fFD=1/acq,FD=120 Hz. The noise level of 200 mK rms for this 120 Hz bandwidth thus corresponds with a noise figure of 200 mK/1201/2=18 mK ∙ Hz-1/2, somewhat better than in the frequency domain scenario. The proposed method for fast stroboscopic thermometry offers an adequate approach to investigate the heat capacity relaxation behavior of glass forming liquids [6] that results from the finite relaxation time needed for cooperative molecular motions to respond to a stimulus (e.g. mechanical force, aligning 60

electric field, heat input,…) in a glassy network [7]. With the current setup, the temporal resolution of 10 ns results in a bandwidth of 100 MHz, which is 100 times higher than state of the art techniques to determine the frequency dependence of the heat capacity of relaxing materials [6]. As mentioned above, provided employing a probe and pump laser type that allow for more adequate timing control and synchronization with a minimum of jitter, the bandwidth offered by the stroboscopic approach is only limited by the laser pulse duration and the fluorescence time of the thhermochromics dye, the latter being of the order of a few nanoseconds. 4.4

Conclusion

We have also demonstrated the concept of using fluorescence thermometry to perform time-resolved photothermal detection, on one hand the determination of a gradual temperature rise due to the thermal accumulation of heat flux portions supplied by repetitive laser pulses, and on the other hand the determination of the transient temperature evolution after a single pump laser pulse. The demonstrated remote temperature monitoring approach can e.g. be useful for monitoring the temperature evolution during a chemical reaction in a semitransparent system [8]. Spatially resolved temperature monitoring can also be useful in the emerging research field of photothermal therapy [9, 10], in which laser light, ultrasound, or microwave absorption are induced heat to kill targeted malignant cells. Real time monitoring of the temperature distribution of the target region would be highly beneficial for tuning the intensity and duration of the excitation light source such that the heating is localized in the target tissue, with minimum impact on the surrounding healthy tissue. With respect to applications for faster temperature evolutions, the proposed spectral shape-based thermometry approach in a CW-probe configuration is limited by the 115 Hz sampling rate of the spectrometer and by the signal to noise decay with frequency due to the 1/ dependence of photothermally induced temperature modulation signal in the case of uniform heating. However, by

61

making use of pulsed laser excitation, larger temperature variations, combined with a large bandwidth, was achieved, allowing remote temperature detection in a very large temporal bandwidth, from 10 ns to 4 seconds, where the lower limit reflects the specifications on the timing control of the used pump and probe laser pulse triggering, but which is intrinsically only limited by the pump and probe laser pulse duration and the fluorescence time of the used thermochromic agent (here 2.5 ns for RhB). Fast stroboscopic thermometry can be interesting to study the temperature response of glassforming liquids to impulsive photothermally induced heating, and to derive therefrom the broadband relaxation behavior of the heat capacity. It should be noted that the proposed approach for remote fluorescence-based thermometry is generic. Research on an implementation that makes use of thermosensitive luminophores is ongoing.

62

References

1.

Liu, Y., D. Cheng, G. Sonek, M. Berns, C. Chapman, and B. Tromberg, Evidence for localized cell heating induced by infrared optical tweezers. Biophysical journal, 1995. 68(5): p. 2137-2144.

2.

Huang, X., P.K. Jain, I.H. El-Sayed, and M.A. El-Sayed, Plasmonic photothermal therapy (PPTT) using gold nanoparticles. Lasers in medical science, 2008. 23(3): p. 217-228.

3.

Zondervan, R., F. Kulzer, H. van der Meer, J.A.J.M. Disselhorst, and M. Orrit, Laser-driven microsecond temperature cycles analyzed by fluorescence polarization microscopy. Biophysical Journal, 2006. 90(8): p. 2958-2969.

4.

Marin, E., Characteristic dimensions for heat transfer. 2010.

5.

Maus, M., E. Rousseau, M. Cotlet, G. Schweitzer, J. Hofkens, M. Van der Auweraer, F.C. De Schryver, and A. Krueger, New picosecond laser system for easy tunability over the whole ultraviolet/visible/near infrared wavelength range based on flexible harmonic generation and optical parametric oscillation. Review of Scientific Instruments, 2001. 72(1): p. 36-40.

6.

Bentefour, E.H., C. Glorieux, M. Chirtoc, and J. Thoen, Thermal relaxation of glycerol and propylene glycol studied by photothermal spectroscopy. The Journal of chemical physics, 2004. 120(8): p. 3726-3731.

7.

Birge, N.O., Specific-heat spectroscopy of glycerol and propylene glycol near the glass transition. Physical Review B, 1986. 34(3): p. 1631.

8.

Herrero, M.A., J.M. Kremsner, and C.O. Kappe, Nonthermal microwave effects revisited: On the importance of internal temperature monitoring and agitation in microwave chemistry. Journal of Organic Chemistry, 2008. 73(1): p. 36-47.

63

9.

Bian, X., Z.L. Song, Y. Qian, W. Gao, Z.Q. Cheng, L. Chen, H. Liang, D. Ding, X.K. Nie, Z. Chen, and W.H. Tan, Fabrication of Graphene-isolated-Au-nanocrystal Nanostructures for Multimodal Cell Imaging and Photothermal-enhanced Chemotherapy. Scientific Reports, 2014. 4.

10.

Shah, J., S. Park, S. Aglyamov, T. Larson, L. Ma, K. Sokolov, K. Johnston, T. Milner, and S.Y. Emelianov, Photoacoustic imaging and temperature measurement for photothermal cancer therapy. Journal of Biomedical Optics, 2008. 13(3).

64

Chapter 5

Spatially resolved photothermal fluorescence spectroscopy in frequency domain

By probing and analyzing optically excited temperature changes in materials, photothermal techniques [1] are used to determine thermal and/or optical properties of a sample, in a wide range of in material characterization applications, [2] chemical analysis, [3, 4] and environmental research. [5-7] In this Chapter, along with the development of the ultrafast remote thermometry, special attention is given to the photothermal application of the fluorescence based thermometry. First, the spatial dependence of the calibrated amplitude and phase of photothermally induced temperature oscillations along the axis of the pump laser are determined at different modulation frequencies. A 1D multi-layer thermal diffusion model is employed to fit the spatial and frequency dependence of the extracted temperature signals. 5.1

Experimental setup

In the following, we verify the feasibility of the concept of spatially resolved detection of photothermally induced periodical temperature changes in a semitransparent sample by fluorescence-based thermometry, by taking advantage of its feasibility for remote, local and fast temperature monitoring. The experimental configuration is shown in Figure 5.1. An infrared pump laser (Vector, Coherent®), of which the intensity was sinusoidally modulated by using an acousto-optic modulator, was used to generate a photothermal response in the sample, while a green laser (532 nm, Samba 100, Cobolt®) was used to evoke 65

fluorescence in the sample, a RhB (2×10-6 molar) and CuCl2 (0.1 molar) dyed glycerol solution. The sample was sealed in a cuvette (optical path length 10.0 mm, total surface 45(L)×12.5(W) mm2), which was embedded in a copper sample holder, situated inside an optical cryostat. The temperature control of the sample cell was accomplished by using the same approach as described in previous Chapter. The path of the green excitation laser (inset, Figure 5.1) was imaged from the right side (perpendicular to the incident light) by a lens (F=100 mm, D=50.8 mm) with numerical aperture NA= 0.25, onto the entrance of the spectrometer fiber, and sent through a fiber to a spectrometer (USB 4000, Ocean Optics®), where the fluorescence spectrum was recorded. Both the object distance and image distance were 200 mm (2F configuration) in order to have an equal-sized image. The spectrometer was mounted on a micrometerdriven translation stage, allowing for scanning the detection location of temperature oscillations, at different distances from the cuvette-sample interface where the pump and excitation laser were entering the sample liquid. The power of the probe laser was about 3 mW and the pump laser power was modulated sinusoidally between 0 mW and 300 mW. The probe and pump laser beams, with diameters 0.8 mm and 5 mm, respectively, were aligned to overlap in the bulk of the cuvette. While photothermally exciting the sample, the spectrometer continuously collected the spectra at a sampling frequency of 50 Hz. The recorded spectra contained both a fluorescence and reference contribution.

66

Figure 5.1

Experimental setup for spatially resolved detection of photothermally induced temperature oscillations in a semitransparent sample. The sample cuvette (10 mm optical path) was mounted inside a vacuum cryostat. The path of the green excitation laser (circled in the inset) was imaged by a lens (focal length=F) onto the entrance of the spectrometer fiber, which was mounted on a translation stage scanning in Z axis. M, mirror; IF, interference filter; L, lens; DM, dichroic mirror; BT, beam trap; and BP, bandpass filter.

The temperature dependent fluorescence of RhB (2×10-6 molar) in CuCl2 (0.1 molar) dyed glycerol is summarized in Figure 5.2. It can be seen from the left axis of Figure 5.2, the UV-VIS absorption spectrum of the sample (dashed area) and as well that of the sample without CuCl2 (shaded area) at the room temperature (293 K), that the CuCl2 enhances the infrared absorption dramatically. The absorption coefficient of the sample at 1064 nm was measured by optical transmission and found to be IR=0.38±0.02 Neper.mm-1 at room temperature, compared to 0.01 Neper.mm-1 for the sample without CuCl2. 67

The right axis of Figure 5.2 shows the fluorescence spectra of the sample collected at different steady state temperatures, in steps of 10 K, ranging from 285 K to 335 K, in which both the integrated and peak (inset, left axis) intensity evolve in a monotonic way. The fluorescence intensity is clearly dependent on temperature. As discussed in Chapter 2, the observed decrease in fluorescence intensity with rising temperature can be attributed to a decrease of the quantum yield with temperature, which is a consequence of the increase of the rate constants for internal conversion or intersystem crossing, which compete with the fluorescence when the temperature is increased. Also the width and maximum position of the spectrum turn out to be strongly temperature dependent (inset, right axis). The full width at half maximum (FWHM) is broadening as the temperature is increasing, due to the emission bands becoming narrower at the low temperatures. The blue-shift covers a range of 3 nm wide over a temperature decrease of nearly 60 K. The blue shift is related to the temperature dependence of the interactions between the fluorophores and the surrounding solvent molecules, which assist in stabilizing and further lowering the energy level of the excited state from the Frank-Condon (FC) state to the relaxed state (R) by re-orienting. Towards lower temperatures, this solvent relaxation becomes slower due to the increasing viscosity, which implies more possibilities for emission from the F state, and hence a blue-shift [8, 9].

68

1.0 0.5

0.0 450

Figure 5.2

586

240

585 584

200

583 38.5 38.0 1.0

37.5 37.0

0.8 285

285.2 K 293.8 K 302.6 K 311.2 K 319.7 K 327.6 K

510

540

1.0 0.8 0.6

RhB RhB with CuCl2

300 315 330 temperature (K)

480

1.2

0.4 0.2

570

600

630

wavelength (nm)

fluorescence intensity (a.u.)

1.5

280

FWHM (nm) emission maximum (nm)

absorbance

2.0

peak intensity (a.u.) integrated intensity (a.u.)

2.5

1000 0.0

Temperature dependent fluorescence (right axis) of RhB in a mixture of CuCl 2 and glycerol for a selection of temperatures. The inset figure shows the temperature dependence of the integrated intensity, the peak intensity, the emission maximum and the FWHM. The left axis shows one representative absorption spectra of the sample with (dashed area) and without CuCl2 (shaded area) at room temperature.

Again, on the basis of the spectra collected at those steady state temperatures, a spectral shape based NN was trained for later temperature extraction. Figure 5.3 depicts the performance of the trained NN in the temperature between 285 K and 327 K, with a mean 106 mK for test examples (bottom). Taking into account the integration time (4 ms) of the spectrometer and the number (10) of spectra that were averaged before being fed to the NN, the mean  value of test examples correspond with 21 mK·Hz-1/2. Figure 5.4 shows some typical reconstructed dynamic temperature evolutions under 0.05 Hz sinusoidal (top) and

69

square-wave (bottom) modulation respectively. The amplitude and phase signal of the photothermally modulated temperature can be extracted by performing a fast Fourier transform (FFT) of the reconstructed temperature in time domain. 0.6 0.3

TNN-Ttrue (K)

0.0 -0.3 0.4 0.2 0.0 -0.2 -0.4

285.2 K

293.8 K

302.6 K

311.2 K

319.7 K

327.6 K

temperature (K) Figure 5.3

Reconstruction performance of the trained spectral shape based NN, represented by boxwhisker-plots at different calibration temperatures.

70

Figure 5.4

The reconstructed temperature evolution resulting from a sinusoidal wave (top) and square wave (bottom) modulated (0.05 Hz) pump laser (black line) by NN shown in Figure 5.3.

