ORIENTATIONAL DYNAMICS OF WATER AT HYDROPHOBIC INTERFACES

By

Florian Figge

A DISSERTATION

Submitted to

Michigan State University

in partial fulfillment of the requirements

for the degree of

Physics – Doctor of Philosophy

2018

ABSTRACT

ORIENTATIONAL DYNAMICS OF WATER AT HYDROPHOBIC INTERFACES

By

Florian Figge

Water surfaces play an important role in many biological, chemical and physical processes. Many

reactions occur at the water interface and some are even favored to happen there. Microscopically,

many of the interfacial properties, such as the structure or dynamics of water molecules, are related

to the termination of the dynamic hydrogen-bond network that is present in bulk water.

Time resolved pump-probe vibrational sum-frequency generation spectroscopy at the water/hy-

drophobic interface in combination with total internal reflection geometry allows the observation of

the surface dynamics with a large signal-to-noise ratio. Pump and probe beams are both resonant

with the surface-specific dangling OH stretch vibration. The pump-induced orientational anisotropy

is measured by pumping with polarizations parallel and perpendicular to the probe polarization.

The measurements show that the decay of the dangling OH stretch excitation is dominated by a
1.61± 0.10ps jump to a hydrogen-bonded configuration. The dangling OH jump time at hydropho-
bic interfaces is twice as fast as a jump between hydrogen-bonded configurations in bulk water and

about 50% slower than what is reported for the water/air interface.

Upon introduction of salt, changes in structure of water are observed in the bulk and at aqueous

interfaces. This is reflected in macroscopic observables such as altered viscosity or surface tension.

Moreover, research on bulk salt solutions shows significant changes in dynamics that, in some

cases, extend beyond the first solvation shell and might be expected to alter the interfacial dynamics.

However, our experiments suggest that the reorientation dynamics associated with the jump from a

dangling to a bonded OH is not affected by the dissolved salts used in our research.

ACKNOWLEDGEMENTS

First of all, I would like to thank my advisor John McGuire. I have never met anybody this dedicated

to research and science. I am very appreciative of his support, patience, guidance, and wisdom

throughout the years.

I want to thank my fellow grad students Cheng Sun for the introduction to the lab and Yanhao Tang

for many helpful discussions. Completing our research group, I want to thank postdoc Shunhao

Xiao for the work that he put into the experimental setup that I eventually took over and was doing

my dissertation research with.

A lot of appreciation goes out to the staff at the physics department of Michigan State University,

especially secretary Kim Crosslan, the machine shop crew, Reza Loloee, and whoever paid for the

pizza at the CORE-CM (Center of Research Excellence in Complex Materials) seminars.

I would like to thank my dear friends, close by and far away. Even though I have not seen many of

them recently but I do not know what I would have done without them.

Of course, this list would be incomplete without my parents, Gabriele Göhringer and Manfred Figge.

Without their unlimited support, I would be nowhere near where I am today. Their encouraging

words via Skype were essential during grad school.

Last, but not least, I would like to thank my girlfriend Celina Wanek for her unconditioned love

and support that ranged from keeping me fed to meeting me for little coffee dates during the day at

times when I left our apartment in the early morning and did not return until midnight on a regular

basis. Furthermore, I want to thank her for all the countless sacrifices that she has made during the

last few years. I will never forget them.

iii

TABLE OF CONTENTS

LIST OF TABLES .

LIST OF FIGURES .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vi

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

CHAPTER 1

INTRODUCTION .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1 Structure and orientational dynamics in bulk water . . . . . . . . . . . . . . . . .
.
1.2 Structure and orientational dynamics at the water interface . . . . . . . . . . . . .

1
2
3

CHAPTER 2 THEORY .

.
2.1 Light-matter interaction .

.

.

.

. . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . .

Fresnel equations and transmission coefficients
2.2.6.1
2.2.6.2
2.2.6.3
2.2.6.4

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.1 Non-linear polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Sum-frequency generation spectroscopy . . . . . . . . . . . . . . . . . . . . . . .
Surface specificity of SFG . . . . . . . . . . . . . . . . . . . . . . . . . .
Polarization and susceptibility . . . . . . . . . . . . . . . . . . . . . . . .

6
6
7
8
8
2.2.1
2.2.2
9
2.2.3 Connection between macroscopic and microscopic description . . . . . . . 10
2.2.4 Measured SFG intensity . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.5 Laser beam polarization and non-zero tensor elements of χ(2)
. . . . . . . 11
. . . . . . . . . . . . . . . 14
2.2.6
. 14
s-polarized light
p-polarized light
. 15
Fresnel coefficients at the angle of total internal reflection . . . . 16
Fresnel coefficients for experiments involving a monolayer . . . . 18
2.3 Time- and polarization-resolved SFG spectroscopy . . . . . . . . . . . . . . . . . 20
. 20
In-plane orientational dynamics . . . . . . . . . . . . . . . . . . 21
. 23
. . . . . . . 25
. . . . . . . . . . . . . . . . 29
. . . . . . . . . . . . . . . . . . . . . . . 29
Intramolecular vibrational relaxation - kd/b
IV R . . . . . . . . . . . . . . . . . 30
Intramolecular energy transfer - kF/B
. . . . . . . . . . . . . . . . . . . . 30
IET
2.4.3.1 Calculation of γ and δ . . . . . . . . . . . . . . . . . . . . . . . 31
IET and kB
2.4.3.2 Calculation of kF
. . . . . . . . . . . . . . . . . . . 32
IET
p−p . . . . . . . . . . . . . . . . . . . . . . . 32
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.3.1.1
2.3.1.2 Out-of-plane orientational dynamics . . . . . . . . . . . . . . .
2.3.1.3 Mathematical description of orientational dynamics

2.4.1 Molecular reorientation - kF/B
jump
2.4.2
2.4.3

2.4.4 Dipole-dipole interactions - ki− j
Isotopic dilution .

2.3.1 Orientational dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.4 Decay mechanisms of the dangling OH stretch mode

2.5

.

.

.

.

CHAPTER 3 EXPERIMENTAL SETUP . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.1 Laser source . .
3.2 Optical parametric amplifier (OPA) . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3
IR resonant with the dangling OH stretch vibration . . . . . . . . . . . . . . . . . 38

. .

.

.

.

iv

.

.
.

IR pump and probe beams

. . . . . . . . . . . . . . . . . . . . . . . . .

3.4.1
3.4.2 VIS 800nm . .
.

3.5 Signal detection . .

3.6 General procedure .
.
3.7 Sample preparation .

3.5.1 Cosmic ray removal . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.4 Sum frequency generation with IR pump/probe and VIS 800nm . . . . . . . . . . . 39
. 39
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
. 45
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.7.1 Growing the octadecyltrichlorosilane self-assembled monolayer (OTS SAM) 49
. . . . . . . . . . . . . . . . . . . . . . . . . 50
3.7.2 Contact-angle measurement
3.7.3 Removing the OTS SAM . . . . . . . . . . . . . . . . . . . . . . . . . .
. 50
. . . . . . . . . . . . . . . . . . . . . . . . . 51
3.7.4
. 52
. 52
. 52

Preparation of salt solutions
Sodium chloride . . . . . . . . . . . . . . . . . . . . . . . . .
3.7.4.1
3.7.4.2
Sodium iodide . . . . . . . . . . . . . . . . . . . . . . . . . .
3.7.4.3 Magnesium sulfate . . . . . . . . . . . . . . . . . . . . . . . .

.
.

.
.

.
.

CHAPTER 4 ORIENTATIONAL DYNAMICS OF WATER AT AN EXTENDED HY-

.

.

.

.

.

.

.
Isotopically Diluted Water

DROPHOBIC INTERFACE . . . . . . . . . . . . . . . . . . . . . . . . . . 53
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.1 Neat water .
4.2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.3 Experimental rates and uncertainties . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.3.1 Determination of the uncertainty of kF
IET . . . . . . . . . . . . . . . . . . . 62
4.3.1.1 Additional constrains for kF
IET based on the research literature . . 63
4.3.2 Determination of the uncertainty of kF
jump . . . . . . . . . . . . . . . . . . 65
4.3.3 Determination of the in-plane reorientation . . . . . . . . . . . . . . . . . 65
4.4 Summary and interpretation by MD simulations . . . . . . . . . . . . . . . . . . . 67
4.4.1 MD simulations and interpretation . . . . . . . . . . . . . . . . . . . . . . 67

CHAPTER 5 SALT SOLUTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.1 Previous research on bulk salt solutions
. . . . . . . . . . . . . . . . . . . . . . . 70
5.2 Previous research on interfaces of salt solutions . . . . . . . . . . . . . . . . . . . 73
5.3 Rotational dynamics of aqueous salt solutions at an extended hydrophobic interface
79
3M NaCl solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.3.1
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.3.2
3M NaI solution .
5.3.3 MgSO4 solution .
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.4 Summary and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

CHAPTER 6 SUMMARY AND OUTLOOK . . . . . . . . . . . . . . . . . . . . . . . . 88
6.1 Orientational dynamics of neat water at hydrophobic interfaces . . . . . . . . . . . 88
6.2 Orientational dynamics of salt solutions at hydrophobic interfaces
. . . . . . . . . 88
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.3 Outlook .