5.2

Results and discussion

As a test to validate the concept for thermal and/or optical property determination in depth, the frequency dependent photothermal response at three locations at different distances (Z1, Z2, and Z3) from the sample from the front cuvette-sample interface where the probe and pump beam were entering the sample compartment was determined by the proposed fluorescence-based thermometry approach. The distance between each position was fixed at 0.5 mm by the micrometer of the translation stage. The thermal diffusion length at 0.05 Hz, given by

  2 

71

Eq. 5.1

with9.5×10-8 m2∙s-1 the thermal diffusivity of glycerol and fhe modulation angular frequency is 0.78 mm. Since this is much smaller than the radius of pump beam (5 mm), the thermal diffusion model

1.0 0.8 0.6 10 0 -10 10

Figure 5.5

-1

frequency (Hz)

10

normalized phase (degrees)

normalized amplitude

of a 1D multilayered flat [10] could be employed to fit the experimental data.

0

Experimental data and the best fit of the photothermal signal at two distances from the cuvette wall through which the probe and pump beam entered the sample compartment. (a) Frequency dependent amplitude (top) signals and phase (bottom) signals at distance Z2 (circles) and Z3 (triangles), normalized to Z1 by amplitude division and phase subtraction, and the corresponding best fit (solid line) by a 1D layered thermal wave model. The signals in the range of 0.05-0.5 Hz were used for fitting due to the noisy signal at higher frequencies.

Figure 5.5 shows the frequency dependent amplitude and phase signals at distance Z2 (circles) and Z3 (triangles), normalized to the signal at Z1, and corresponding best fitting curves (solid lines). The data

72

were normalized by amplitude division and phase subtraction. Two fitting parameters were used for the fit: the distance Z1 from the cuvette-sample interface, and the optical absorption coefficient (), which was measured to be 0.38 ±0.02 Neper.mm-1 by optical transmission method at room temperature. The thermal properties of glycerol (thermal conductivity 0.28 W∙m-1∙K-1 and thermal diffusivity 9.5×10-8 m2∙s) and the cuvette material (thermal conductivity 1.38 W∙m-1∙K-1 and thermal diffusivity 8.2×10-7 m2∙s-1),

1

along with the cuvette wall thickness (1.25 mm) were fixed as known parameters.

Figure 5.6

Contour plot of two fitting parameters Z1 (0.77±0.03 mm) and absorption coefficient  (334±14 m-1).

73

The contour plot in Figure 5.6, representing the dependence of the normalized sum of squared fitting errors on the two fitting parameters, reveals a satisfactory fitting quality. This confirms the feasibility of fluorescence-based thermometry to determine photothermally induced temperature variations at different locations, with spatial resolution determined by the entrance aperture of the used spectrometer (in this work the fiber aperture was 100 micrometers) and extract the underlying thermal and optical material parameters.

5.3

Conclusion

The presented all-optical fluorescence spectral shape based dynamic thermometry exploits the advantages of NN recognition of temperature dependent spectral features, leading to an accuracy of 21 mK∙Hz-1/2 , allowing to monitor the photothermally induced temperature oscillation in a RhB and CuCl2 dyed glycerol sample.

A proof of concept of spatially resolved detection of photothermally induced temperature

oscillations, with spatial resolution determined by the entrance aperture of the fluorescence collecting optical fiber (100 micrometers in this work), was demonstrated by determining the frequency dependent photothermal response of the sample, at different depths inside the sample, based on the fluorescencebased thermometry. The satisfactory fitting of the experimental data suggests the technique has the perspective to be applied to profile the thermal properties and optical properties of semitransparent systems. The reported technique has the potential to extend photothermal spectroscopy (which has been widely applied in environmental research, chemical analysis and nuclear material characterization [1, 5, 6]) to in-depth investigation of dynamical temperature variations, optical absorption and thermal properties of semitransparent systems, and can be further extended to perform 3D thermotomography, by performing a combined scan of the probe beam and of the focal spot of the lens that collects the fluorescent light, or by means of 3D (confocal) fluorescence microscopy.

74

References

1.

Bialkowski, S., Photothermal spectroscopy methods for chemical analysis. Vol. 134. 1996: Wiley-Interscience.

2.

Malacarne, L.C., N.G. Astrath, G.V. Lukasievicz, E.K. Lenzi, M.L. Baesso, and S.E. Bialkowski, Time-Resolved Thermal Lens and Thermal Mirror Spectroscopy with Sample–Fluid Heat Coupling: A Complete Model for Material Characterization. Applied Spectroscopy, 2011. 65(1): p. 99-104.

3.

Terazima, M., Temperature lens and temperature grating in aqueous solution. Chemical physics, 1994. 189(3): p. 793-804.

4.

Terazima, M., N. Hirota, H. Shinohara, and Y. Saito, Photothermal investigation of the triplet state of carbon molecule (C60). The Journal of Physical Chemistry, 1991. 95(23): p. 9080-9085.

5.

Franko, M., Recent applications of thermal lens spectrometry in food analysis and environmental research. Talanta, 2001. 54(1): p. 1-13.

6.

Horne, K., H. Ban, A. Mandelis, and A. Matvienko, Photothermal radiometry measurement of thermophysical property change of an ion-irradiated sample. Materials Science and Engineering B-Advanced Functional Solid-State Materials, 2012. 177(2): p. 164-167.

7.

Franko, M. and C.D. Tran, Analytical thermal lens instrumentation. Review of Scientific Instruments, 1996. 67(1): p. 1-18.

8.

Birks, J.B., Organic Molecular Photophysics. 1975, New York: J. Wiley.

9.

Lakowicz, J.R., Principles of fluorescence spectroscopy. 3 ed. 2009: Springer.

75

10.

Glorieux, C., J. Fivez, and J. Thoen, Photoacoustic Investigation of the Thermal-Properties of Layered Materials - Calculation of the Forward Signal and Numerical Inversion Procedure. Journal of Applied Physics, 1993. 73(2): p. 684-690.

76

Chapter 6

Lock-in photothermal photoluminescence spectroscopy in frequency domain

With the aim of exploring the possibilities of using fluorescence based thermometry for application in non-destructive testing (NDT), this Chapter presents a digital lock-in scheme of photothermal photoluminescence (PL) spectroscopy in frequency domain. In a series of feasibility tests, photothermally induced temperature oscillations, T(f), at different frequencies (f) are recovered from the associated PL intensity oscillation of a thin layer of CdSe/ZnS-polystyrene (PS) composite, IPL(f). Advantage is taken of the linear temperature response of the PL intensity of the nanocomposite layer in the temperature range from 30 oC to 60 oC. In order to verify the feasibility of the proposed approach, the experimental results are fitted by a 1D thermal wave model and the fitting quality is evaluated.

6.1

Introduction

Quantum dots (QDs) are tiny nanocrystals made of semiconducting material, with diameter in the range of 2-10 nanometers. Their exceptional photostability, large luminescence quantum yield, and, in particular, the unique size-based tenability of their electronic and optical properties make them very attractive for a wide variety of emerging technologies and applications, e.g., light-emitting diodes (LEDs) [1-3], displays

77

[4-6], photovoltaics [7-9], biological labels [10-12], and optical sensing [11, 13-15]. Previous work has shown that the spectroscopic characteristics for both an ensemble of QDs [16, 17] and individual QD [18, 19] shift with temperature, thus enabling their use for optical temperature readout through changes in their spectral characteristics. Most often the integrated PL intensity has been utilized, due to the forthright measurement and relatively low cost involved, and in spite of the PL intensity sometimes being affected by mechanical, electrical, and optical instabilities of the measurement arrangement. The concept of utilizing QDs as optical thermometers has been demonstrated both in near and far field configurations. Near field configurations rely on the spatial-controlled scanning of a single QD over the sample surface to be thermally imaged. This is usually accomplished by modifying the tip of an atomic force microscopy (AFM) with QDs [20, 21]. Far-field configurations [18, 22, 23] often involve the direct injection of the QDs into the system of interesting in advance and afterward a superimposition of a fluorescence image of QDs and an optical transmission image, from which thermal imaging can be recovered on the basis of a beforehand calibration of PL intensities and temperatures. Generally, as the temperature is increased, the PL emission of QDs decreases. This can be attributed to thermally assisted energy transfer processes from bulk to surface (non-radiative) states, as well to the escalated interaction between exciton and longitudinal optical (LO) phonon at higher temperatures, leading to enhanced nonradiative decay [16, 24, 25]. In different types of QDs, such as CdSe [18], CdSe/ZnS [17, 26], CdTe [27], CdTe/ZnS QDs [28], a notable feature of thermally induced PL quenching is its linear temperature response near ambient temperature (20-50 oC). This is of particular interesting for application in photothermal detection, since it provides a constant thermal sensitivity in the working temperature range. In this work, we investigate the temperature dependence of the PL properties of a CdSe/ZnS core/shell QDs- polystyrene composite layer, spin-coated on a copper substrate, in the temperature range 30–60 oC. The linear response of the PL integration intensity to temperature is exploited to set up a lock-in scheme to monitor PL-detected

78

photothermally induced temperature oscillations up to 150 Hz, and verify the feasibility of exploiting their frequency dependence for remote thermal characterization of the substrate.

6.2

Temperature dependence of the QD photoluminescence

CdSe/ZnS core/shell QDs, purchased from Sigma-Aldrich (Lumidot™ CdSe/ZnS), were incorporated in a polystyrene (PS) matrix, by spin-coating (Spin coater Model P6700) a QDs (2 mg∙mL-1) and PS (0.8 g∙mL-1) blend in toluene, on the cleaned and polished copper substrate of thickness 2 mm, at 1000 rpm for 1 min. Polystyrene (Mw~35,000), toluene (99%) were all used as purchased from Sigma-Aldrich without further purification. Prior to the spin-coating, besides, in between the QDs-PS and copper layer, a layer of black paint (matt black) was implemented, acting as an absorbing layer for the photothermal application. The thicknesses of the QDs-PS layer and black absorber layer, measured by a micrometer (Digimicro, Nikon®), were found to be 13.1 m and 27.3 m, respectively. By increasing the doping of QDs to ensure sufficient photoluminescence for detection, the thickness of QDs-PS layer could be decreased down to submicron scale, which would of benefit to decrease the influence of the active layer (sensing layer) on signals.

79

M

probe laser

DM fiber L

BF

spectrometer OBL USB 10 m

PC 10 mm RS232

Figure 6.1

S PL TC

25 oC

Experimental setup (left) for investigation of the temperature dependent PL of QDs-PS film, and (right) a fluorescence image of the composite taken at 25 oC. M, mirror; DM, dichroic mirror; L, lens, BF, bandpass filter; OBL, objective lens, S, sample, TC, temperature controller.

The temperature dependence of the sample was investigated by using the setup in Figure 6.1 (left). A CW 532-nm laser (Samba 100, Cobolt®) beam was focused on the sample through an 10X-objective lens (OBL, NA=0.25), to excite the PL of the QDs-PS layer, with a spot diameter at the sample surface of approximately 20 m. The emitted PL light was collected by the same OBL and focused by a lens (f=10 mm) onto the entrance of fiber of the USB 4000 spectrometer (Ocean Optics®), where the spectrum was recorded. The sample was mounted on a hotplate (RET, control-visc, IKA®), with temperature control accuracy of 0.1 K. The temperature control and spectrum acquisition were programmed by a home-made Labview (National Instruments®) program via an RS232 and USB interface respectively. One fluorescence image of the QDs-PS film, taken at room temperature, is presented in Figure 6.1 (right). The cracks were caused by shrinking during drying of the film [29].

80

The steady-state PL spectrum of the CdSe/ZnS QDs-PS composite was measured at several selected temperatures, between 30 oC to 60 oC, in steps of 3 oC, as reported in Figure 6.2.

energy (eV) 2.05

2.00

1.85

5 4

1.75

0.8

o

o

slope=-1.3%/ C

59.8 C

3

1.80

1.0

o

normalized PL

3

1.90

30.7 C

6

PL intensity (10 counts)

1.95

0.6

2

30 35 40 45 50 o55 60 temperature ( C)

1 600

Figure 6.2

620

640

660

680

wavelength (nm)

700

720

PL spectra measured at different temperatures, and temperature response of PL intensity (triangles, inset) fitted by a linear function (solid line, inset).

6.3

Lock-in detection of photothermal photoluminescence signals

As can be seen in Figure 6.2, QD spectra are all characterized by a quasi-Gaussian emission shape, which is attributed to the band-edge recombination of electron–hole pairs within the CdSe core of the QDs [30]. The bandwidth is narrow, approximately 30 nm (0.1 eV) at half maximum. The PL emission is strongly dependent on temperature. The shift of the emission maximum, which is characteristic for the average bandgap of the QDs, evolves monotonically towards longer wavelengths with increasing temperature, as a

81

consequence of energy bandgap shrinkage at higher temperatures [31], which can be quantified through Varshni [32] equation:

AT 2 Eg (T )  Eg (0)  T B

Eq. 6.1

Where Eg(0) is the energy gap at 0 K, A is the temperature coefficient, and B is close to the Debye temperature of a material, typically around 150-180 K [33] for bulk CdSe. The Varshni relation suggests a quadratic low temperature asymptotic behavior while linear dependence at high temperatures. The observed redshift over a temperature range of 30 K is nearly 3 nm, corresponding to 8.5 meV of bandgap widening, which corresponds to a temperature dependent bandgap change of 2.8×10-4 eV∙K-1, which is consistent with literature values for bulk CdSe, (2.8-4.1) ×10−4 eV∙K-1[24, 25, 33]. The full widths at half maximum (FWHM) of PL emission also broaden slightly as temperature increases, from 30 nm (90 meV) at 30 oC to 31.3 nm (94 meV) at 60 oC. The broadening is due to increased exciton scattering by acoustic and LO phonons [24]. Clearly, the most pronounced spectral change occurs in the PL intensity, which is reduced by 40% throughout the investigated temperature range (inset, Figure 6.2). Interestingly, the PL intensity decreases quasi linearly over the range from 30 to 60 oC, with approximately -1.3%/K, in agreement with observations in previous works [17, 18] . The linear response to temperature changes of the PL intensity of the CdSe/ZnS QDs over a wide temperature range give them an excellent potential to be used as an optical temperature probe. This is of particular interest for applications in photothermal spectroscopy, which involve the detection of photothermally induced heating in time domain (pulsed heating) or frequency domain (modulated heating). In the following, taking advantage of this linearity, we present a digital lock-in photothermal PL spectroscopy in frequency domain.

82

M1

probe laser

APL

DM

APD

pre-amp

L

LOCK-IN

∆PL

BF

M3

FG

OBL

PL pump laser

Figure 6.3

BE

M2

S

Experimental setup for lock-in photothermal PL spectroscopy in frequency domain. M, mirror; DM, dichroic mirror; L, lens, BF, bandpass filter; OBL, objective lens, S, sample, BE, beam expander; APD, avalanche photodiode

The experimental setup for performing lock-in photothermal PL spectroscopy is depicted in Figure 6.3. An 808-nm pump laser (FAP 800, Coherent®) beam, of which the power was sinusoidally modulated by an external function generator (3320A, Agilent®) at frequency f, was expanded, collimated and uniformly impinging on the sample surface with a spot size of 8 mm, to guarantee 1D heat propagation, at an incident power of 0.5 W. Due to the energy conversion of partially absorbed laser radiation (with intensity variations I(t)), mainly at the surface of the black paint layer, into heat, the temperature at the sample surface was modulated. The thermal wave induced variations T(t) in turn modulated the PL intensity, IPL(t). At the center of the pump beam, where the probe beam was located, as illustrated in Figure 6.4a, the PL intensity variations were detected by an avalanche photodiode (APD130A, Thorlabs®). The voltage signal produced by the APD was firstly amplified (SR560, Stanford research systems ®) and then fed into a digital lock-in amplifier (SR830, Stanford research systems ®), from which the amplitude (APL) and the phase (PL) of the modulated PL signal was determined. By making 83

use of the following relations between the PL intensity oscillations and the underlying temperature oscillations,

T  I PL   APLeiPL  AT  APL  T  PL  

Eq. 6.2

Eq. 6.3

the amplitude (AT) and phase (T) signal of the photothermal induced thermal wave can be reconstructed. Note that the negative sign in right term in Eq. 6.2 is responsible for the inverse-linear response of the PL to temperature as discussed in the previous Section, which also implies a- phase delay between the PL intensity oscillation and temperature oscillation, as shown in Figure 6.4b.

84

(a)

pump probe PL

QDs

(b)

T

matrix

absorber

substrate

PL APL

pump wave

PL wave

AT

thermal wave

time (a.u.) Figure 6.4

Beam configuration of the pump and probe beam (a), the probe beam was located at the center of the pump beam. (b) one- phase delay between the PL (red) and thermal wave (blue) induced by the modulated pump laser (wine) heating.

85

6.4

Experimental results and discussion

Figure 6.5 shows the amplitude (top plot) and phase (bottom plot) of the periodically modulated PL signals, versus the modulation frequency, which was varied in logarithmic steps from 5 Hz to 150 Hz. The error bars were obtained by statistical analysis of 5 measurements at each frequency. Also the system response of the setup, H(f), which is mainly reflecting the frequency dependence of the electronic response of the APD detector and pump laser intensity modulation electronics, was determined, by

APL (mv)

detecting the intensity variations of the reflected pump light instead of the PL light.

8 6 4 2

o

PL ( )

175 150 125 100 75

10

Figure 6.6

frequency (Hz)

100

Amplitude (top) and phase (bottom) signal of the photothermally modulated PL in the frequency domain from 5 Hz to 150 Hz.

Since H(f) (Figure 6.7) is not flat, a PL data were normalized by amplitude division and phase subtraction of H(f). In addition, 180 degrees was subtracted from the normalized phase signal, in order to compensate for the sign difference between the PL signal oscillation and the temperature oscillation in Eq. 6.2.

86

AH (mv)

9.4 9.2 9.0 8.8

0

o

H ( )

10

-10 -20

Figure 6.7

10

100

frequency (Hz)

Non-flat system response of the experimental setup determined from the reflected pump light oscillations.

35

0.8 0.6 0.4

amplitude best fit

30

Labs.(m)

normalized AT

1.0

25

-20 o

T ( )

-40 -60

phase best fit

10

18

13

18

4.4

20

-80

Figure 6.8

8.8

13

frequency (Hz)

15 8

100

10

12

14

16

LQDs-PS (m)

Normalized photothermal response (left) of the sample in the frequency domain (dots) and the best fit (solid line). Contour plot (right) of normalized 2-error (dB) of two fitting parameters LQDs-PS (12.3±0.1 m) and Labs.(26.6±0.1 m) suggests a satisfactory fitting.

87

Figure 6.8 depicts the frequency dependent photothermal response of the sample and the best fit by a 1D3-layer thermal wave model [34], with parameters summarized in Table 6.1. The thermal properties of the three layers were taken from literature. The thicknesses of the QDs-PS layer and of the paint layer were used as fitting parameters. As shown in the last column of the Table 6.1, the best fitting values, 2.3±0.1 m, and 26.6±0.1 m respectively, are in good agreement with the values measured by a micrometer. The contour plot depicts the dependence of the normalized sum of squared fitting errors on the two fitting parameters. The sharp minimum validates the feasibility of the proposed technique to perform photothermal characterization of solid samples.

Material

Thermal

Thermal

conductivity

diffusivity

(W-1m-1K-1)

(m2s-1)

Thickness Best-fit

Nikon DIGIMICRO

Top layer

QDs-PS [35]

0.10

9.8×10-8

12.3±0.1 m

13.1 m

Middle layer

Paint [36]

1.45

2.1×10-7

26.6±0.1 m

27.3 m

Substrate layer

red brass [37]

60.6

1.8×10-5

Table 6.1

2.05 mm

Fitting parameters used in Figure 6.8

Furthermore, as a cross validation of the feasibility of the method for thermal characterization of materials, in Figure 6.9 the thicknesses measured by micrometer were utilized as known parameters to fit the thermal conductivity of the top two layers. During the fitting, the volumetric heat capacity (Cp) of the top two layers were fixed at 1.0×107 J∙K-1 ∙m-3, and 6.9×106 J∙K-1 ∙m-3 for top layer and middle layer respectively, as calculated from Table 6.1, while the thermal diffusivity was adapted through =/(Cp). As shown in Figure 6.9, the experimental data (left) were again fitted well by the 1D thermal wave model, and the best fitting thermal conductivity values, 0.12±0.02 W∙m-1∙K-1, and 1.3±0.1 W∙m-1∙K-1 respectively,

88

are consistent with values from literature [35, 36]. This confirms the potential application of the proposed

1.0

1.7

0.8

1.6

amplitude bestfit

0.6 0.4 -20

o

T ( )

-40

phase bestfit

-60

8.0 6.0

1.5

abs.(W/m/K)

normalized AT

approach in the field of material evaluation, as a tool for remote thermal property determination.

4.0

1.4

2.0

1.3 1.2 1.1

6.0 4.0

2.0

1.0

-80 10

Figure 6.9

frequency (Hz)

0.10

100

0.11

0.12

0.13

0.14

QDs-PS (W/m/K)

0.15

0.16

Cross validation of the fitting by using micrometer-measured thicknesses as known parameters to fit the thermal conductivity of the top two layers. The best fitting values, 0.12±0.02 W∙m-1∙K-1, and 1.3±0.1 W∙m-1∙K-1 are in line with the values from literatures listed in Table 6.1.

6.5

Conclusion

We have presented a lock-in photothermal photoluminescence spectroscopy in frequency domain by taking advantage of the linear response of the PL integration intensity of CdSe/ZnS QDs to the temperature near ambient condition. The amplitude and phase of the photothermally induced temperature oscillation at the sample surface were detected by a lock-in amplifier at different frequencies between 5 and 150 Hz, and fitted well by a 1D thermal-wave model. The bestfit thickness of the QDs-PS matrix layer and the absorber layer are consistent with values measured by a micrometer, which confirms the

89

potential application of the reported technique in field of NDT, i.e. remote thermal property characterization and/or thickness measurement. The reported approach can be extended to 2D lock-in thermography by implementing either an X-Y scanning of the sample, or by implementing a CCD camera to record the full-filed PL image of the sample surface, from which a temperature image with optical resolution can be extracted on the basis of a temperature calibration process, i.e. neural network recognition of the PL intensity at each pixel.

90

References

1.

Coe, S., W.-K. Woo, M. Bawendi, and V. Bulović, Electroluminescence from single monolayers of nanocrystals in molecular organic devices. Nature, 2002. 420(6917): p. 800-803.

2.

Sun, Q., Y.A. Wang, L.S. Li, D. Wang, T. Zhu, J. Xu, C. Yang, and Y. Li, Bright, multicoloured light-emitting diodes based on quantum dots. Nature Photonics, 2007. 1(12): p. 717-722.

3.

Kim, S., B. Fisher, H.-J. Eisler, and M. Bawendi, Type-II quantum dots: CdTe/CdSe (core/shell) and CdSe/ZnTe (core/shell) heterostructures. Journal of the American Chemical Society, 2003. 125(38): p. 11466-11467.

4.

Kim, T.-H., K.-S. Cho, E.K. Lee, S.J. Lee, J. Chae, J.W. Kim, D.H. Kim, J.-Y. Kwon, G. Amaratunga, and S.Y. Lee, Full-colour quantum dot displays fabricated by transfer printing. Nature Photonics, 2011. 5(3): p. 176-182.

5.

Wood, V., M.J. Panzer, J. Chen, M.S. Bradley, J.E. Halpert, M.G. Bawendi, and V. Bulovic, Inkjet-printed quantum dot-polymer composites for full-color AC-driven displays. Adv. Mater, 2009. 21(21): p. 2151-2155.

6.

Moeller, G. and S. Coe-Sullivan, Quantum-dot light-emitting devices for displays. Information Display, 2006. 22(2): p. 32.

7.

Nozik, A., Quantum dot solar cells. Physica E: Low-dimensional Systems and Nanostructures, 2002. 14(1): p. 115-120.

8.

Kamat, P.V., Quantum dot solar cells. The next big thing in photovoltaics. The Journal of Physical Chemistry Letters, 2013. 4(6): p. 908-918.

9.

Sargent, E.H., Colloidal quantum dot solar cells. Nature Photonics, 2012. 6(3): p. 133-135.

91

10.

Jaiswal, J.K., H. Mattoussi, J.M. Mauro, and S.M. Simon, Long-term multiple color imaging of live cells using quantum dot bioconjugates. Nature biotechnology, 2003. 21(1): p. 47-51.

11.

Medintz, I.L., H.T. Uyeda, E.R. Goldman, and H. Mattoussi, Quantum dot bioconjugates for imaging, labelling and sensing. Nature materials, 2005. 4(6): p. 435-446.

12.

Jaiswal, J.K. and S.M. Simon, Potentials and pitfalls of fluorescent quantum dots for biological imaging. Trends in cell biology, 2004. 14(9): p. 497-504.

13.

Costa-Fernández, J.M., R. Pereiro, and A. Sanz-Medel, The use of luminescent quantum dots for optical sensing. TrAC Trends in Analytical Chemistry, 2006. 25(3): p. 207-218.

14.

Ma, Q. and X. Su, Recent advances and applications in QDs-based sensors. Analyst, 2011. 136(23): p. 4883-4893.

15.

Murphy, C.J., Peer reviewed: optical sensing with quantum dots. Analytical Chemistry, 2002. 74(19): p. 520 A-526 A.

16.

Jing, P., J. Zheng, M. Ikezawa, X. Liu, S. Lv, X. Kong, J. Zhao, and Y. Masumoto, Temperaturedependent photoluminescence of CdSe-core CdS/CdZnS/ZnS-multishell quantum dots. The Journal of Physical Chemistry C, 2009. 113(31): p. 13545-13550.

17.

Walker, G.W., V.C. Sundar, C.M. Rudzinski, A.W. Wun, M.G. Bawendi, and D.G. Nocera, Quantum-dot optical temperature probes. Applied Physics Letters, 2003. 83(17): p. 3555-3557.

18.