.

.

.

.

.

.

.

.

APPENDIX .

.

.

.

BIBLIOGRAPHY .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. 92

v

LIST OF TABLES

Table 2.1: Summary of the rate constants of the dominant processes involved in relaxation

of the pump-probe SFG signal. . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

Table 3.1: Galvanometer power supply (Topward Dual-Tracking DC power supply 63061)

settings .

.

.

.

. .

. .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

Table 4.1: Fast and slow time constants extracted from a fit to the data shown in figure 4.2 . 54
Table 4.2: Time constants extracted from fits to ∆β and ¯β(cid:48) shown in figure 4.3. * = fixed

. .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

parameter. .

. .

. .

Table 4.3: Summary table of experimentally obtained and calculated values.

. . . . . . . . 67

Table 5.1: Summary of all time constants. * = fixed parameter. . . . . . . . . . . . . . . . . 79

Table 5.2: Summary table of experimentally obtained and calculated values for NaCl.

. . . 81

Table 5.3: Summary table of experimentally obtained and calculated values for NaI.

. . . . 84

Table 5.4: Summary table of experimentally obtained and calculated values for MgSO4.

. . 86

Table 5.5: Summary of experimentally obtained constants. . . . . . . . . . . . . . . . . .

. 86

vi

LIST OF FIGURES

Figure 1.1: Polarization of a water molecule and the resulting hydrogen bonds with
neighboring water molecules. The oxygen of the central water molecule is
considered to be an acceptor of two hydrogen bonds (with the water molecules
in the lower left corner of the figure) and a donor of two hydrogen bonds (with
the water molecules in the top and the lower right corner of the figure).

. . . .

.

1

Figure 1.2: Schematic reorientation in bulk illustrated by Laage and Hynes[1]. a-b)
Initially fluctuations lead to an overcoordinated water molecule (Oa) and an
undercoordinated water molecule (Ob). The water molecules rearrange so
that O∗Oa = O∗Ob and therefore reach a symmetric configuration where both
hydrogen-bond acceptors are energetically equivalent. c) During the transition
state the original hydrogen bond O∗H∗···Oa bifurcates symmetrically. d-e)
The formation of the new hydrogen bond and the breaking of the old hydrogen
bond happen simultaneously which is accompanied by an angular jump of
O∗H∗. The initial overcoordination of Oa and undercoordination of Ob are
interchanged.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. .

.

Figure 1.3: Dangling OH (dOH) at the water/air interface.
. . . . . . . . . . . . . . . . .
Figure 1.4: Water spectrum from C.S. Tian et al. showing the broad bOH feature (3000−
3600cm−1) and narrow dOH feature at the interface with OTS (3680cm−1)
and with air (3695cm−1).[2]

. . . . . . . . . . . . . . . . . . . . . . . . . . .

.

Figure 2.1: Sum-frequency generation at the interface between two media with respective

refractive indices ni and nt.

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

2

3

5

9

Figure 2.2:

Illustration of the direction of beam polarizations for the polarization combi-
nation ppp. .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

.

.

.

.

.

.

Figure 2.3: Comparison of the Fresnel coefficients Lzz and Lxx at the water-OTS-FS interface. 13

Figure 2.4: Fresnel transmission coefficients Lzz for p-polarized VIS at 800nm (solid
plots) and p-polarized IR at 3680cm−1 (dashed plots). The blue plots describe
the coefficient for the water-air interface. The red plots describe the coefficient
for the water-FS interface. The orange plots describe the coefficients for the
water-OTS-FS interface. Note that |Lzz| for SFG would be essentially the same
as Lzz for VIS. This is due the fact that Lzz depends on indices of refraction
which are only weakly wavelength-dependent between λSFG = 618nm and
λV IS = 800nm for water and OTS-FS.

. . . . . . . . . . . . . . . . . . . . . . 17

vii

Figure 2.5: Sum-frequency generation at the interface of a monolayer and the medium below. 18

Figure 2.6: Schematic of time-resolved SFG spectroscopy using an additional IR pump
beam. The probe SFG polarization is ppp. The pump polarization can be
changed from p to s with a waveplate.

. . . . . . . . . . . . . . . . . . . . . . 19

Figure 2.7: Schematic probe SFG spectra for pump-off (Ioff(λ ;tk)) and pump-on (Ion(λ ;tk))

at a pump-probe delay of tk. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

Figure 2.8: A water molecule at the interface excited by an IR laser beam resonant with
the dangling OH stretch mode. The z-axis is normal to the interface; x- and
y-axes are in the surface plane.

. . . . . . . . . . . . . . . . . . . . . . . . . . 21

Figure 2.9:

Illustration of the azimuthal distribution of the excited dOH groups assuming
a dOH tilt angle θ0 = 59◦ [3]. Green indicates the probability distribution of
excited dOH produced by an s-polarized pump; red indicates the probability
distribution created by a p-polarized pump. . . . . . . . . . . . . . . . . . . . . 21

Figure 2.10: Time evolution of the excited-state angular distribution shown in figure
2.9 created by the p-polarized pump pulse (left) and the s-polarized pump
pulse (right). To highlight how in-plane reorientational effects contribute
to anisotropy decay, population decay is turned off: The pump-induced az-
imuthal anisotropy decays and azimuthal isotropy is re-established. Color
key: Red t = 0, purple t = 0.25T , orange t = T , where T is the decay time
associated with in-plane reorientation.

. . . . . . . . . . . . . . . . . . . . . . 22

Figure 2.11: Time evolution of the excited-state angular distribution shown in figure 2.9
created by the p-polarized pump pulse (left) and the s-polarized pump pulse
(right). To highlight how population decay affects the excited-state angular
distribution, in-plane reorientation is turned off: The pump-induced anisotropy
persists in this case. Color key: Red t = 0, purple t = 0.25T , orange t = T ,
where T is the governing population decay time. . . . . . . . . . . . . . . . . . 22

Figure 2.12: The average and difference of p- and s-pump are calculated for the case that
population decay is turned off. The results reveal that probing the distribution
with a p-polarized probe (shown in black) leads to a large sensitivity to
azimuthal anisotropy in the difference of p- and s-pump. The average of the
two quantities shows little sensitivity. Color key: Red t = 0, purple t = 0.25T ,
orange t = T , where T is the decay time associated with in-plane reorientation. . 24

Figure 2.13: The average and difference of p- and s-pump are calculated for the case that no
reorientation is allowed. The results reveal that probing the distribution with a
p-polarized probe (shown in black) leads to a large sensitivity to population
decay in both quantities. Color key: Red t = 0, purple t = 0.25T , orange
t = T , where T is the decay time associated with in-plane reorientation. . . . . . 24

viii

Figure 2.14: Schematic illustration of the pump induced bleach for p-polarized pump

(cid:0)βp(t)(cid:1), s-polarized pump (βs(t)), and scaled s-polarized pump.

. . . . . . . . 26

Figure 2.15: Schematic plot of the difference and average of βp and βs for time constants

observed in the experiment.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

Figure 2.16: Schematic overview of the dangling OH stretch mode decay mechanisms.

. .

. 28

Figure 2.17: Overview from Xiao et al.[4] of the processes involved in the decay of the

excited dangling OH stretch vibration.

. . . . . . . . . . . . . . . . . . . . . . 34

Figure 3.1: Schematic illustration of the Spectra Physics Spitfire PRO-XP system.

. . . .

. 36

Figure 3.2: Schematic overview of the experimental setup used to investigate the dangling OH. 37

Figure 3.3: Schematic overview of the homemade OPA. The red beam represents the
800nm, the green beam is the supercontinuum white light, the orange beam
is the signal and idler after the first pass through the BBO crystal, the brown
beam is its reflection from the spherical mirror and the cyan beam is the signal
and idler after the second pass through the BBO crystal.

. . . . . . . . . . . . . 38

Figure 3.4: Experimental schematic of the beam geometry and sample. The fused FS
prism is coated with an OTS SAM. The dotted blue line is the pump SFG
(pump+VIS) which is discarded using an iris during the experiment but is of
importance for alignment purposes. . . . . . . . . . . . . . . . . . . . . . . . . 39

Figure 3.5:

Illustration of the spectral narrowing by the pulse shaper.

. . . . . . . . . . .

. 40

Figure 3.6: Schematic overview of the 4f pulse shaper. To illustrate the pulse shaper’s
operation, the figure uses three distinct colors (red, green, blue) which would
correspond to the red tail, the center and blue tail of the 800nm spectrum.

. .

. 41

Figure 3.7: Schematic of the galvanometer-driven separation of “pump-on” and “pump-
off” beams where α(t) is the galvanometer mirror angle at time t. The gray
(red) laser pulses indicate the probe SFG signals corresponding to by the
chopper blocked (unblocked) IR pump pulses.

. . . . . . . . . . . . . . . . . . 42

Figure 3.8: Schematic illustration of a typical normalized intensity trace Inorm(tk).

. . . . . 44

Figure 3.9: Schematic illustration of a typical trace of the pump-induced bleach.