Li, S., K. Zhang, J.-M. Yang, L. Lin, and H. Yang, Single quantum dots as local temperature markers. Nano letters, 2007. 7(10): p. 3102-3105.

19.

Labeau, O., P. Tamarat, and B. Lounis, Temperature Dependence of the Luminescence Lifetime of Single C d S e/Z n S Quantum Dots. Physical Review Letters, 2003. 90(25): p. 257404.

20.

Jaque, D. and F. Vetrone, Luminescence nanothermometry. Nanoscale, 2012. 4(15): p. 4301-4326.

92

21.

Aigouy, L., B. Samson, G. Julié, V. Mathet, N. Lequeux, C.N. Allen, H. Diaf, and B. Dubertret, Scanning near-field optical microscope working with a CdSe∕ ZnS quantum dot based optical detector. Review of Scientific Instruments, 2006. 77(6): p. 063702.

22.

Cadiau, A., C.D. Brites, P.M. Costa, R.A. Ferreira, J. Rocha, and L.D. Carlos, Ratiometric nanothermometer based on an emissive Ln3+-organic framework. ACS nano, 2013. 7(8): p. 7213-7218.

23.

Maestro, L.M., E.M. Rodríguez, F.S. Rodríguez, M.I.-d. la Cruz, A. Juarranz, R. Naccache, F. Vetrone, D. Jaque, J.A. Capobianco, and J.G. Solé, CdSe quantum dots for two-photon fluorescence thermal imaging. Nano letters, 2010. 10(12): p. 5109-5115.

24.

Valerini, D., A. Creti, M. Lomascolo, L. Manna, R. Cingolani, and M. Anni, Temperature dependence of the photoluminescence properties of colloidal Cd Se∕ Zn S core/shell quantum dots embedded in a polystyrene matrix. Physical Review B, 2005. 71(23): p. 235409.

25.

Joshi, A., K. Narsingi, M. Manasreh, E. Davis, and B. Weaver, Temperature dependence of the band gap of colloidal CdSe∕ ZnS core/shell nanocrystals embedded into an ultraviolet curable resin. Applied Physics Letters, 2006. 89(13): p. 131907.

26.

Kuznetsov, S.N., A.B. Cheremisin, and G.B. Stefanovich, Photoluminescence response of colloidal quantum dots on VO2 film across metal to insulator transition. Nanoscale research letters, 2014. 9(1): p. 1-7.

27.

Wang, S., S. Westcott, and W. Chen, Nanoparticle luminescence thermometry. The Journal of Physical Chemistry B, 2002. 106(43): p. 11203-11209.

28.

Han, B., W.L. Hanson, K. Bensalah, A. Tuncel, J.M. Stern, and J.A. Cadeddu, Development of quantum dot-mediated fluorescence thermometry for thermal therapies. Annals of biomedical engineering, 2009. 37(6): p. 1230-1239.

93

29.

Balazs, A.C., T. Emrick, and T.P. Russell, Nanoparticle polymer composites: where two small worlds meet. Science, 2006. 314(5802): p. 1107-1110.

30.

Hines, M.A. and P. Guyot-Sionnest, Synthesis and characterization of strongly luminescing ZnScapped CdSe nanocrystals. The Journal of Physical Chemistry, 1996. 100(2): p. 468-471.

31.

Wan, J., J. Brebner, R. Leonelli, G. Zhao, and J. Graham, Temperature dependence of freeexciton photoluminescence in crystalline GaTe. Physical Review B, 1993. 48(8): p. 5197.

32.

Varshni, Y., Temperature dependence of the energy gap in semiconductors. Physica, 1967. 34(1): p. 149-154.

33.

Madelung, O., Numerical data and functional relationships in science and technology: crystal and solid state physics. semiconductors. special systems and topics, comprehensive index for III/17a··· i. 1985: Springer-Verlag.

34.

Glorieux, C., J. Fivez, and J. Thoen, Photoacoustic investigation of the thermal properties of layered materials: Calculation of the forward signal and numerical inversion procedure. Journal of applied physics, 1993. 73(2): p. 684-690.

35.

Mark, J.E., Physical properties of polymers handbook. 1996: Springer.

36.

Raghu, O. and J. Philip, Thermal properties of paint coatings on different backings using a scanning photo acoustic technique. Measurement science and technology, 2006. 17(11): p. 2945.

37.

Kothandaraman, C., Heat and mass transfer data book. 2004: New Age International.

94

Chapter 7

Temperature dependent time-resolved photothermal spectroscopy of glycerol

Driven by research group's historical interest [1-8] in the complex relaxation behavior in glass forming liquids, in this chapter, we explore the feasibility of two time domain photothermal techniques to investigate the relaxation behavior of specific heat capacity, which offers a unique window on the glass transition and/or the behavior of relaxing materials. Being very well characterized [9-11] glycerol was used as material of interest in this work. In the first photothermal approach, density and heat capacity relaxation information was extracted from photothermally generated transient thermal lens signals. In the photothermal experiment, the developed ultrafast fluorescence thermometry, as described in Chapter 4, were explored to extract the pure temperature response to photothermally induced heating, and from that the relaxation behavior of the heat capacity.

7.1

Introduction 

Relaxation dynamics in supercooling systems

Upon cooling below the freezing point, a majority of liquids (molecular liquids, polymeric liquids, etc.) form a glass if cooled sufficiently fast to avoid the crystallization [12, 13]. One of the most essential 95

features of glass forming materials is their frequency dependent response to different kinds of stimuli [1416], with a strongly temperature dependent relaxation frequency [17, 18], and with a moderately temperature dependent relaxation strength [17]. For many applications, the most important response quantity is the elastic modulus [19-21]: with increasing frequency, due to a decreasing amount of possibilities for cooperative molecular motions [22-24] within an oscillation period, glass forming materials behave more stiff. The relaxation frequency, below which the material behaves substantially more softly than in the high frequency limit, drastically decreases with decreasing temperature, as cooling down goes along with the amount of available kinetic energy of the molecules dropping below characteristic activation energy [21, 25]. The so-called fragility [26, 27] of a system, representing how rapidly the dynamics of a material slows down as it is cooled toward the glass transition, reflects to what extent the activation energy decreases with increasing temperature [28]. As the temperature dependence of the relaxation frequency of glassforming liquids is very strong, many efforts have been done to develop and combine different experimental approaches to investigate relaxation in a frequency range from GHz-THz down to mHz. The study of mechanical relaxation is typically done by a combination of classical mechanical techniques for frequencies below 100 Hz [29, 30], to acoustic (audio frequencies) [31], ultrasonic (up to MHz frequencies) [32] and impulsive stimulated scattering (MHz-GHz) techniques [1, 33-36]. One interesting approach to cover a wide frequency range is to exploit the phenomenon of time-temperature superposition[37],by extrapolating information on the temperature dependence of the relaxation frequency and the frequency dependence of the elastic modulus from an experimentally accessible frequency range to the frequency range of interest. An alternative way to extract the temperature dependence of the relaxation frequency of the mechanical response is to determine the dielectric response [7, 38, 39], which gives access to a very wide frequency range between mHz and GHz. The idea is to exploit the empirically confirmed universality of the fragility of glass formers across different response functions. In casu the Arrhenius plot of mechanical relaxation can be

96

extrapolated from a narrow experimentally accessible frequency range to the wide frequency range of interest, by employing information obtained in the wide bandwidth covered by dielectric spectroscopy. 

Relaxation of specific heat capacity

Given the underlying physics of cooperativity of molecular motions being responsible for relaxing behavior, it can be expected that the fragility, which is a fingerprint of the involved processes, has been found to be the same for different response functions and excitation mechanisms[40, 41], such as dynamical mechanical excitation (mechanical relaxation) and a dynamically orienting electric field (dielectric relaxation). Relaxation is also observed when a glassforming liquid is dynamically heated: a time scale dependent part of the dynamical supply of vibrational energy turns out to be channeled to the potential energy of the molecular network. This relaxation phenomenon, which involves heat capacity relaxation and thermal expansion relaxation is often referred to as structural relaxation[42, 43] and has been experimentally observed as a frequency dependence of the specific heat capacity, by means of the 3 technique, photopyroelectric (PPE) spectroscopy and nanocalorimetry [44]. For glycerol and other glass formers the fragility connected to the relaxation of the heat capacity [45, 46] was found to be the same as the one connected with dielectric and mechanical relaxation. PPE spectroscopy experiments revealed a relaxation strength of around 0.43 for the heat capacity of glycerol. Heat capacity spectroscopy is thus a valuable technique to investigate and determine the relaxation behavior of glass formers. Till now, this possibility has been much less exploited than mechanical and dielectric relaxation spectroscopy. This is a consequence of the frequency range that could be experimentally covered by thermal response techniques being too narrow. It is well known that a DSC cooling run on a glass former allows to determine the calorimetric value of T g, which is often not so far from the kinematic value of Tg, which in turn corresponds to the point on the Arrhenius plot with relaxation time equal to 100 seconds. Combining DSC cooling curves acquired at different cooling rates, 97

it is also possible to estimate the fragility of a glass former [47-49], but this estimation relies on an effective frequency range which is quite narrow and low. Using the 3- technique, heat capacity spectroscopy has been shown to be possible from mHz frequencies till frequencies of the order of 1 kHz [36]. By employing PPE spectroscopy, the experimental frequency range was further extended to 100 kHz [50, 51]. In this work, we explore the feasibility of two time domain photothermal techniques to extend the frequency range for heat capacity spectroscopy further, till MHz and higher. Thermal relaxation behavior in a time domain experiment, with impulsive heat supply, can be expected be revealed as follows. At short times, the supercooled liquid is too slow to take up heat in its slow degrees of freedom, so that the effective, instantaneous heat capacity is low, and the temperature rise is high. Only at later times, energy can flow from the vibrational energy degrees of freedom towards the cooperative rearrangements of the amorphous network, so that the effective, instantaneous heat capacity becomes higher, and the temperature is reduced. Being very well characterized [9, 10], glycerol is used as material of interest. We investigate the feasibility of disentangling the frequency dependences of the heat capacity and the thermal expansion coefficient from two experimental approaches: transient thermal lens signals and fluorescence based thermometry. The interpretation of the results is supported by earlier heat capacity relaxation and structural relaxation results on glycerol, obtained by impulsive stimulated scattering[8, 52, 53], photopyrolectric spectroscopy [3, 51] and the 3- technique [37].

98

7.2

Time-resolved thermal lens spectroscopy of supercooled glycerol 

Experimental setup

Since its discovery by Gordon et al. in 1965 [54], the thermal lens (TL) effect has been widely employed to determine weak optical absorbance values of gases and liquid phase samples [55, 56]. TL spectroscopy is a photothermal method that detects the temperature rise in a sample due to heat generated from nonradiative relaxation processes resulting from optical absorption of light. By virtue of its high sensitivity and versatility of detection, the TL method has become widely used for electro-optical characterization of materials [57, 58], spectrometry of photochemical reactions [55, 59], microvolume and trace analyses [60] of gas, liquid, and solid samples. liquid N2

temperature controller

vacuum pump

TL sample probe beam pump beam

OSC

cryostat

sample M1

Figure 7.1

DM

L

IF

PH

PD

Experimental setup of nanosecond laser induced thermal lens spectroscopy, and mode mismatched beam configuration (inset). M, mirror; DM, dichroic mirror; L, lens; IF, interference filter; PH, pinhole; PD, photodetector; OCS, oscilloscope.

99

In this work a time-resolved TL scheme, in which nanosecond pulsed laser light is used to impulsively heat the sample was employed to measure the photothermal impulse response of the density of glycerol and thus investigate the structural relaxation behavior. The global density response to impulsive heating can be considered as a convolution between the temperature response to impulsive heating (with the specific heat as response parameter) and the density response to a sudden temperature rise (with the thermal expansion coefficient as response parameter). Due to its time-varying and often spatially nonuniform character, the local thermal expansion response is unavoidably accompanied by the launching of acoustic waves, which carry information on the (relaxation behavior of the) elastic modulus. In many respects the pulsed TL scheme is very similar to the one of impulsive stimulated scattering. The main difference between the two photothermal approaches lies in the geometry of the optical excitation pattern: while in impulsive stimulated scattering (ISS), also called the transient grating (TG) method, the light pattern is spatially quasi-periodic and characterized by a single wavenumber, the typically Gaussian pattern used in a TL configuration results in a wide spectrum in wavenumber domain. This difference in spectral content has mainly consequences with respect to the thermal diffusion tail of the signal, which is purely exponential for ISS signals in bulk transmission mode and more complicated for TL signals. ISS signals also contain a very clear, sinusoidal acoustic wave contribution with acoustic frequency f a=c/, with  the wavelength of the periodic light grating, and c the speed of sound. The wideband acoustic wave contribution in TL signals has a more complicated character, and, due to the excitation light beam typically being quite wide compared to typical transient grating wavelengths, does not contain high acoustic frequencies. Figure 7.1 shows the experimental scheme of the used TL setup. A pulsed ND: YAG laser (Lab-130-10, Quanta-Ray®), operating at 10 Hz, was used to produce the thermal lens, while the time evolution of which was simultaneously monitored by means of a CW 532-nm laser (Samba 100, Cobolt®). The two beams were collimated, with diameter of 8 mm and 2 mm for pump beam and probe beam respectively,

100

merged coaxially through a dichroic mirror, thereafter, and focused by a Plano-convex lens (F=125 mm, D=50.8 mm) into the bulk of the pure glycerol sample (≥99%), which had been purchased from Sigma Aldrich and further purified with vacuum evaporation at 75 oC for 4 hours. The sample was sealed in a cuvette (optical path length 2.0 mm, total surface 45(L)×12.5(W) mm2), which was embedded in a copper sample holder, situated inside an optical cryostat. The temperature control of the sample cell was accomplished by using the same approach as in Chapter 4. Due to the aberration of the lens, the focal points/ waists of the two beams were displaced with each other inside the sample, as illustrated in the inset of Figure 7.1, in a so-called mode-mismatched configuration [57, 61]. Given that the curvature of the lens (LA1314, Thorlabs®) was 64.4 mm, and the refractive index of the lens material N-BK7 is 1.5066 and 1.5195 at 1064 nm and 532 nm [62], respectively, the axial displacement between the two beam axes, predicted by lens maker’s equation was approximately 3 mm. Near the pump beam waist, local photothermal heating resulting from the Gaussian beam produced a transverse temperature gradient which induced a refractive index gradient, thermal lens. Since glycerol (as most materials) expands upon heating and the refractive index is proportional to the density, dn/dT is negative and thus a thermal lens behaving like a concave lens was produced, which in turn modified the wavefront of the propagating probe beam and caused beam divergence. After passing through the sample, the pump beam was blocked by an interference filter. The intensity of the 15-mm diameter transmitted probe beam was first windowed by a 1-mm pinhole located in far field (around 1.5 meters away from the sample) and then focused again to a homemade photodetector, with bandwidth from DC to 100 MHz. The produced voltage signal was monitored with a fast oscilloscope (LC564A, Lecroy®) and 500 consecutive TL cycles, corresponding to 50 seconds of acquisition time, were recorded and averaged by the oscilloscope in order to get better SNR. For the sake of extracting a maximum of information from the TL signal, each signal was acquired for two different acquisition time windows, with 100 s sampling at 1 GHz, and 5 ms sampling at 2 MHz,

101

respectively, and then off-line combined together, and resampled on a logarithmic scale for the purpose of optimum data reduction.

normalized amplitude

1.0

T=224 K

0.9 0.8 1.0

0.9

0.8 1E-9

1E-8

1E-7

1E-6

1E-5

1E-4

1E-3

0.01

0.1

time (seconds)

Figure 7.2

Typical TL signal measured at a short (100 s, top) and long (5 ms, bottom) time window. The TL signal of each was normalized to unity for the short time limit.



Results and discussion

Figure 7.2 shows one representative measurement conducted at 224 K and acquired in two different time windows. Since in this work such two signals were not recorded simultaneously, their relative magnification was adjusted by normalizing the signal at short time limit to unity in order to combine them in an optimum smooth way. Alternatively, one can employ two oscilloscopes to realize simultaneous acquisition of the fast and slow time part of the signal respectively.

102

normalized TL

1.0

0.9

0.8

rescaled TL

1.0 0.8 0.6 0.4 0.2 0.0

1E-9

1E-8

1E-7

1E-6

1E-5

1E-4

1E-3

0.01

0.1

time (seconds) Figure 7.3

Full time evolution of the TL signal (top) at 224 K obtained by combining the two measurements in Figure 7.2, and rescaled TL signal (bottom).

Figure 7.3 depicts a full-scale TL signal (top) between 1 ns and 50 ms, which is composed of 2000 data points (circles), obtained by interpolation of the two combined data sets (solid line) on a logarithmic scale. This procedure yields a full view of the TL signal while representing it by only 2000 data points, much less than the original 200k data points, without losing important signal features. This allows to dramatically reduce the computing time involved in later fitting of the data. The resampled signal was further rescaled (bottom) to [0 1]. It clearly shows three different processes separately, (i) a rapid increase in amplitude (1 ns to 80 ns), which can be attributed to the fast part of the temperature rise and resulting thermal expansion response to the sudden heat input, followed by (ii) a continued slow rise due to the slow part of the thermal expansion and underlying temperature dynamics, until (iii) finally the slow 103

thermal diffusion from heat radially out of the heated laser beam path dominates, and the amplitude exponentially decays to zero, with a characteristic thermal diffusion decay time determined by the width of the pump laser beam at the focal point. In order to interpret the observed transient TL signal, it is instructive to take a look at the temperature and density response obtained in a transient grating experiment [35, 36, 52] which is typically performed by exciting the sample with an impulsive and quasi-periodic diffraction light pattern with wave number q=2/ and spatial period  along the x-axis, as shown in Eq. 7.1 and 7.2,

T ( x, t ) 

 ( x, t ) 

Q0 cos( qx ) exp  q 2 t  H (t )  T0 exp  q 2 t  H (t ) C

Eq. 7.1

  Q0 sin( qx )  q 2 exp  q  t  cos( qc t )  sin( qc t )     H (t ) L L cL  q 2 2   

 Cq  1  

Eq. 7.2

 cL 2 

with  [kg.m-3] the static part of the density, C [J.kg-1.K-1] the specific heat capacity, α [m2.s-1] the thermal diffusivity,  [K-1] the thermal expansion, cL [m.s-1] the longitudinal acoustic velocity, Q0 [J.s-1.m-3] the power density amplitude and H(t) the Heaviside step function. From the density response (x,t), a compact expression can be obtained for the ISS signal response of a glassforming material:

s  ( A  B)et / th  Aet / ac cos(2 f ac t )  Be( t / st )



Eq. 7.3

Eq. 7.3 shows that the ISS signal dynamics is determined by three components: a two-step thermal expansion jump, with amplitudes A (instantaneous density response) and B (relaxation part of the density response) respectively, an oscillating acoustic part with amplitude A and frequency given by f ac=c/grating, and a thermal diffusion tail, with decay time th=1/(q2), with  the thermal diffusivity. In the case of a 104

glassforming liquid, the relaxation character of the response is revealed by the finite acoustic decay time, ac, and the structural relaxation time st. A+B represents the total amount of thermal expansion that ultimately tends to build up (except when preceded by thermal diffusion decay) in response to sudden heating. In a TL configuration, the Gaussian optical excitation pattern, and therefore the signal, consists of a continuous spectrum of wavenumber components, and is therefore difficult to express analytically. However, taking into account that due to the wide beam diameter the thermal diffusion decay is very slow, the acoustic frequencies are quite low, the acoustic contribution to the signal is very small, and considering times before thermal diffusion sets in, the expression for the TL signal can be reduced to

s  ( A  B) H (t )  Aet / ac  Be( t / st )



Eq. 7.4

with H(t) the Heaviside step function. Equation 7.3 and its simplified version Eq. 7.4 assume real and frequency independent thermal parameters (heat capacity and thermal diffusivity =/(0C), with  the thermal conductivity and 0 the density, and hence do not explicitly take into account possible relaxation behavior of the response of temperature heat input, i.e. frequency dependence of the heat capacity. In principle, relaxing behavior of the heat capacity could be implicitly incorporated in the term Be(  t /

st

)

, provided the relaxing part of the temperature

dynamics and of the thermal expansion dynamics are occurring on the same time scale, and can be fitted together by an effective, stretched exponential. ISS signals for glycerol [52] and salol [42, 43] were indeed fitted with good quality by Eq. 7.3, confirming this scenario. For materials that exhibit structural relaxation behavior, the frequency dependence and non-zero imaginary part of the heat capacity, C(), and of the thermal expansion,  can also be explicitly plugged into the Fourier domain (,q-domain) expressions for the temperature and density response,

105

T ( x,  ) 

Q0 cos(qx) 2  q 2  i C 

Eq. 7.5

and

u ( x,  ) 

q cL2 Q0 sin(qx)

2 C  q 2  i  2  q 2 cL2 

Eq. 7.6

Assuming Debye behavior,

C ( )  C 

 ( )    

C0  C 1 i

 C

0   1 i 

Eq. 7.7

Eq. 7.8

with C∞ and ∞ the high frequency limit, and C0 and 0 the low frequency limit or static limit of the heat capacity and thermal expansion respectively, the spatiotemporal behavior of the temperature and density response to impulsive heating of relaxing materials can then be simply obtained by taking the 2D inverse Fourier transform (kx,t) of the frequency-wavenumber domain expression. A detailed parameter study of the combined effect of C∞ and ∞ on the temperature and density response to a sudden laser heating was made by J. Fivez et al [53] for a transient grating configuration. The values of parameters that were varied in the simulations are given in the caption of Figure 7.4, which was taken from that work.

106

density response (a.u.)

time (seconds) Figure 7.4

Simulated transient density response in the time domain, taking into account relaxing behavior of the longitudinal acoustic modulus, the specific heat capacity C() and the thermal expansion . From left to right: transient grating wavelengths 1, 15, 200 m. From top to bottom: =104, 105, 106 s-1. The simulations show the case-dependent effect of varying the relaxation frequency of the heat capacity (C=0 (top red curve), C=(middle green curve), C=3 (bottom black curve)).

The ISS response curves, which also give a qualitative idea on the behavior of TL response curves (the difference between both lies mainly in the acoustic part of the response, and to a lesser extent in the thermal diffusion decay), reveal relaxation features of both the specific heat capacity and the thermal

107

expansion in the responses. The initial temperature jump is followed by a decay due to the onset of relaxation, which goes along with an flow of energy from the fast degrees of freedom (vibrations) of the glassformer to the slow ones (molecular network conformation). The temperature evolution acts as a driving force for the thermal expansion, which in turn consists of an immediate response, followed by a slow, additional response. The temperature and density decrease ultimately decay again, due to thermal diffusion washing away the thermal contrast that was initiated by the optical light pattern. The double step in the density response is mainly visible when the relaxation time of C() and  is much longer than the thermal diffusion time. In Section 7.3 we will further discuss which scenarios are compatible with experimental observations.

205.6 K 210.1 K 215.3 K 219.3 K 223.9 K 228.5 K 233.2 K 242.4 K 251.6 K 261.1 K 270.1 K 279.2 K

1.4

normalized TL signal

1.2 1.0 0.8 0.6 0.4 0.2

279.2 K

205.6 K

0.0 -9

10

Figure 7.5

-8

10

-7

10

-6

-5

10 10 time (seconds)

-4

10

-3

10

-2

10

Normalized TL signals in glycerol at a selection of temperatures (symbols) and best fits (solid line) by Eq. 7.4.

108

Figure 7.5 presents experimentally determined TL spectra (symbols) for glycerol at a selection of temperatures between 205 K and 280 K, and the best fits (full line) by Eq. 7.4 with the stretching exponent fixed at 0.6 for glycerol from literature [52] and B fixed at one, while A, ac, and st as fitting parameters. As mentioned earlier, TL signals are expected to similar to ISS signals in terms of the signal features related to the heat capacity relaxation (double step in temperature) and the thermal expansion relaxation (double step in density increase). Figure 7.5 clearly shows a strongly (DC) temperature dependent structural relaxation like double density increase feature at intermediate times. With decreasing DC temperature, the second density step occurs at later times, till it is quenched by the thermal diffusion decay of the signal. With increasing DC temperature, it occurs at earlier times, till it is overlapping with the initial, fast part of the thermal expansion.

0.7 0.6

fDW

0.5 0.4 0.3 0.2

Figure 7.6

210

220 230 240 temperature (K)

250

Temperature dependence of the DW factor from TL responses (symbols) and average value (line).