. . . . . . 45

Figure 3.10: Schematic of the data collection and cosmic ray removal process. Each box
represents a spectrum collected during scan number l at pump-probe delay tk.

. 46

ix

Figure 3.11: Chemical structure and space-filling-model of octadecyltrichlorosilane, CH3
(CH2)17 SiCl3. Colour code: Carbon, C: grey-black; Hydrogen, H: white;
Chlorine, Cl: green; Silicon, Si: turqouise-grey. [5]

. . . . . . . . . . . . . . . 48

Figure 3.12: OTS coated prism. A circular fit yields a contact-angle of θC = 100.4± 1.4◦.

An elliptic fit yields a contact-angle of θE = 104.8± 0.94◦.

. . . . . . . . . . . 49

Figure 3.13: Setup by Lamour et al.[6] for the contact-angle measurement optimized for

analysis with “ImageJ”.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

Figure 3.14: Apparatus equipped with a Whatman Carbon Cap 75 to filter salt solutions. . .
Figure 4.1: Red: IR spectrum with a ≈ 45cm−1 full-width half-maximum determined
from the probe SFG spectrum from a Ag mirror. Black: SFG spectrum of the
dOH stretch at the fused-silica/ODS/H2O interface.

. . . . . . . . . . . . . . . 53

. 51

Figure 4.2: Time-resolved sum-frequency signal of the dOH stretch at the ODS/water
interface normalized to the SFG signal at t < −1.2ps for pure H2O and
H2O:D2O. The blue shading highlights the difference between p- and scaled s-
pump data. The black curve is the third-order cross-correlation at ω = ωVIS +
ωpump + ωprobe recorded at the same configuration as the measurements to
indicate t = 0 and the temporal resolution of our instrument. The bright green
lines show biexponential fits to the data for t ≥ 0.6ps.

. . . . . . . . . . . . . . 55

Figure 4.3: dOH stretch SFG dynamics at the ODS/H2O and ODS/H2O:D2O interfaces
showing the decay of the average bleach ¯β(cid:48)(t) (open blue circles) relative to
the long-decay bleach and the difference ∆β (t) (solid red circles) between
the p-pump and scaled s-pump data. Solid curves show biexponential fits
(cyan) for t ≥ 2.0ps and single-exponential fits (orange) for t ≥ 0.6ps. The
third-order cross correlation is shown in solid black.

. . . . . . . . . . . . . . . 59

Figure 4.4: MD simulations by Stirnemann and Laage: a) MD snapshot of the water/air
interface showing the instantaneous interface as a red mesh. b) Definition of
the OH group angle θ and the distance to the instantaneous interface z. c) Pop-
ulation decay of dOH→bOH for water/air and water/hydrophobic interfaces.
d) Free energy difference ∆FdOH of the water/air and water/hydrophobic
interface.[4]

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

. .

. .

Figure 5.1: Solvated anions, like Cl−, form hydrogen bonds with surrounding water
molecules (OH···Cl−, orange). These hydrogen bonds can have a different
absorption spectrum than bOH in bulk water (OH···O, blue).

. . . . . . . . . . 70

x

Figure 5.2: Solvated cations, like Na+, do not form hydrogen bonds with the surrounding
water molecules. Due to the positive charge of the cation, the water molecules
orient themselves so that the partially positive hydrogen atoms face away
from the cation. The hydrogen bonds formed by water molecules in the first
solvation shell (OH···O, blue) already exhibit the same frequencies as bOH
in bulk. .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

.

.

.

.

.

.

.

.

Figure 5.3: Tielrooij et al.[7] show that in a semi-rigid hydration shell water molecules are
only locked in one direction (A,B) and the cooperative interaction of cation
and anion that locks water molecules in two directions (C) while involving
water molecules beyond the first solvation shell.

. . . . . . . . . . . . . . . . . 71

Figure 5.4: Schematic representation by Laage et al.[8] of the molecular jump mechanism
in a solvation shell around a chloride atom showing the “departure” of a water
molecule from the solvation shell. (Blue sticks: H-bonds involving molecules
in the surrounding solution. Green sticks: H-bonds directly involved in
reorientation.) The sequence starts with overcoordination of the chloride ion
and an undercoordinated water molecule in the second solvation shell. This is
followed by a bifurcation and abrupt switch of the hydrogen bond, followed
by the departure of the water molecules from the first solvation shell.

. . . . . . 73

Figure 5.5: MD simulations from Jungwirth and Tobias show number densities of water,
oxygen atoms, and ions, normalized by the bulk water density ρb, for several
1.2M salt solutions.[9]

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

Figure 5.6: Schematic illustration of dangling OH bonds at the water/air interface showing
dOH groups within the first (illustrated by the dashed circle) and second
solvation shell of a iodine ion. . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

Figure 5.7: NaCl normalized SFG intensities . . . . . . . . . . . . . . . . . . . . . . . .

. 80

Figure 5.8: NaCl difference and average of bleach . . . . . . . . . . . . . . . . . . . . . . 81

Figure 5.9: NaI normalized SFG intensities . . . . . . . . . . . . . . . . . . . . . . . . . . 82

Figure 5.10: NaI difference and average of bleach . . . . . . . . . . . . . . . . . . . . . . . 83

Figure 5.11: MgSO4 normalized SFG intensities . . . . . . . . . . . . . . . . . . . . . . . . 84

Figure 5.12: MgSO4 difference and average of bleach . . . . . . . . . . . . . . . . . . . .

. 85

xi

CHAPTER 1

INTRODUCTION

The properties of water have fascinated humans for millennia, and many of its properties are well

known from introductory physics and chemistry lessons. Compared to fluids of similar molecular

size it provides us with a variety of interesting properties, including its phase transitions and density

anomaly. For example, molecules made of hydrogen and elements surrounding oxygen in the

periodic table are all gases at standard conditions (N, F, P, S, Cl). In contrast, water is of course

liquid under the same conditions. Many of these unusual properties are due to the fact that water

forms strong hydrogen bonds with four neighboring water molecules. These bonds, schematically

displayed in figure 1.1, form due to the much larger electronegativity of oxygen compared to

hydrogen.

Figure 1.1: Polarization of a water molecule and the resulting hydrogen bonds with neighboring
water molecules. The oxygen of the central water molecule is considered to be an acceptor of two
hydrogen bonds (with the water molecules in the lower left corner of the figure) and a donor of two
hydrogen bonds (with the water molecules in the top and the lower right corner of the figure).

1

1.1 Structure and orientational dynamics in bulk water

In bulk, a water molecule generally forms four hydrogen bonds with its neighboring water

molecules. Focusing on the water molecule in the center of figure 1.1, the oxygen supplies two lone

pairs of electrons that can each form a hydrogen bond with a hydrogen atom of an adjacent water

molecule. Conversely, each hydrogen atom of the central water molecule can form a hydrogen

bond with lone electron pairs of an oxygen atom of a neighboring water molecule. Throughout

this dissertation this type of bond is called a bonded oxygen-hydrogen group or bOH. At room

temperature, hydrogen bonds are very dynamic: They break and form rapidly and enable the bOH

to reorient. The dynamics of bulk water has been studied in theory and experiment. Molecular

dynamics (MD) simulations show that the reorientation of bOHs occurs with a large angular jump
motion covering 68◦ on a 3.3ps time scale.[10] Further MD studies suggest that hydrogen atoms can
participate in two hydrogen bonds simultaneously − they bifurcate − as an “old” hydrogen bond
breaks and a “new” one forms.[11, 1] Experimentally, the orientational dynamics in bulk water can

be accessed via infrared (IR) pump-probe spectroscopy by separately measuring the pump-induced
change in absorption with pump and probe polarizations being parallel (∆α(cid:107)) and then orthogonal

Figure 1.2: Schematic reorientation in bulk illustrated by Laage and Hynes[1]. a-b) Initially
fluctuations lead to an overcoordinated water molecule (Oa) and an undercoordinated water
molecule (Ob). The water molecules rearrange so that O∗Oa = O∗Ob and therefore reach a
symmetric configuration where both hydrogen-bond acceptors are energetically equivalent. c)
During the transition state the original hydrogen bond O∗H∗···Oa bifurcates symmetrically. d-e)
The formation of the new hydrogen bond and the breaking of the old hydrogen bond happen
simultaneously which is accompanied by an angular jump of O∗H∗. The initial overcoordination of
Oa and undercoordination of Ob are interchanged.

2

(∆α⊥) with respect to each other. The rotational anisotropy can then be calculated[12, 13] to be

=

2
5

R(t) ≡ ∆α(cid:107)(t)− ∆α⊥(t)
∆α(cid:107)(t) + 2∆α⊥(t)

(1.1)
with the second-order Legendre polynomial orientational correlation function C2(t) =(cid:104)P2[ ˆµ(t)· ˆµ(0)](cid:105)
and the transition dipole ˆµ. Since bulk water is an isotropic medium, the population dynamics

C2(t),

can be easily disentangled from the orientational dynamics by performing pump-probe spec-

troscopy in a magic-angle geometry: Adjusting the angle between pump and probe polarizations to
≈ 54.7◦ has the effect that all interactions governed by the second order Legendre
θm = arccos 1√
3
polynomial (P2(cos54.7◦) = 0), including orientational dynamics, cancel out and one measures the
population dynamics directly. This is equivalent to measuring ∆α(cid:107) + 2∆α⊥.