Figure 7.6 shows the temperature dependence of the relative magnitude of the relaxing part compared to the total expansion B/(A+B), also termed as Debye-Waller factor (fDW), for which mode-coupling theory predicts a square root cusp at the cross temperature in glassforming liquids. K. Nelson et al. [52] have

109

reported the observation of such cusp for salol [63]and Ca4K6(NO3)14 (CKN) [64] in the temperature range from 270 to 275 K, and the absence of a cusp for glycerol [52]. E.H. Bentefour et al. reported indications of a cusp for glycerol by PPE technique. In this work, TL signals in the temperature range studied do not show a square root cusp in the temperature dependence of the DW factor, or distinctive behavior in the temperature dependence. Besides, the average DW factor in this work is 0.55, which is close to the value (0.66) observed by Paolucci et al. [52]for glycerol. Figure 7.7 shows the temperature dependence of average relaxation time (red squares), which was determined after fitting the signal with Eq. 7.4 from the resulting time constant st and the stretching parameter,  through

 

 st * (1/  ) 

Eq. 7.9

with (1/  ) the gamma function of  As mentioned earlier, fitting the data with Eq. 7.3 implicitly assumes effective stretched exponential time dependence of combined effects of thermal and structural relaxation occurring at a very similar time scale. The satisfactory fitting quality infers that, as in the results of Paolucci et al. [52], this assumption is justified. The temperature dependence of st was fitted by the Vogel–Fulcher- Tamman (VFT) form (red line), defined by



T

  0e

B T TVFT

Eq. 7.10

In the fit, values of 0 (1.7×10-15 seconds) and B (2210 K) were taken from Ref. [52]. TVFT, the only fit parameter was found to be TVFT =126.8±0.8 K. The best fit is represented by the red line in Figure 7.7. Except for the highest temperatures, for which the relaxation time is saturated due to the finite bandwidth 110

of the 10 ns-pump pulses, the Arrhenius curve obtained by TL is parallel with the one from ISS. Since both TL and ISS experiments probe the photothermal response of density to heating, one could expect the respective Arrhenius curves to be coincident. The offset between the TL and ISS curves is probably the result of a difference in temperature calibration between the two experiments, as a result of unavoidable and difficult to assess gradients between the probed and photothermally heated part of the sample in the excitation laser beam, and the temperature sensor, which was placed at some distance from the laser beam. 3

structural relaxation (TL) bestfit structural relaxation (ISTS) specific heat capacity (PPE) specific heat capacity  dielectric spectroscopy ultrasonic relaxation

seconds

10 2 10 1 10 0 10 -1 10 -2 10 -3 10 -4 10 -5 10 -6 10 -7 10 -8 10 -9 10 -10 10

glycerol 3.4

3.6

3.8

4.0

4.2

4.4

4.6

4.8

5.0

5.2

-1

1000/T (K )

. Figure 7.7

Arrhenius plot of the VFT behavior of the structural relaxation of glycerol determined by TL spectroscopy (symbols) and the best linear fit (red line).

The VFT behavior

determined by other different techniques, as summarized in Table 7.1, was also plotted as reference.

The VFT curve obtained by TL is also parallel with the VFT curve of other response parameters, i.e. specific heat capacity by 3- [65] and PPE [3], the dielectric permittivity by broadband dielectric spectroscopy [38], and ultrasonic [32] by heterodyne ultrasonic spectrometer, which confirms once more 111

that the fragility (curvature of the Arrhenius plot) for a given material is universal between the different response functions, in spite of the characteristic relaxation frequencies being somewhat different between different physical susceptibilities.

Relaxation dynamic

Measurement technique

Log 10()

B (K)

T0 (K)

Structural

TL (this work)

-14.76

2210

126.8

Structural

ISS [52]

-14.76

2210

133

Specific heat capaciy

PPE [3]

-11.9

1593

141.6

Specific heat capaciy

3 [65]

-14.6

2500

128

Ultrasonic

Ultrasonic spectrometer [32]

-14.39

2310

129

Dielectric

Dielectric spectroscopy [38]

-13.8

1740

137

Table 7.1

7.3

Comparison of VFT behavior of glycerol determined by different techniques

Photothermal fluorescence spectroscopy of supercooled glycerol

The TL spectroscopy results in Section 7.2 clearly show the effect of structural relaxation on the measured density response. The absence of an initial temperature relaxation induced signal overshootdecay sequence suggests that the relaxation time of the heat capacity is probably of the same order as the one of the thermal expansion coefficient so that the temperature relaxation effect is masked by the thermal expansion relaxation delayed density increase. However, the data do not allow to unravel the individual contributions of the heat capacity C() and the thermal expansion coefficient () to the overall density response. Therefore, in this section, an effort is made to extract purely the heat capacity relaxation from

112

the impulsive response of the temperature of the sample to pulsed laser heating, by employing the ultrafast fluorescence based thermometry approach that was introduced in Chapter 4. 

Simulation of C() effect on the impulsive temperature response

The analytical solution of the impulsive temperature response in a glassforming liquid to transient grating excitation, assuming Debye relaxation behavior of the heat capacity C() (cfr Eq 7.7) , has been shown to be [53]:

T ( x, t ) 

Q0eiqx 1  iC  ei1t  2  iC  ei2t  H (t )  C 1  2 

Eq. 7.11

with

i 2

 C0  C C  C )C  4q 2C  (q 2  (1  0 )C ) 2  C C 

i

 C0  C C  C )C  4q 2C  (q 2  (1  0 )C ) 2  C C 

1   q 2  (1 

2   q 2  (1  2 

and q the wavenumber of the transient grating. The corresponding temperature response was shown to be a combination of two exponential components that together make up an overshoot-decay type of behavior. At short times, the heat capacity (C∞) is small, due to the supercooled liquid being too slow to take up heat. As a result, the initial temperature rise is substantial. At later times, as a result of more energy flowing to evoke cooperative rearrangements of the amorphous network, the heat capacity increases towards C0, and the temperature decreases. The time it takes for the relaxation heat uptake to emerge increases with decreasing temperature, due to reduced molecular mobility, causing a slowing down of cooperative rearrangements. If the heat capacity would not undergo a relaxation process (C∞ towards C0), then 1=iC, and 2=iq2, which reduces the temperature response to a simple, transient mono-exponential decay function, reflecting the thermal diffusion induced smearing out of the temperature grating: 113

temperature response (a.u.)

T ( x, t ) 

Q0 eiqx  q2 t e H (t )  C

Eq. 7.12

T=280 K d=200 m

T=240 K d=200 m

T=200 K d=200 m

T=280 K d=50 m

T=240 K d=50 m

T=200 K d=50 m

T=280 K d=5 m

T=240 K d=5 m

T=200 K d=5 m

log10 time (seconds) Figure 7.8

Simulated temperature response in the time domain, taking into account relaxing behavior of the specific heat capacity C(). From left to right: =280, 240, 200 K. From top to bottom: transient grating spacing d=200, 50, 5 m. The simulations in each subfigure show the case-dependent effect of the heat capacity relaxation time. From left to right: 0.1, 0.4, 1, 2.5, and 10 times heat capacity value at the corresponding temperature, 11.5 ns, 3 s, and 0.47 s at 280, 240, 200 K respectively. The figure was extracted from the PhD thesis of Robbe Salenbien (KU Leuven, 2012), in which also all parameter values are listed.

114

Figure 7.8 shows a detailed parameter study of the effect of C)on the temperature response to sudden laser heating, which was made by Robbe Salenbien [8] for a transient grating configuration. The values of parameters that were varied in the simulations are given in the caption of Figure 7.8, which was taken from that work. Figure 7.8 clearly illustrates the expected bi-exponential decay, consisting of (i) a strongly temperature dependent overshooting regime (the first exponent) followed by (ii) thermal diffusion decay with characterized time determined by grating space, in temperature response curves for all cases, except at 200 K, where the heat capacity relaxation is so slow that the thermal diffusion has already washed away the temperature before the overshooting occurs. Besides, the thermal diffusion tail can be clearly shifted towards the later times, by increasing the grating space, while by decreasing the temperature, the tail of the relaxation part (overshooting tail) can be shifted towards later times as well, as a result of the increased relaxation time of heat capacity. The temporal separation of the two steps is more pronounced for larger grating spacing, up to 4 orders of magnitude in the top left corner (T=280 K, d=200 m). This suggests that during experiments, it would be preferable to implement a beam configuration with larger grating space, in order to distinguish the two exponential components in a wide temperature range. Although the simulations were conducted on the basis of a grating laser beam configuration, the conclusions are also applicable to the uniform laser beam configuration, where the influence of the beam spot size is analogous to that of the grating space in a transient grating experiment. For the case of spatially uniform heating, which is often employed in photothermal measurements to yield simple a 1D thermal conduction model, the wavenumber q can be set to zero, so that 1=0, and



2  i  (1  

 C0  C )C  C 

By substituting 1 and 2 into Eq. 7.12, the 1D temperature response of a glassforming liquid can be obtained:

115

T (t ) 

Q0  C

C C C  C  C0 C t  e   0  H (t )  C0  C 0  

Eq. 7.13 predicts an instantaneous temperature jump, T (t  0) 

Eq. 7.13

Q0 representing the temperature  C

response in the high frequency limit, followed by relaxation for T (t  ) to the value

Q0 , i.e. the  C0

temperature response for the low frequency limit. The timescale on which the relaxation takes place is determined by the relaxation frequency c. Figure 7.9 simulates the temporal temperature response of glycerol at different temperatures from 210 K to 300 K with a step of 10 K. The simulation clearly shows the shift of temperature response towards short times as the temperature is higher. As expected, the simulation presents purely the whole heat capacity relaxation part without convolving any thermal diffusion part, due to the applied 1D (infinite) laser spot. In this simulation C∞ and C0 was fixed at 1060 and 1850 J ∙ Kg-1 ∙ K-1 [8] respectively for each temperature, while the relaxation frequency c was calculated by the VFT curve (Figure 7.10) of glycerol that is represented by heat capacity and measured by 3- technique [65], as shown in Table 7.1.

116

temperature response (a.u.)

1.0 C 0.9 0.8

210 K

300 K

0.7 0.6 0.5

C0 1E-10 1E-9 1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 0.01 time (seconds)

Figure 7.9

0.1

Simulated time domain temperature responses of glycerol, induced by 1D uniform laser heating, by using Eq.7.13.

-1

10

relaxation time (seconds)

10

-3

10

-4

10

-5

10

-6

10

-7

10

-8

10

-9

10

7

10

6

10

5

10

4

10

3

10

2

10

relaxation frequency (Hz)

8

-2

10

1

10

210 220 230 240 250 260 270 280 290 300

temperature (K) Figure 7.10

VFT curve of relaxation time (left) and relaxation frequency (right) of the heat capacity of glycerol. The values Log10(0)=-14.6 seconds, B=2500 K, and TVFT= 128 K, taken from Ref. [65], were used to define the VFT curve by Eq. 7.8.

117



Experiment results and discussion

As mentioned earlier, one of the incentives for developing ultrafast fluorescence thermometry was to be able to determine the wideband relaxation behavior of the heat capacity, without interference of the relaxation of the density, which is a complicating factor in the analysis of structural relaxation experiments such as impulsive stimulated scattering and transient thermal lens detection. The experimental setup illustrated in Figure 4.7 in Chapter 4 was utilized to determine the nanosecond laser induced temperature evolution of the glycerol sample, dyed by rhodamine B (RhB), with a concentration of 2×10-6 molar acting as temperature probe, and copper chloride (CuCl2), with a concentration of 0.1 molar serving as absorber for the pump laser (1064 nm) light, enhancing the photothermal heating compared to pure glycerol. In the measurement setup, the pump and probe laser beams were aligned coaxially in the bulk of the sample, and diameter of the pump beam was around 3 mm, substantially larger than the one of the probe beam (0.8 mm). This was done in order to achieve a situation of uniform laser heating and temperature field on the time scale of interest, which is of benefit for temporally separating the heat capacity relaxation signal part from the thermal diffusion regime, as simulated in Figure 7.9. Note that, as in thermal lens spectroscopy, due to the impulsive excitation and finite time of thermal expansion and acoustic wave propagation, both volume (thermal expansion) and pressure (acoustic wave propagation) vary during a measurement. As a consequence, the determined the heat capacity (relaxation) was a combination of CV and CP (relaxation), as discussed by T. Christensen et.al. [66] The pulse energy of the pump laser and probe laser was 65 mJ/pulse and 25 J/pulse, respectively. 1D impulsive temperature response measurement was carried out for 6 temperatures, as depicted in Figure 7.11, where the stroboscopic measurement was arranged with the delay times between -300 ns and 10 ms. The pump laser was fired at 0.1 Hz in order to allow sufficient time (10 s before next pulse issued) for heat to diffuse

118

away and do subsequent measurements at equal starting temperature.

For each measurement the

determined temperature response is expected to be initiated by a quasi-instantaneous (≤10 ns, the duration of the longest laser pulse) temperature jump, with magnitude determined by the infinite frequency response value of the heat capacity as expected from Eq. 7.9. Given the pump beam radius of Rpb=1.5 mm, the temperature response is expected to decay with a thermal diffusion tail decay time of about decay=R2/(4glycerol)=6 seconds (with the thermal diffusivity of glycerol, glycerol=9.5×10-8 m2∙s-1), which is close to the experimental observation value, 4 seconds. Nevertheless, the thermal diffusion time is nearly 100 times longer than the 40 ms (VFT curve in Figure 7.10) relaxation time of the heat capacity at 210 K. Therefore one can expect the overshooting temperature response due to the relaxation of heat capacity can be clearly distinguished with the thermal diffusion regime in the complete investigated temperature range from 216 K to 261 K.

216.6 K

10

-8

10

-7

10

-6

10

-5

10

-4

t (seconds)

2.0 1.6 1.2 0.8 0.4 0.0 -0.4 -7 -3.0x10

T (K)

T (K)

1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -7 -3.0x10

10

-3

10

-2

10

-1

10

0

10

1

1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -7 -3.0x10

0.4

224.9 K

0.0

-3.0x10

-7

10

-8

10

-7

10

-6

10

-5

10

-4

t (seconds)

10

-3

10

-2

10

-1

10

0

10

1

1.2 0.4

234.2 K

-0.4 -3.0x10

-7

Figure 7.11

10

-8

10

-7

10

-6

10

-5

10

-4

t (seconds)

10

-8

10

-7

10

-6

10

1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -7 -3.0x10

-5

-3

10

-2

10

-1

10

0

10

1

10

-4

t (seconds)

10

-3

10

-2

10

-1

10

0

10

1

-3

10

-2

10

-1

10

0

10

1

-3

10

-2

10

-1

10

0

10

1

251.9 K

10

-8

10

-7

10

-6

10

-5

10

-4

t (seconds)

T (K)

T (K)

0.8 0.0

10

T (K)

T (K)

0.8

243.6 K

10

260.9 K 10

-8

10

-7

10

-6

10

-5

10

-4

t (seconds)

10

Experimental data for the impulse response of the temperature to sudden laser heating acquired at 6 temperatures between 216 K to 261 K.

119

Unlike these expectations, in Figure 11, only two measurements present an overshoot-like temperature response. At 243.6 K, the initial temperature jump decays to a lower value, with a relaxation time around 0.1 ms. The heat capacity relaxation time that one would expect from the extrapolated VFT [65] curve of glycerol, Figure 7.10, is 15 s. The curve collected at 260.9 K infers a relaxation time of about 800 ns, compared to a value of 400 ns predicted by VFT curve. The respective relaxation strengths are about 0.25 and 0.5, the latter value being close to the values of 0.5 observed for glycerol [65] by the 3 approach, 0.43 observed by [3] PPE technique, and 0.58 [8].

normalized TL signal

1.0 0.8 0.6 0.4 0.2 0.0 1E-9 1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 0.01

time (seconds)

Figure 7.12

Thermal lens spectroscopy of pure glycerol (black) at 205.6 K and CuCl 2 dyed glycerol (red) at 203.3 K.

This unexpected observation could in principle be due to the added amount of absorber CuCl2 of 0.1 molar being too high so that it has substantially changed the relaxation behavior of glycerol. This was examined by performing a thermal lens spectroscopy measurement of the dyed sample, as shown in Figure 7.12. This showed expected relaxation behavior. Thus, the speculation that CuCl2 would have

120

modified the relaxation behavior of glycerol can be excluded. Additional research efforts will be needed to address the unexplained absence of relaxation in the fluorescence thermometry data.

7.4

Conclusion

In this Chapter, we have investigated the density and temperature response to impulsive laser heating of supercooled glycerol, by using two types of time-resolved photothermal spectroscopy. In a first approach, a nanosecond laser induced thermal lens scheme was implemented to study the temperature dependent structural relaxation behavior, which is a combined effect of heat capacity relaxation and thermal expansion relaxation, of a supercooled glycerol. The obtained VFT curve of the structural relaxation time is parallel with that of other response parameters, which confirms previous observations that the fragility for a given material is universal between the different response functions. Additionally, in order to detect the purely the relaxation of heat capacity, in a second approach, ultrafast fluorescence thermometry was utilized to determine the temperature dependence of the temperature response to impulsive heating of glycerol. A priori, extrapolating information on the frequency dependence of the heat capacity from low frequency measurements, an initial overshoot of the temperature response was expected. However, indications of an overshoot were only observed at 244 K and 261 K. The duration of the overshoot, which is expected to correspond with the relaxation time of the heat capacity, was about 0.1 ms for the 244 K data, compared to 15 s predicted by the VFT curve for glycerol, as extrapolated from low frequency data. The latter one showed an overshoot duration/relaxation time of about 800 ns, compared to 400 ns predicted by the VFT curve.

121

References

1.

Glorieux, C., K. Nelson, G. Hinze, and M. Fayer, Thermal, structural, and orientational relaxation of supercooled salol studied by polarization-dependent impulsive stimulated scattering. The Journal of chemical physics, 2002. 116(8): p. 3384-3395.

2.

Leys, J., M. Wübbenhorst, C.P. Menon, R. Rajesh, J. Thoen, C. Glorieux, P. Nockemann, B. Thijs, K. Binnemans, and S. Longuemart, Temperature dependence of the electrical conductivity of imidazolium ionic liquids. The Journal of chemical physics, 2008. 128(6): p. 064509.

3.

Bentefour, E.H., C. Glorieux, M. Chirtoc, and J. Thoen, Thermal relaxation of glycerol and propylene glycol studied by photothermal spectroscopy. The Journal of chemical physics, 2004. 120(8): p. 3726-3731.

4.

Sinha, G., C. Glorieux, and J. Thoen, Broadband dielectric spectroscopy study of molecular dynamics in the glass-forming liquid crystal isopentylcyanobiphenyl dispersed with aerosils. Physical Review E, 2004. 69(3): p. 031707.

5.

Glorieux, C., J. Descheemaeker, J. Vandenbroeck, J.-P. Groby, L. Boeckx, P. Khurana, and N. Roozen, Temperature and frequency dependence of the visco-elasticity of a poro-elastic layer. Applied Acoustics, 2014. 83: p. 123-126.

6.

van den Berg, O., W.G. Sengers, W.F. Jager, S.J. Picken, and M. Wübbenhorst, Dielectric and fluorescent probes to investigate glass transition, melt, and crystallization in polyolefins. Macromolecules, 2004. 37(7): p. 2460-2470.

7.

Lupaşcu, V., H. Huth, C. Schick, and M. Wübbenhorst, Specific heat and dielectric relaxations in ultra-thin polystyrene layers. Thermochimica Acta, 2005. 432(2): p. 222-228.

122

8.

Salenbien and Robbe, Photoacoustic and photothermal investigation of ultrathin structures and viscoelastic materials. Dec, 2012.

9.

Lunkenheimer, P., A. Pimenov, M. Ssel, Y.G. Goncharov, R. Böhmer, Loidl, and A., Fast Dynamics of Glass-Forming Glycerol Studied by Dielectric Spectroscopy. Physical Review Letters, 1996. 77(2): p. 318-321.

10.

Lunkenheimer, P. and A. Loidl, Dielectric spectroscopy of glass-forming materials: alpharelaxation and excess wing. Chemical physics, 2002. 284(02): p. 205-219.

11.

Ediger, M.D., C.A. Angell, and S.R. Nagel, Supercooled Liquids and Glasses. J.phys.chem, 1996. 100(31): p. 13200-13212.

12.

Angell, C.A., Formation of glasses from liquids and biopolymers. Science, 1995. 267(5206): p. 1924-1935.

13.

Ediger, M.D., C. Angell, and S.R. Nagel, Supercooled liquids and glasses. The Journal of Physical Chemistry, 1996. 100(31): p. 13200-13212.

14.

Wong, J. and C.A. Angell, Glass: structure by spectroscopy. 1976: M. Dekker New York.

15.

Dixon, P.K., L. Wu, S.R. Nagel, B.D. Williams, and J.P. Carini, Scaling in the Relaxation of Supercooled Liquids. Physical Review Letters, 1990. 65(9): p. 1108-1111.

16.

Bengtzelius, U., W. Gotze, and A. Sjolander, Dynamics of supercooled liquids and the glass transition. Journal of Physics C: solid state Physics, 1984. 17(33): p. 5915.

17.

Debenedetti, P.G. and F.H. Stillinger, Supercooled liquids and the glass transition. Nature, 2001. 410(6825): p. 259-267.

18.

Gotze, W. and L. Sjogren, Relaxation processes in supercooled liquids. Reports on progress in Physics, 1992. 55(3): p. 241.

19.

Dyre, J.C., Colloquium: The glass transition and elastic models of glass-forming liquids. Reviews of modern physics, 2006. 78(3): p. 953.

123

20.

He, H. and M. Thorpe, Elastic properties of glasses. Physical Review Letters, 1985. 54(19): p. 2107.

21.

Goldstein, M., Viscous liquids and the glass transition: a potential energy barrier picture. The Journal of chemical physics, 1969. 51(9): p. 3728-3739.

22.

Adam, G. and J.H. Gibbs, On the temperature dependence of cooperative relaxation properties in glass‐forming liquids. The Journal of chemical physics, 1965. 43(1): p. 139-146.

23.

Stevenson, J.D., J. Schmalian, and P.G. Wolynes, The shapes of cooperatively rearranging regions in glass-forming liquids. Nature Physics, 2006. 2(4): p. 268-274.

24.

Donth, E., The size of cooperatively rearranging regions at the glass transition. Journal of NonCrystalline Solids, 1982. 53(3): p. 325-330.

25.

Williams, M.L., R.F. Landel, and J.D. Ferry, The temperature dependence of relaxation mechanisms in amorphous polymers and other glass-forming liquids. Journal of the American Chemical Society, 1955. 77(14): p. 3701-3707.

26.

Novikov, V. and A. Sokolov, Poisson's ratio and the fragility of glass-forming liquids. Nature, 2004. 431(7011): p. 961-963.

27.

Martinez, L.-M. and C. Angell, A thermodynamic connection to the fragility of glass-forming liquids. Nature, 2001. 410(6829): p. 663-667.

28.

Novikov, V., Y. Ding, and A. Sokolov, Correlation of fragility of supercooled liquids with elastic properties of glasses. Physical Review E, 2005. 71(6): p. 061501.

29.

Meng, Y. and S.L. Simon, Pressure relaxation of polystyrene and its comparison to the shear response. Journal of Polymer Science Part B: Polymer Physics, 2007. 45(24): p. 3375-3385.

30.

Meng, Y., P. Bernazzani, P.A. O’Connell, G.B. McKenna, and S.L. Simon, A new pressurizable dilatometer for measuring the time-dependent bulk modulus and pressure-volume-temperature properties of polymeric materials. Review of Scientific Instruments, 2009. 80(5): p. 053903.

124

31.

Sane, S. and W.G. Knauss, The time-dependent bulk response of poly (methyl methacrylate). Mechanics of Time-Dependent Materials, 2001. 5(4): p. 293-324.

32.

Jeong, Y.H., S.R. Nagel, and S. Bhattacharya, Ultrasonic investigation of the glass transition in glycerol. Physical Review A, 1986. 34(1): p. 602.

33.

Büchner, B. and P. Korpiun, Frequency-dependent specific heat near the glass transition measured with the photoacoustic effect. Applied Physics B, 1987. 43(1): p. 29-33.

34.

Kojima, S., Study of Phase Transitions by Frequency Dependent Photoacoustic Measurements, in Physical Acoustics, O. Leroy and M. Breazeale, Editors. 1991, Springer US. p. 399-404.

35.

Yang, Y. and K.A. Nelson, Impulsive stimulated light scattering from glass‐forming liquids. I. Generalized hydrodynamics approach. The Journal of chemical physics, 1995. 103(18): p. 77227731.

36.

Yan, Y.X., E.B. Gamble Jr, and K.A. Nelson, Impulsive stimulated scattering: General importance in femtosecond laser pulse interactions with matter, and spectroscopic applications. The Journal of chemical physics, 1985. 83(11): p. 5391-5399.

37.

N B, O., C. T, and D. J C, Time-temperature superposition in viscous liquids. Physical Review Letters, 2001. 86(7): p. 1271-1274.

38.

Stickel, F., E.W. Fischer, and R. Richert, Dynamics of glass ‐ forming liquids. II. Detailed comparison of dielectric relaxation, dc ‐ conductivity, and viscosity data. The Journal of chemical physics, 1996. 104(5): p. 2043-2055.

39.

Stickel, F., E.W. Fischer, and R. Richert, Dynamics of glass‐forming liquids. I. Temperature‐ derivative analysis of dielectric relaxation data. The Journal of chemical physics, 1995. 102(15): p. 6251-6257.

40.

Debenedetti, P.G. Amp, and F.H. Stillinger, review article Supercooled liquids and the glass transition. Nature, 2001(6825): p. 259-267.

125

41.

Mukta, T. and K.S. Schweizer, The influence of shape on the glassy dynamics of hard nonspherical particle fluids. II. Barriers, relaxation, fragility, kinetic vitrification, and universality. Journal of Chemical Physics, 2009. 130(24): p. 244907-244907-9.

42.

Yang, Y. and K.A. Nelson, Structural Relaxation of Supercooled Liquids from Impulsive Stimulated Light Scattering. Acs Symposium, 2009.

43.

Yang, Y. and K.A. Nelson, Structural Relaxation and Orientational Relaxation Dynamics in Salol Studied by Impulsive Stimulated Light Scattering. American Physical Society, 1996.

44.

Chua, Y.Z., G. Schulz, E. Shoifet, H. Huth, R. Zorn, J.W.P. Scmelzer, and C. Schick, Glass transition cooperativity from broad band heat capacity spectroscopy. Colloid & Polymer Science, 2014. 292(8): p. 1893-1904.

45.

Birge, N.O., Specific-Heat Spectroscopy of Glycerol and Propylene-Glycol near the GlassTransition. Physical Review B, 1986. 34(3): p. 1631-1642.

46.

Birge, N.O., P.K. Dixon, and N. Menon, Specific heat spectroscopy: Origins, status and applications of the 3 omega method. Thermochimica Acta, 1997. 305: p. 51-66.

47.