1.2 Structure and orientational dynamics at the water interface

At the water interface several issues arise that make its study more challenging. In order to

study the interface, it is necessary to separate the surface from the bulk response. This rules out

the possibility of obtaining the population dynamics of the system at the interface through a non-

surface-specific magic angle measurement. Particularly, on the experimental side new approaches

were needed.

At the interface with air or other hydrophobic matter the hydrogen bond network terminates and

Figure 1.3: Dangling OH (dOH) at the water/air interface.

3

dOHleaves behind dangling oxygen-hydrogen groups (dOH) on at least 20% of the water molecules

at the surface.[14] A dOH is an OH group that by definition does not form hydrogen bonds with

other water molecules. They stick out of the bulk, with the vector from the oxygen to the hydrogen

atom pointing towards the air or the hydrophobic matter (figure 1.3). dOH groups can also briefly

exist in bulk water for < 200fs before they participate in a hydrogen bond with another water

molecule again.[15] They can be neglected in our experiment due to their short lifetime and their on

average random orientation. Looking at the vibrational OH stretch spectrum of water at an interface

(figure 1.4), one can see a broad and a narrow feature. The broad feature is associated with bOH,

and its width can be explained by the various strengths of hydrogen bonds with which a bOH can

be involved. In general, the weaker the hydrogen bond is the higher the OH stretch vibrational

frequency. Of course, this means that we will find the dOH feature, which is associated with OH

groups that are not involved in any hydrogen bonds, at a higher, narrower range of frequencies

compared to the bOH. In figure 1.4 we can also see that the spectrum of the water/air interface

differs from the spectrum of the water/OTS interface. Although the interfaces with air and OTS are

both considered to be hydrophobic interfaces, it is interesting to see such a significant difference

in their spectra. This raises the question of whether the rotational dynamics of the dOH at those

interfaces differ, too.

In deed, such a difference has been found theoretically: Stirnemann et al. [10] investigated the

reorientation dynamics at a model hydrophobic surface with MD simulations. Their result puts
the dOH reorientation time at 1.6ps, which is in contrast to the 0.8− 1.1ps reorientation time at
the water/air interface obtained computationally by Vila Verde et al. [16] and experimentally by

Hsieh et al. [17]. Even though the difference is significant, it is hard to distinguish the difference

experimentally because of the challenge of obtaining a sufficiently high signal-to-noise ratio. This

dissertation elaborates on how this problem can be solved.

4

Figure 1.4: Water spectrum from C.S. Tian et al. showing the broad bOH feature
(3000− 3600cm−1) and narrow dOH feature at the interface with OTS (3680cm−1) and with air
(3695cm−1).[2]

5

CHAPTER 2

THEORY

In order to measure the interfacial orientational dynamics of water, one needs a surface-specific

probe. This can be understood easily because only after about 5Å into the bulk water molecules

adopt bulk properties.[18, 19] Sum-frequency generation, a second order process, is an established

technique for this task. Its surface specificity is rooted in the fact that for systems with inversion

symmetry, for example bulk water, the second order susceptibility vanishes in the electric-dipole

approximation. Only at the surface, where inversion symmetry is broken, is the electric dipolar

contribution to the second order susceptibility non-zero. The following section covers the theoretical

aspects of sum-frequency generation (SFG) spectroscopy which allow the collection of surface-

specific spectra. The connection is made between the orientational dynamics of a molecules and

the orientational dynamics of the orientational distribution of vibrational excitations. In addition,

I discuss how measurements in a total-internal reflection geometry can lead to an improved SFG

signal compared to measurements at the interface between water and air. Moreover, this chapter

introduces the concept of time-resolved pump-probe SFG that enables the measurement of dynamics

at the interface and the challenge of separating population relaxation from orientational dynamics.

2.1 Light-matter interaction

Let’s start with the macroscopic description of the light-matter interaction. In a dielectric

medium an electric field E induces electric dipoles, whose density is described by the polarization

density P, also just called polarization. For homogeneous and isotropic dielectrics the relationship

between E and P is of linear and scalar nature:

P = ε0χE

(2.1)

6

The scalars ε0 and χ are the electric constant and the electric susceptibility, respectively. The
susceptibility is related to the relative permittivity by the following relationship:

εr = 1 + χ

(2.2)

χ and εr are dimensionless and in general complex quantities. εr also determines the complex

refractive index of a material:

√
εrµr,

n =

(2.3)

where µr is the relative permeability, which for many materials is ≈ 1. The real and imaginary parts
of n = n + iκ are the refractive index n and the extinction coefficient κ, respectively: The latter is
related to the attenuation coefficient α = 4πκ/λ0 with the vacuum wavelength λ0, known from the
Lambert’s law describing the transmittance of a material sample of a given thickness l:

T (l) =

I(l)
I(0)

= exp(−αl),

(2.4)

where I(l) is the light intensity after a sample thickness of l. In the case of anisotropic dielectrics,

for which the polarization and the electric field are not in the same direction, the susceptibility in

equation 2.1 becomes a tensor:

Pi = ε0∑
j

χi jE j

(2.5)

2.1.1 Non-linear polarization

In the presence of a sufficiently strong electric field, deviations from a linear response are seen.

Assuming sufficiently weak fields and no permanent dipole moments, a Taylor series in E can

describe the polarization:

Pi = P(1)
= ε0∑
j

i + P(2)
i + P(3)
χ(1)
i j E j + ε0∑
jk

i + . . .
χ(2)
i jk E jEk + ε0∑
jkl

χ(3)
i jklE jEkEl + . . .

(2.6)

(2.7)

7

We are especially interested in the second term, which is the source of polarization for sum-frequency

generation spectroscopy and is described in the following section in more detail:

P(2)
i = ε0∑
jk

χ(2)
i jk E jEk

(2.8)

From here on ε0 is omitted to improve the readability, unless otherwise noted.

2.2 Sum-frequency generation spectroscopy

Sum-frequency generation (SFG) spectroscopy is a non-linear process where two laser beams
with frequencies ω1 and ω2 overlap at a surface of a material or the interface between two materials
and create an output beam with frequency ω3 = ω1 + ω2. More specifically, the two input photons
at ω1 and ω2 are annihilated, while one output photon at ω3 is created. For this second order process
to be observed sufficiently intense laser beams have to interact with non-centrosymmetric matter,

for example, surfaces or interfaces.

2.2.1 Surface specificity of SFG

For any system with inversion symmetry, for example bulk water, the second order polarization

vanishes. This can be understood by the following proof. Consider the inversion operator Iinv
that inverts any arbitrary vector V: IinvV = −V. Applying Iinv to the second order polarization
then results in IinvP(2) = −P(2). Using the definition in equation 2.12 and the fact that for a
centrosymmetric medium the electric susceptibility satisfies Iinv

(cid:110)←→χ (2) ˜EVIS ˜EIR

←→χ (2) = ←→χ (2) yields
(cid:111)

IinvP(2) = Iinv
= P(2)

(2.9)

(2.10)

From −P(2) = P(2) one can deduce that P(2) = 0. The consequence is that the second order
susceptibility has to vanish: ←→χ (2) = 0. By extension, all polarization terms of even order vanish in
the electric-dipole approximation for systems with inversion symmetry. At interfaces this symmetry
is broken, allowing a non-zero ←→χ (2) and the generation of a second order signal.

8

Figure 2.1: Sum-frequency generation at the interface between two media with respective refractive
indices ni and nt.

2.2.2 Polarization and susceptibility

To further examine the properties of SFG, let us consider two electromagnetic plane waves ˜En(r,t) =
En(r,t)exp (−iωnt) +c.c., where En(r,t) = An(t)exp (ikn · r), with different frequencies ωVIS and
ωIR to be incident onto a nonlinear medium with the second order susceptibility, a third rank tensor,
←→χ (2) with the components χ(2)

i jk . The total electric field is

˜E(r,t) = ˜EVIS(r,t) + ˜EIR(r,t) = EVIS(r,t)exp (−iωVISt) + EIR(r,t)exp (−iωIRt) + c.c.

The resulting second order polarization is then of the form (and dropping r and t)

(cid:105)
(cid:105)
VIS exp (−2iωVISt) + c.c.
E2
IR exp (−2iωIRt) + c.c.
E2

˜P(2)(r,t) =←→χ (2) ˜E(r,t)2
=←→χ (2)(cid:104)
+←→χ (2)(cid:104)
+ 2←→χ (2) [EVISEIR exp (−i(ωVIS + ωIR)t) + c.c.]
+ 2←→χ (2) [EVISE∗
+ 2←→χ (2)(cid:2)EVISE∗
IR exp (−i(ωVIS − ωIR)t) + c.c.]
VIS + EIRE∗

(cid:3)

IR

(SHG)

(SHG)

(SFG)

(DFG)

(OR)

(2.11)

(2.12)

(2.13)

(2.14)

(2.15)

(2.16)

(2.17)

The terms are grouped to illustrate the different interactions that contribute to the second order

polarization. They are second harmonic generation (SHG), sum frequency generation (SFG),

9

θSFGVISIRθIRθVISSFGnintzyxdifference frequency generation (DFG), and optical rectification (OR). All terms conserve en-

ergy and momentum. Looking specifically at the SFG term of the polarization, this means that
¯hωSFG = ¯hωVIS + ¯hωIR and ¯hkSFG = ¯hkVIS + ¯hkIR. Momentum and energy conservation can
be used to separate the SFG signal from the other signals. This can be accomplished spectrally

or for noncollinear input beams, spatially because each signal propagates in a different direction
according to its momentum. Remembering that En(r,t) = An(t)exp(ikn · r) one can rewrite the
SFG polarization term

PSFG = 2←→χ (2) [EVISEIR exp (−i(ωVIS + ωIR)t) + c.c.]
(cid:2)E j,VISEk,IR exp (−i(ωVIS + ωIR)t) + c.c.(cid:3)
PSFG
(cid:2)A j,VISAk,IR exp (−iωSFGt + ikSFG · r) + c.c.(cid:3)
i

χ(2)
i jk

χ(2)
i jk

= 2∑
jk
= 2∑
jk

(2.18)

(2.19)

(2.20)

where the indices i, j, k are the Cartesian components of the fields.

2.2.3 Connection between macroscopic and microscopic description

The second order susceptibility in equation 2.19 can be described with a resonant and a non-resonant
term labeled χ(2)

NR, respectively:[20]

R and χ(2)

χ(2)
i jk = χ(2)

NR,i jk + χ(2)

R,i jk = χ(2)

NR,i jk +

α(2)
i jk (ωIR)

(2.21)

(cid:68)

(cid:69)

Ω

The resonant part of the macroscopic susceptibility χ(2)
ensemble average (cid:104). . .(cid:105)Ω of the molecular hyperpolarizability tensor[21]

R is related to the microscopic level by the

(ˆi·←→α (Ω)· ˆj)(µ(Ω)· ˆk)

α(2)
i jk (ωIR) = ∑
q

(2.22)
with the molecular orientation Ω(t) = (θ (t),φ (t)), the Raman polarizability tensor ←→α , the dipole
derivative µ, the frequency of the IR beam ωIR, the resonant frequency ωq and damping constant
Γq of the q-th vibrational mode. Within this dissertation the IR laser is resonant with the vibrational

ωIR − ωq + iΓq

stretch mode of the dangling OH. Therefore, by dropping the sum over all vibrational modes, the

10

(cid:68)
(cid:69)
(ˆi·←→α (Ωt)· ˆj)(µ(Ωt,ωq)· ˆk)

ωIR − ωdOH + iΓdOH

Ω

(2.23)

resonant term becomes:

χ(2)
R,i jk =

2.2.4 Measured SFG intensity

A quantitative treatment of the nonlinear response requires that we take into account that the media

that form the interface have different indices of refraction. In that case the intensity of the sum

frequency signal created by the polarization is
ω2
SFG

ISFG =

8ε0c3ni(ωSFG)ni(ωVIS)ni(ωIR)cos2(θSFG)

(cid:12)(cid:12)(cid:12)χ(2)

eff

(cid:12)(cid:12)(cid:12)2

IVISIIR.

(2.24)

where χ(2)

eff is the effective surface nonlinear susceptibility:

eff = [LSFG · ˆeSFG]·←→χ (2) : [LVIS · ˆeVIS] [LIR · ˆeIR] .
χ(2)

(2.25)

ˆe j is the unit polarization vector of the optical field at ω j, and L(ω j) are the local field corrections
(the transmission Fresnel factors), which are covered in more detail during section 2.2.6. But first,
let us address how the polarization of the laser beams affects χ(2)
eff .

2.2.5 Laser beam polarization and non-zero tensor elements of χ(2)

A common way to describe the polarization of laser beams is if they are parallel (“p-polarized”)

or perpendicular (“s-polarized”) to the optical plane. Using the coordinate system in figure 2.1 to

describe the direction of the polarization, “s” polarization is parallel to the y-axis. Consider, for
example, an s-polarized (ˆeSFG = ˆey) SFG term in equation 2.25:

LSFG · ˆeSFG = Lyy,SFGˆey · ˆey = Lyy,SFG

(2.26)

In contrast, since a “p”-polarized beam has polarization components in the x-z plane (ˆeSFG =
−cos(θSFG)ˆex + sin(θSFG)ˆez) the SFG term in 2.25 would have two contributions to the Fresnel
transmission factor:

LSFG · ˆeSFG = −Lxx,SFG cos(θSFG) + Lzz,SFG sin(θSFG)

(2.27)

11

Figure 2.2: Illustration of the direction of beam polarizations for the polarization combination ppp.

The minus sign originates in the fact that the x-component of ˆeSFG is pointing in the negative
x-direction, as visualized in figure 2.2. In the experiment the beam polarization combinations

ssp (denoting the polarization of the SFG signal, the VIS and the IR beam, respectively) and ppp

(analogously: SFG, VIS and IR all polarized in p-direction) are used. The effective susceptibility

for the polarization combination ssp can be calculated using the previous examples and equation

2.25:

eff,ssp = Lyy,SFG · Lyy,VIS · Lxx,IR · cosθIR · χ(2)
χ(2)

yyx + Lyy,SFG · Lyy,VIS · Lzz,IR · sinθIR · χ(2)
yyz

(2.28)

However, the water interface has C∞,ν symmetry, which reduces the 27 elements of the second-order
susceptibility tensor to just four distinct non-zero elements:

zxx = χ(2)
χ(2)
zyy ,

xzx = χ(2)
χ(2)
yzy ,

xxz = χ(2)
χ(2)
yyz ,

χ(2)
zzz ,

(2.29)

where we have taken the z-axis to be parallel to the surface normal. Since χ(2)
susceptibility χ(2)

eff,ssp in equation 2.28 simplifies to only one term:

yyx = 0, the effective

eff,ssp = Lyy,SFG · Lyy,VIS · Lzz,IR · sinθIR · χ(2)
χ(2)

yyz

(2.30)

12

Figure 2.3: Comparison of the Fresnel coefficients Lzz and Lxx at the water-OTS-FS interface.

Analogously, for the polarization combination ppp, the effective susceptibility χ(2)
from 23 = 8 terms to four non-zero terms:

eff,ppp simplifies

eff,ppp =− Lxx,SFG · Lxx,VIS · Lzz,IR · cosθSFG · cosθVIS · sinθIR · χ(2)
χ(2)
xxz
− Lxx,SFG · Lzz,VIS · Lxx,IR · cosθSFG · sinθVIS · cosθIR · χ(2)
xzx
+ Lzz,SFG · Lxx,VIS · Lxx,IR · sinθSFG · cosθVIS · cosθIR · χ(2)
zxx
+ Lzz,SFG · Lzz,VIS · Lzz,IR · sinθSFG · sinθVIS · sinθIR · χ(2)
zzz

(2.31)

(2.32)

(2.33)

(2.34)

For large incident angles θVIS and θIR, the exit angle of the SFG beam, θSFG, becomes large
too. This enhances the χ(2)
zzz component compared to the other three contributions to the effective
susceptibility. In the next chapter it will become apparent that the Fresnel coefficient Lxx approaches

zero (figure 2.3) at the angle of total internal reflection and one can approximate:
eff,ppp ≈ Lzz,SFG · Lzz,VIS · Lzz,IR · sinθSFG · sinθVIS · sinθIR · χ(2)
χ(2)

zzz

(2.35)

The Fresnel coefficients Lxx, Lyy and Lzz are derived and their role for our experiment discussed in

more detail in section 2.2.6.

13

0204060800.00.51.01.52.02.53.03.5angleofincidenceθi[deg]ComparisonofFresnelcoefficientsLxxandLzzLxxIR3680cm-1LzzIR3680cm-1Lxx(VIS800nm)Lzz(VIS800nm)2.2.6 Fresnel equations and transmission coefficients

The Fresnel equations determine the transmission and reflection coefficients of an electromagnetic

wave at an interface between two non-magnetic media (B = H). Consider an incident, a reflected

and a transmitted electromagnetic wave with:

Ei = E0

i exp(i(ki · r− ωt))
r exp(i(kr · r− ωt))
t exp(i(kt · r− ωt))

Er = E0

Et = E0

Hi =

Hr =

Ht =

n1
ki × Ei
ki
n1
kr × Er
kr
n1
kt × Et
kt

(2.36)

(2.37)

(2.38)

The directions of the reflected and transmitted beams can be easily determined (θi = θr and Snell’s
law ni sinθi = nt sinθt). To determine the field amplitudes one needs to look at the boundary
conditions:

(Ei + Er)× n = Et × n
(Hi + Hr)× n = Ht × n,

(2.39)

(2.40)

where n is the unit normal vector of the interface pointing into medium 2. In experiments the fields

are usually either s or p polarized.