Lu, Z.P., Y. Li, S.C. Ng, and Y.P. Feng, Study of glass transition of metallic glasses by temperature-modulated differential scanning calorimetry (MDSC). Thermochimica Acta, 2000. 357: p. 65-69.

48.

Boller, A., C. Schick, and B. Wunderlich, Modulated differential scanning calorimetry in the glass transition region. Thermochimica Acta, 1995. 266: p. 97-111.

49.

Song, M., A. Hammiche, H.M. Pollock, D.J. Hourston, and M. Reading, Modulated Differential Scanning Calorimetry .1. A Study of the Glass-Transition Behavior of Blends of Poly(Methyl Methacrylate) and Poly(Styrene-Co-Acrylonitrile). Polymer, 1995. 36(17): p. 3313-3316.

50.

Bentefour, E.H., C. Glorieux, M. Chirtoc, and J. Thoen, Thermal relaxation of glycerol and propylene glycol studied by photothermal spectroscopy. Journal of Chemical Physics, 2004. 120(8): p. 3726-3731. 126

51.

Bentefour, E.H., C. Glorieux, M. Chirtoc, and J. Thoen, Broadband photopyroelectric thermal spectroscopy of a supercooled liquid near the glass transition. Journal of applied physics, 2003. 93(12): p. 9610-9614.

52.

Paolucci, D.M. and K.A. Nelson, Impulsive stimulated thermal scattering study of structural relaxation in supercooled glycerol. Journal of Chemical Physics, 2000. 112(15): p. 6725-6732.

53.

Fivez, J., R. Salenbien, M.K. Malayil, W. Schols, and C. Glorieux. Dynamics of specific heat and other relaxation processes in supercooled liquids by impulsive stimulated scattering. in Journal of Physics: Conference Series. 2011.

54.

Gordon, J., R. Leite, R.S. Moore, S. Porto, and J. Whinnery, Long‐transient effects in lasers with inserted liquid samples. Journal of applied physics, 1965. 36(1): p. 3-8.

55.

Kitamori, T., M. Tokeshi, A. Hibara, and K. Sato, Peer reviewed: Thermal lens microscopy and microchip chemistry. Analytical Chemistry, 2004. 76(3): p. 52 A-60 A.

56.

Snook, R.D. and R.D. Lowe, Thermal lens spectrometry. A review. Analyst, 1995. 120(8): p. 2051-2068.

57.

Lima, S., J. Sampaio, T. Catunda, A. Bento, L. Miranda, and M. Baesso, Mode-mismatched thermal lens spectrometry for thermo-optical properties measurement in optical glasses: a review. Journal of Non-Crystalline Solids, 2000. 273(1): p. 215-227.

58.

Lima, S., T. Catunda, R. Lebullenger, A. Hernandes, M. Baesso, A. Bento, and L. Miranda, Temperature dependence of thermo-optical properties of fluoride glasses determined by thermal lens spectrometry. Physical Review B, 1999. 60(22): p. 15173.

59.

Terazima, M. and N. Hirota, Population lens in thermal lens spectroscopy. The Journal of Physical Chemistry, 1992. 96(18): p. 7147-7150.

60.

Malik, A.K. and W. Faubel, Photothermal and light emitting diodes as detectors for trace detection in capillary electrophoresis. Chemical Society Reviews, 2000. 29(4): p. 275-282.

127

61.

Shen, J., R.D. Lowe, and R.D. Snook, A model for cw laser induced mode-mismatched dual-beam thermal lens spectrometry. Chemical physics, 1992. 165(2): p. 385-396.

62.

Schott, A., Optical glass data sheets. Retrieved February, 2011.

63.

Yang, Y. and K.A. Nelson, Impulsive stimulated light scattering from glass‐forming liquids. I. Generalized hydrodynamics approach. Journal of Chemical Physics, 1995. 103(18): p. 7722-7731.

64.

Yang, Y. and K.A. Nelson, Impulsive stimulated thermal scattering study of α relaxation dynamics and the Debye–Waller factor anomaly in Ca0.4K0.6(NO3)1.4. Journal of Chemical Physics, 1996. 104(14): p. 5429-5436.

65.

Birge, N.O., Specific-heat spectroscopy of glycerol and propylene glycol near the glass transition. Physical Review B, 1986. 34(3): p. 1631.

66.

Christensen, T., N.B. Olsen, and J.C. Dyre, Conventional methods fail to measure c p (ω) of glass-forming liquids. Physical Review E, 2007. 75(4): p. 041502.

128

Chapter 8

General conclusion and perspective

8.1

Conclusion

In this work, an all-optical ultrafast thermometry technique, allowing remote temperature determination down to 10 ns, has been developed. The fast response the fluorescence characteristics of a dye to temperature changes was exploited by making use of a stroboscopic approach. Neural network recognition based calibration was used to extract temperature from the fluorescence spectrum, with high robustness against mechanical, electrical and optical disturbances, by simultaneously exploiting multiple features of fluorescence spectra of the thermosensitive dye.

The concept was implemented on a

rhodamine B dyed mixture of copper chloride and glycerol, of which the full temperature evolution induced by a nano-second laser (10 ns) could be determined with a sensitivity of 15 K∙ Hz-1/2 with 100 MHz bandwidth. The ultrafast fluorescence based thermometry method was implemented in a photothermal application, both in (i) time domain, where the gradual temperature rise resulting from accumulation of heat supplied by repetitive laser pulses was carried out, and in (ii) frequency domain, where spatially resolved detection of photothermally induced temperature oscillations, with spatial resolution determined by the entrance aperture of the fluorescence collecting optical fiber, 100 micrometers in this work (ideally optical 129

resolution can be achieved), was demonstrated by determining the frequency dependent photothermal response of the a sample, at different depths inside the sample. In addition, lock-in based photothermal photoluminescence spectroscopy, with potential applications in the field of NDT, e.g. for remote thermal property characterization or layer thickness determination, was also illustrated in frequency domain by taking advantage of the linear temperature response of the PL integration intensity of CdSe/ZnS QDs at ambient conditions. Furthermore, driven by the research group's interest in the complex relaxation behavior in glass forming liquids, two types of time-resolved photothermal spectroscopy were explored, with the goal of extracting the relaxation behavior of the specific heat capacity in supercooled glycerol. In a first approach, a nanosecond laser induced thermal lens technique was implemented. The TL signals clearly show strongly temperature dependent structural relaxation behavior, which involves the combined effect of heat capacity relaxation and thermal expansion relaxation. The temperature dependence of the Debye-Waller (DW) factor of glycerol, extracted from the relaxation strength of the response, is quite flat, with an average value around 0.55, consistent with the value obtained by D. Paolucci and K. Nelson et al. by means of the impulsive stimulated scattering technique. The VFT curve of the structural relaxation time obtained by the thermal lens method is parallel with the one of other response parameters. This confirms previous observations that the fragility (the curvature of the Arrhenius plot) for a given material is universal between the different response functions, in spite of the characteristic relaxation frequencies being somewhat different between different physical susceptibilities. In a second approach, the developed ultrafast fluorescence thermometry was used to determine the DC temperature dependence of the temperature response to impulsive heating of glycerol, in which RhB and CuCl2 were doped, serving as temperature probe and absorber of pump light respectively. Against expectations, indications of an overshoot temperature response were only observed at 243.6 K and 260.9

130

K. For the first temperature overshoot the duration was around 0.1 ms, compared to 15 s predicted by the VFT curve of glycerol. The second temperature showed an overshoot duration/relaxation time of approximately 800 ns, compared to 400 ns predicted by VFT curve. The relaxation strengths for these two temperatures were about 0.25 and 0.5 respectively. The latter value is comparable to values reported for glycerol by other techniques: 0.5 by the 3- approach, 0.43 by the photopyroelectric technique, and 0.58 by impulsive stimulated scattering.

8.2

Perspective

Additional research efforts will be needed to address the absence, or suppression of the expected overshoot of the temperature response to impulsive heat input. This unexpected observation could be due to the pulsed pump laser having induced some photochemical reaction between RhB and CuCl2 as a result of the rather high concentration of CuCl2, the high absorption of CuCl2 for the pump light, and the high peak power of the pulsed pump laser, so that RhB was not only sensing the temperature variations but also fast photochemical effects. One issue complicating the interpretation of the results is the poor pointing stability of the pump laser. Given the stroboscopic nature of the measurement, the full temperature evolution was constructed by measurements for different delay times performed sequentially with 10 seconds in between. In this scheme, geometrical drifting of the pump laser beam position induces a change of magnitude of the temperature induced by the pump laser throughout the acquisition of the temperature evolution versus pump-probe delay time, and thus to fluctuations in the determined temperature response curve. The possible occurrence of CuCl2 induced photochemical reactions could be examined by measuring the time-resolved absorbance spectrum of the sample with white probe light. If any photochemical reaction between CuCl2 and RhB is triggered by the pump laser, then one can expect to observe a significant 131

change in the absorbance change. Alternatively, chemically more inert and physically less invasive gold nanoparticles can be employed as optical absorbers avoiding the usage of CuCl2. The effect of pointing instability of the laser beam on the measurement results could be estimated by monitoring the long term pump - probe beam geometrical position with a CCD camera, in parallel with the collection of fluorescence thermometry data. Once a correlation is established between the respective fluctuations, this can be exploited to normalize the data and reduce the noise level. Besides, it would be very interesting to employ two fluorescent dyes in parallel, order to implement stroboscopic fluorescence thermometry with one dye acting as a temperature sensor, and the other one as a reference probe. For the selection of appropriate dyes for a two-dye stroboscopic fluorescence thermometry system, there are several important factors that should be taken into account. First, a temperature-insensitive dye is preferred as a reference probe since it will maximize the temperature sensitivity of the ratio of the temperature-sensing dye and the reference dye, respectively. Secondly, the two dyes should share similar absorption characteristics yet quite different emission spectra, which allows for fluorescence excitation of two dyes by a single light source, and, more importantly, efficient separation of two emission band in the spectrum recorded by a spectrometer. Given that RhB was proven to be a suitable temperature sensor, sulforhodamine-101 (SR101) and rhodamine 110 (Rh110) are interesting candidates as reference dyes.

132

Curriculum Vitae

Personal data Liwang Liu Place and date of birth: Jiangsu Province, December 8th, 1986 Nationality: Chinese Education 

2011-2015: KU Leuven, Belgium Laboratory for Soft Matter and Biophysics/ Ph. D’s degree in Science: Physics



2008- 2011: Soochow University, P.R. China Institute of Modern optical technologies/Master’s degree in Optical Engineering



2004-2008: Northwestern Polytechnical University, P.R. China School of Science /Bachelor’s degree in Optical Information Science and Technology

133

List of publications

Publications in peer-reviewed journals



Liu Liwang, Zhong Kuo, Munro Troy, Alvarado Salvador, Côte Renaud, Creten Sebastiaan, Fron Eduard, Ban Heng, Van der Auweraer Mark, Roozen N.B., Matsuda Osamu and Glorieux Christ (2015). Wideband fluorescence-based thermometry by neural network recognition: Photothermal application with 10 nanoseconds temporal resolution. Journal of Applied Physics.(in press)



Depuydt Daphne, Liu Liwang, Glorieux Christ, Dehaen Wim, and Binnemans Koen (2015). Homogeneous liquid-liquid extraction of metal ions with nonfluorinated bis(2-ethylhexyl)phosphate ionic liquids having a lower critical solution temperature in combination with water. Chemical Communications, 51(75), 14183-14186.



Liu Liwang, Creten Sebastiaan, Firdaus Yuliar, Flores Cuautle Jose Jesus Agustin, Kouyaté Mansour, Van der Auweraer Mark, Glorieux Christ (2014). Fluorescence spectra shape based dynamic thermometry. Applied Physics Letters. 104, 031902.

134

Scientific conferences attended



Liu Liwang and Glorieux Christ (2015). Development of ultrafast all-optical techniques for remote temperature detection and thermophysical property determination, 18th International Conference on Photoacoustic and Photothermal Phenomena (ICPPP18), Novi Sad, Serbia, September 6-10, 2015. (oral presentation)



Liu Liwang, Zhong Kuo, and Glorieux Christ (2015). Non-contact photothermal characterization of thermophysical properties by using fluorescence based thermometry, 6th International Conference on Emerging Technologies in Non Destructive Testing (ETNDT6), Brussels, Belgium - 27-29 May 2015. (oral presentation)



Liu Liwang, Creten Sebastiaan, Glorieux Christ (2013). Using fluorescence spectral shape based dynamic thermometry to determine a photothermally induced temperature evolution, 17th International Conference on Photoacoustic and Photothermal Phenomena (ICPPP17), Suzhou, China, 20-24 October 2013. (oral presentation)



Liu Liwang, Xiong Jichuan, Wang Chinhua, Glorieux Christ (2013). Thermal conductivity depth profiling of hardened solids using infrared photothermal radiometry technique, Optical Measurement Techniques for Structures & Systems (OPTIMESS 2012): 243-252, Antwerp, Belgium, 04-05 April, 2012. (poster and proceedings)

135