2.2.6.1

s-polarized light

In the case of s-polarized light the electric field is perpendicular to the optical plane. The first

boundary condition gives:

E0
i + E0

r = E0
t

(2.41)

14

Using the relationships between E and H, the second boundary condition can be rewritten with the

use of expansion of the triple product in the following form:

n× (ki × Ei) + n× (kr × Er) = n× (kt × Et)

ki(n· Ei)− Ei(n· ki) + kr(n· Er)− Er(n· kr) = kt(n· Et)− Et(n· kt)

Ei(n· ki) + Er(n· kr) = Et(n· kt)
Eiki cosθi + Erkr cosθr = Etkt cosθt

(2.42)

(2.43)

(2.44)

(2.45)

Using the fact that momentum parallel to the surface is conserved (ki sinθi = kr sinθr = kt sinθt),
which is more commonly known in the form of Snell’s law (ni sinθi = nt sinθt), and θr = θi:

(cid:16)

i − E0
E0

r

(cid:17)

=

ki = kr
nt
kt
ni
ki
nt
ni

cosθi =

E0
t cosθt

Combining the two boundary conditions (equ. 2.41, 2.48), one can solve for Er and Et:

E0
r =

E0
t =

sin(θt − θi)
sin(θt + θi)
2cosθi sinθt
sin(θt − θi)

E0
i = rSE0
i

i = tSE0
E0
i

(2.46)

(2.47)

(2.48)

(2.49)

(2.50)

The dependence on θt of the coefficient tS can be easily eliminated by using Snell’s law. One obtains
the following relationship which is typically denoted as Lyy in the literature:

Lyy =

2cosθi sinθt
sin(θt − θi)

.

(2.51)

2.2.6.2 p-polarized light

If the light is p-polarized, the E-field is parallel to the optical plane (x-z plane) and the H-field

points out of the optical plane. Essentially the roles of E and H have been interchanged. Knowing

this, one can use the result from the s-polarized case to write down the two boundary conditions

15

(H0

j = n jE0
j ):

Solving for Er and Et yields:

E0
r =

E0
t =

(cid:17)

= ntE0
t

(cid:16)
E0
i + E0
ni
r
i − E0
E0

(cid:17)

r

(cid:16)

cosθi = E0

t cosθt.

E0
i = rPE0
i

tan (θt − θi)
tan (θt + θi)
sin (θi + θt)cos (θi − θt)

2cosθi sinθt

i = tPE0
E0
i .

(2.52)

(2.53)

(2.54)

(2.55)

The Fresnel coefficients for the electric field projections in x and z directions, rxx, rzz, txx and tzz

can be easily obtained. Here we are only interested in the transmission coefficients:

t cosθt

Exx
t = E0
= tPE0

= tP

i cosθt
Exx
i
cosθi
= txxExx
i

.

cosθt

(2.56)

(2.57)

(2.58)

(2.59)

where txx = tP
SFG spectroscopy, txx and tzz are denoted as Lxx and Lzz, respectively.

. Analogously, one can derive tzz = tP

cosθt
cosθi

sinθt
sinθi

. When using the typical notation for

Lxx =

Lzz =

2cosθi sinθt

sin (θi + θt)cos (θi − θt)
sin (θi + θt)cos (θi − θt)

2cosθi sinθt

cosθt
cosθi
sinθt
sinθi

(2.60)

(2.61)

2.2.6.3 Fresnel coefficients at the angle of total internal reflection

In our experiment we have chosen the water-fused silica (FS) interface in the form of a prism over

the water-air interface because significant signal enhancements can be made by operating with a

beam geometry that is close to the angle of total internal reflection (TIR). TIR is not possible at the

water-air interface because ni < nt. At the water-FS interface ni > nt, so signal enhancement through

16

Figure 2.4: Fresnel transmission coefficients Lzz for p-polarized VIS at 800nm (solid plots) and
p-polarized IR at 3680cm−1 (dashed plots). The blue plots describe the coefficient for the water-air
interface. The red plots describe the coefficient for the water-FS interface. The orange plots
describe the coefficients for the water-OTS-FS interface. Note that |Lzz| for SFG would be
essentially the same as Lzz for VIS. This is due the fact that Lzz depends on indices of refraction
which are only weakly wavelength-dependent between λSFG = 618nm and λV IS = 800nm for
water and OTS-FS.

TIR is possible. For example, at 800nm the refractive indices are ni = 1.4533 and nt = 1.3290.
Figure 2.4 shows the angle-dependence of the Fresnel coefficients and gives an estimate of how

much the signal is boosted compared to the water-air interface. Nevertheless, FS is considered,

as opposed to air, a hydrophilic material that forms hydrogen bonds with water. Since we are

interested in investigating the dangling OH it is necessary to create a hydrophobic interface. This

is accomplished by coating the FS prism with an octadecyltrichlorosilane (OTS) self-assembled

monolayer (SAM).

For our purposes, another advantage of this configuration is the absolute magnitude of the Fresnel

transmission coefficients Lzz and Lxx. In figure 2.3 it can be seen that at the angle of TIR, where Lzz
has its maximum, Lxx approaches zero. We can therefore neglect all terms of χ(2)
eff,ppp that contain

17

Figure 2.5: Sum-frequency generation at the interface of a monolayer and the medium below.

Lxx and arrive at equation 2.35. This means if we use the polarization combination ppp, for SFG,

VIS and IR, respectively, the only significant contribution to the sum-frequency signal is going to
be from χ(2)
zzz :

χ(2)
eff,ppp ∝ χ(2)
zzz

(2.62)

2.2.6.4 Fresnel coefficients for experiments involving a monolayer

Now let’s look closer at the case where the FS prism is coated with a monolayer, separating the FS

and water shown in figure 2.5. In this case, the Fresnel coefficients need to be modified slightly

by introducing a factor for the coefficient describing the relative field amplitude in the z-direction

(perpendicular to the monolayer)[22], resulting in:

LSAM
zz = Lzz

n2
t
n2
SAM

where nSAM can be approximated to be

n2
SAM =

n2
t (n2
4n2

t + 5)
t + 2

(2.63)

(2.64)

The angle dependence of |Lzz| for different water interfaces (water/air, water/FS, water/SAM/FS) is
shown in figure 2.4 for VIS and IR light.

18

θSFGVISIRθIRθVISSFGnintnSAMzyxFigure 2.6: Schematic of time-resolved SFG spectroscopy using an additional IR pump beam. The
probe SFG polarization is ppp. The pump polarization can be changed from p to s with a waveplate.

Figure 2.7: Schematic probe SFG spectra for pump-off (Ioff(λ ;tk)) and pump-on (Ion(λ ;tk)) at a
pump-probe delay of tk.

19

VISSFGdelaypumpprobeν=1ν=0VISprobeSFGzyxpump2.3 Time- and polarization-resolved SFG spectroscopy

In order to research the dynamics of the water interface, we use time-resolved SFG spectroscopy

by employing a pump beam with polarization and delay with respect to the probe beam that can be

varied through a delay stage and a waveplate (figure 2.6). The pump to excites the water molecules

from the ground state ν = 0 to the first excited state ν = 1 of the dangling vibrational stretch mode

which leads to a decrease or bleach in probe SFG signal. The anharmonicity of the dOH stretch
vibration is such that the excited state transition ν = 1 → 2 is redshifted by 90 cm−1 relative to the
ν = 0 → 1 transition.[23] With our IR having a bandwidth of 45cm−1 this suggests that excited
state absorption is not significant. Overall, this means that with the pump laser on we observe a

reduction of probe SFG signal intensity simply proportional to the square of the difference in excited
and ground state populations (N0 − N1)2 as illustrated in figure 2.7. However, this reduction in
probe SFG signal depends on the pump-probe delay: For τ much shorter than the overall population

relaxation time, the signal reduction is around 15% in our experiment when using collinear polarized

pump and probe. At very long delays between pump and probe the reduction almost completely

recovers. We do not observe a complete recovery on the time scale used in the experiment due to

transient heating described by a population decay of ν = 1 into an intermediate state where the

stretch modes have returned to their ground state, but the hydrogen bonds have weakened leading to

a shift in the bOH spectrum.[24] Until now, we have disregarded the fact that water molecules can

reorient, in- and out-of the optical plane.

2.3.1 Orientational dynamics

The reorientation of water molecules at an interface is modeled by rotation around the two axes

that the two OH groups form. Using the coordinate system in figure 2.8, where the z-axis is normal

to the interface, these two rotations are called "in-plane" and "out-of-plane" reorientation. In the

former case, the azimuthal angle φ changes while the angle θ persists and therefore the OH group

that is pointing up remains in the optical plane. In the latter case, the azimuthal angle stays constant

20

Figure 2.8: A water molecule at the interface
excited by an IR laser beam resonant with the
dangling OH stretch mode. The z-axis is
normal to the interface; x- and y-axes are in the
surface plane.

Figure 2.9: Illustration of the azimuthal
distribution of the excited dOH groups
assuming a dOH tilt angle θ0 = 59◦ [3]. Green
indicates the probability distribution of excited
dOH produced by an s-polarized pump; red
indicates the probability distribution created by
a p-polarized pump.

while the polar angle θ changes because the OH group that is pointing upwards rotates out of the

optical plane.

2.3.1.1

In-plane orientational dynamics

Reorientation in the optical plane can be thought of in terms of a wobbling motion of the dOH

around the z-axis, along the cone shown in figure 2.8. This motion leads to the restoration of the

azimuthal isotropy of excited dOH groups (illustrated in figure 2.10). It is of importance to note that

in-plane reorientation does not contribute to population decay. To illustrate the effect of population

decay, figure 2.11 represents a model scenario in which population relaxation is permitted but that

does not allow for in-plane reorientation. In this case, the projection of the angular distribution of

the vibrational excited-state distribution onto the z-axis diminishes with the declining population. In

reality, the effects of population decay and in-plane orientational dynamics occur at the same time.

21

θφxzyIR laserp polarizeds polarizedRe-establishment of azimuthal isotropy (population decay turned off)

p-pump

s-pump

Figure 2.10: Time evolution of the excited-state angular distribution shown in figure 2.9 created by
the p-polarized pump pulse (left) and the s-polarized pump pulse (right). To highlight how in-plane
reorientational effects contribute to anisotropy decay, population decay is turned off: The
pump-induced azimuthal anisotropy decays and azimuthal isotropy is re-established. Color key:
Red t = 0, purple t = 0.25T , orange t = T , where T is the decay time associated with in-plane
reorientation.

Persistent azimuthal anisotropy (in-plane reorientation turned off)

p-pump

s-pump

Figure 2.11: Time evolution of the excited-state angular distribution shown in figure 2.9 created by
the p-polarized pump pulse (left) and the s-polarized pump pulse (right). To highlight how
population decay affects the excited-state angular distribution, in-plane reorientation is turned off:
The pump-induced anisotropy persists in this case. Color key: Red t = 0, purple t = 0.25T , orange
t = T , where T is the governing population decay time.

22

1.00.50.51.01.00.50.51.01.00.50.51.01.00.50.51.01.00.50.51.01.00.50.51.01.00.50.51.01.00.50.51.02.3.1.2 Out-of-plane orientational dynamics

Out-of-plane reorientation describes the movement where the dOH flips to an orientation that

is tangential to the interface and now participates in a hydrogen bond (forward jump), and the

corresponding backwards motion (backward jump). Assuming the pump is p-polarized, an excited

dOH population with an azimuthal distribution as seen in red in figure 2.9 would be created. The

projection of the z-component of the orientational distribution of the population difference of

excited- and ground-state dOH groups onto the z-axis can be measured by the probe after a time τ.

Within our experiment we follow the development of this projection as a function of the pump-probe

delay.

On the other hand, if the pump is s-polarized, the excited dOH population starts out with an

azimuthal distribution as shown in green in figure 2.9. The same processes as in the case for

p-polarized pump take place, with the difference that the initial population of excited dOH groups

that can be probed by the p-polarized probe is smaller. Only the excited dOH groups that are

associated with the area where the pumped and probed azimuthal distributions in figure 2.9 overlap
contribute to the change in χ(2)
zzz .
A hand-waving argument can be made here that the average bleach in the p-pump and s-pump case

would essentially only be affected by population decay since the orientational dynamics average out

(the reorientation of dOH excited by a p-polarized pump out of the angular distribution seen by the

probe is compensated by the reorientation of dOH excited by an s-polarized pump into the angular

distribution seen by the probe), whereas the difference of the p-pump and s-pump case is sensitive

to orientational dynamics and population decay. These arguments are also supported by calculating

the average and difference of the angular distributions and follow their time dependence. (Shown in

figures 2.12 and 2.13) It turns out that there are many more processes at work besides reorientation

and population relaxation of the dOH groups that dissipate the energy of the stretch vibration and so

the validity of the hand-waving argument might not hold. These processes are discussed in section

2.4.

23

average of p-pump and s-pump

difference of p-pump and s-pump

Figure 2.12: The average and difference of p- and s-pump are calculated for the case that
population decay is turned off. The results reveal that probing the distribution with a p-polarized
probe (shown in black) leads to a large sensitivity to azimuthal anisotropy in the difference of p-
and s-pump. The average of the two quantities shows little sensitivity. Color key: Red t = 0, purple
t = 0.25T , orange t = T , where T is the decay time associated with in-plane reorientation.

average of p-pump and s-pump

difference of p-pump and s-pump

Figure 2.13: The average and difference of p- and s-pump are calculated for the case that no
reorientation is allowed. The results reveal that probing the distribution with a p-polarized probe
(shown in black) leads to a large sensitivity to population decay in both quantities. Color key: Red
t = 0, purple t = 0.25T , orange t = T , where T is the decay time associated with in-plane
reorientation.

24

-1.0-0.50.51.0-1.0-0.50.51.0-1.0-0.50.51.0-1.0-0.50.51.0-1.0-0.50.51.0-1.0-0.50.51.0-1.0-0.50.51.0-1.0-0.50.51.02.3.1.3 Mathematical description of orientational dynamics

Introducing a pump pulse requires the second order susceptibility to have an additional term that

describes the interaction of the pump with the sample (pump-induced bleach) and its evolution until

the probe beam interaction. The following equation 2.66 can be understood with the help of the

pump off

(cid:12)(cid:12)(cid:12)2

(cid:12)(cid:12)(cid:12)χ(2)

Bloch sphere. The probe SFG intensity Ipump on after the pump interaction is proportional to the
square of the difference between ground state N0 and first excited state N1 population. Normalized to
∝ 1), we get Ipump on ∝ (N0 − N1)2. On the Bloch
the pre-pump intensity (Ipump off ∝
sphere, N0 − N1 is the projection of the Bloch vector onto the z-axis and can be described with the
tipping angle θ: N0−N1 = cosθ. Therefore, Ipump on ∝
pump on ∝ cosθ.
For a sparsely populated excited state (N0 (cid:29) N1), the small-angle approximation (cosθ ≈ 1−θ 2/2)
pump on ∝ 1− θ 2/2. The tipping angle θ is related to the electric field of
can be applied so that χ(2)
the pump and its pulse duration (θ ∝ |Epu|Tpu), so we arrive at χ(2)
pu/2. It is
now possible to recognize and interpret the terms in a mathematically more rigorous description of

pump on ∝ 1−|Epu|2T 2

∝ cos2 θ = or χ(2)

(cid:12)(cid:12)(cid:12)χ(2)

(cid:12)(cid:12)(cid:12)2

pump on

the second-order probe susceptibility:

χ(2)
pump on,i jk(ωIR,t) =

(cid:68)
(cid:69)
(ˆi·←→α (Ω)· ˆj)(µ(Ω,ωIR)· ˆk)
(cid:28)
(cid:12)(cid:12)(cid:12)2(cid:29)
(cid:12)(cid:12)(cid:12)µ(Ω0)· ˆl
(ˆi·←→α (Ωt)· ˆj)(µ(Ωt)· ˆk)G(Ωt, Ω0)
(cid:12)(cid:12)2
i jkll(ωIR,t)(cid:12)(cid:12)Epu,l
pump off,i jk(ωIR)− R(4)

Ωt

+

= χ(2)

(2.65)

(cid:12)(cid:12)Epu

(cid:12)(cid:12)2 T 2

pu

¯h2

(2.66)

Ωt ,Ω0

where R(4)
i jkll(ωIR,t) is the fourth-order temporal response function, Epu is the electric field of
the pump and Tpu its temporal length, µ(Ωt,ωIR) is the IR dipole derivative, ←→α is the Raman
polarizability tensor, Ω(t) = (θ (t),φ (t)) is the molecular orientation at time t, and G(Ωt, Ω0) is
the Green’s function describing the temporal evolution of the molecular orientation between times 0
and t. The angled brackets (cid:104)(cid:105)Ωi indicate averages over each Ωi.
Gengeliczki et al. [25] showed how the Green’s function can be written using spherical harmonics

25

Figure 2.14: Schematic illustration of the pump induced bleach for p-polarized pump(cid:0)βp(t)(cid:1),

s-polarized pump (βs(t)), and scaled s-polarized pump.

and the rotational diffusion constant Dor

G(Ωt, Ω0) =

∞
∑
n=0

n
∑
m=−n

n (Ωt)Y m∗
Y m

n

(Ω0)exp (−n(n + 1)Dort)

(2.67)

In our experiment the pump is either s- or p-polarized. Hence, after taking into account the properties

of the spherical harmonics, the C∞,ν symmetry of the water interface, and assuming no population
relaxation and free diffusion of the dOH within a cone of half angle θ0, the response function can
be written as:

p-pump: Rzzzzz(t) ≈ Azz(θ0) + Bzz(θ0)exp(−kort)
s-pump: Rzzzyy(t) ≈ Ayy(θ0) + Byy(θ0)exp(−kort)

The reason for considering only the zzzzz and zzzyy contributions is motivated by the discussion of

the Fresnel coefficients in the vicinity of TIR in section 2.2.6.3. To experimentally access the rate of

molecular reorientation kor with the two available pump-polarizations, we construct the following
(cid:18)
expression, which is proportional to exp(−kort):
Bzz(θ0)− Azz(θ0)
Ayy(θ0)

Rzzzzz(t)− Rzzzzz(t (cid:29) T−1
or )
Rzzzyy(t (cid:29) T−1
or )

Rzzzyy(t) =

(cid:19)

Byy(θ0)

exp(−kort)

(2.68)

26

Figure 2.15: Schematic plot of the difference and average of βp and βs for time constants observed
in the experiment.

It is important to note for later that the coefficients before Bzz(θ0) and Byy(θ0) have opposite signs.
In our experiment, we measure the pump-induced bleach β (t), depicted in figure 2.14, versus the

pump-probe delay t. We do this by measuring the SFG intensity at a given delay with the pump on,
Ion
SFG(t), and the pump blocked, Ioff

SFG, and compute:

β (t) ≡ 1− Ion
SFG
Ioff
SFG

(2.69)

Furthermore the bleach of the different pump polarizations s and p is proportional to the correlation

function Ri jkll:

βs(t) ∝ Rzzzyy(t)

βp(t) ∝ Rzzzzz(t).

(2.70)

(2.71)

The orientational dynamics kor can now be obtained when one calculates the difference between βp
and the scaled βs analogously to equation 2.68 (pictured in figure 2.15):

∆β (t) = βp(t)− βp(t (cid:29) k−1
or )
βs(t (cid:29) k−1
or )

βs(t) ∝ exp(−kort)

(2.72)

27

Up to this point, the population relaxation time T1 = 1/kdOH has been neglected, but it can be
phenomenologically inserted:

∆β (t) ∝ exp (−(kor + kdOH)t) = exp

(2.73)

(cid:16)−k∆βt

(cid:17)

An associated quantity that will help to disentangle the orientational dynamics and the population

relaxation is the average of βp and the scaled βs:

(cid:32)

(cid:33)

βs(t)

(cid:18)

¯β (t) =

1
2

βp(t) +

βp(t (cid:29) k−1
or )
βs(t (cid:29) k−1
or )
1
∝ Azz(θ0)exp (−kdOHt) +
2

(cid:19)

Byy(θ0)

Azz(θ0)
Ayy(θ0)

exp (−(kor + kdOH)t)

(2.74)

Bzz(θ0) +

In equation 2.74, as noted earlier, the opposite signs of Bzz and Byy reduces the contribution of the

second term, which includes orientational dynamics, in comparison to the first term, which only
depends on population decay. To conclude, by measuring ∆β (t) (equation 2.73) and ¯β (t) (equation

2.74) the two unknowns kor and kdOH can be evaluated.

Figure 2.16: Schematic overview of the dangling OH stretch mode decay mechanisms.

28

kIETF/BkjumpF/BkIVRbkp-pb-bkp-pFkp-pd-dkp-pBkIVRd2.4 Decay mechanisms of the dangling OH stretch mode

An excited OH stretch mode can decay through several different channels which are schemati-

cally shown in figure 2.16. The contributing mechanisms are molecular reorientation, intramolecular

vibrational relaxation (IVR), intramolecular energy transfer (IET) and dipole-dipole interactions

(p-p). Isotopic dilution of H2O with D2O can help suppress some of these mechanisms and may
help disentangle their interplay with orientational dynamics. A brief discussion of isotopic dilution

follows in section 2.5.

2.4.1 Molecular reorientation - kF/B
jump

Molecular reorientation is the mechanism of interest in this dissertation. A water molecule at

the surface with a dOH sticking out of the bulk water is typically still involved in three other

hydrogen bonds: One is formed between the OH group that is oriented tangentially to the surface

and lone pair electrons associated with an oxygen atom of another water molecule. The other two

hydrogen bonds form between the lone electron pairs of the oxygen and hydrogen atoms of two
different, neighboring water molecules. The molecular roerientation happens as a ∼ 90◦ rotation
around the bOH axis of the molecule so that the dOH that was sticking out of the bulk is now

tangential to the interface and able to form a hydrogen bond with other water molecules. The rate
constant for this “forward” reorientation process is denoted kF
or “backwards”, process is denoted kB

jump. The generally slower reverse,
jump. The backwards process is slower because there are
many more tangentially oriented OH groups at the surface than dangling OH groups. Since detailed

balance must hold at equilibrium, the rate of forward jumps must exceed that of backward jumps.
jump = (21ps)−1 at the
MD simulations by our collaborators suggest that kF

jump = (1.6ps)−1 and kB
interface between water and an idealized hydrophobic surface.[10] Therefore,

jump = R× kF
kB

jump

(2.75)

where R ≈ 1/13. However, at the water/air interface simulations by Vila Verde et al. and experiments
jump = (1.9ps)−1 - significantly faster
by Hsieh et al. indicate that kF

jump = (0.8− 1.1ps)−1 and kB

29

compared to the water/OTS interface.[17, 16] Using the numbers reported in the paper of Hsieh et
al.[17] yields R ≈ 1/1.7.

2.4.2

Intramolecular vibrational relaxation - kd/b
IV R

IVR occurs when an excited OH stretch vibration ν = 1 couples to the ν = 2 overtone of the
molecular bending mode at ≈ 3150cm−1.[26]. This process occurs in the bulk with the rate kb
IV R
where the initial OH stretch vibration involves a bOH group at 3000− 3500cm−1. In principle IVR
can also occur for a vibrationally excited dOH with the rate kd
IV R. Since the dOH stretch vibration
mode is about 530cm−1 to the blue of the ν = 2 bend overtone, the lack of spectral overlap prevents
this channel from being a significant decay pathway and kb
IV R. The bulk rate in H2O has
IV R = (0.2ps)−1 in several investigations.[26, 27, 28, 29] In the case of
been determined to be kb
IV R ≈ (0.75ps)−1, where the tilde
isotopic dilution, the intramolecular vibrational relaxation is ˜kb
indicates a quantity associated with isotopic dilution.[30] Since the relaxation from the overtone of

IV R (cid:29) kd

the molecular bending mode to its first excited state is very rapid,[26] IVR is for our purposes an

irreversible process.

2.4.3

Intramolecular energy transfer - kF/B
IET

IET from an excited dOH (≈ 3680cm−1)
IET describes the transfer of energy with the rate constant kF
to a bOH (≈ 3500cm−1) on the same water molecule.[31] kF
IET can be calculated through perturba-
tion theory. The wave function of the excited dOH stretch of an interfacial H2O molecule can be
written

|φdOH(cid:105) ∝

γ
δ

|b(cid:105) +|d(cid:105)

(2.76)

with |b(cid:105) and |d(cid:105) being the localized vibrations on the uncoupled bOH and dOH groups, γ being the
coupling between bOH and dOH stretch modes, and δ being the frequency difference between the

uncoupled modes.

30

2.4.3.1 Calculation of γ and δ

γ and δ can be derived for H2O from the work of Stiopkin et al. which focuses on IET in the
case of D2O (and its isotopic dilution with H2O).[32] There, the coupling between a dOD and a
bOD on the same D2O molecule is γD2O = −48cm−1. This is in contrast to the gas-phase value of
−60cm−1. Similarly, for H2O with a gas-phase intramolecular coupling of −50cm−1 this would
lead to a value of

γ =

−50
−60

× (−48cm−1) = −40cm−1.

(2.77)

The detuning δ of the uncoupled modes for H2O can also be derived from what is known about
D2O: First, the observed frequency difference δ(cid:48)
D2O between the dOD and the bOD, which includes
coupling, is calculated as follows:
dOD − ω(cid:48)

bOD = 2745cm−1 − 2580cm−1 = 165cm−1.

(2.78)

δ(cid:48)
D2O = ω(cid:48)

The next step is to calculate the detuning of the uncoupled modes δD2O of dOD and bOD, which
each experience a blue- and redshift of 17cm−1, respectively, caused by their coupling:

δD2O = δ(cid:48)

D2O − 2× 17cm−1 = 131cm−1.

(2.79)
It is important to note that the surface bOD peak at 2580cm−1 does not coincide with the bulk bOD
peak at 2480cm−1 (but is well within the broad bulk bOD feature), which indicates that the average
bulk bOD is involved in hydrogen bonds that are stronger than the surface bOD. The bulk bOD
and the dOD are spectrally separated by 2745cm−1 − 2480cm−1 = 265cm−1. The same should
apply for the H2O case. Therefore, scaling δD2O by frequency. Stiopkin et al. use 300cm−1 for the
frequency difference between dOH and the bOH on the same H2O molecule and 250cm−1 for the
dOD and the bOD on a surface D2O molecule. This leads to

and finally results in

δH2O = δD2O × 300cm−1
250cm−1 = 157cm−1,
(cid:12)(cid:12)(cid:12) = 0.24.

(cid:12)(cid:12)(cid:12) γ

δ

31

(2.80)

(2.81)

2.4.3.2 Calculation of kF

IET and kB
IET

During IET the excited dOH transfers its energy to the uncoupled bOH. Therefore we calculate:

(cid:12)(cid:12)(cid:12) γ

δ

(cid:12)(cid:12)(cid:12)2

|(cid:104)b|φdOH(cid:105)|2 ∝

= 0.242

(2.82)

Since the bOH lifetime is governed by intramolecular vibrational relaxation[30] described in the
previous section, kF

IET can be estimated to be:

IET ≈ 0.242kb
kF

IV R = (13ps)−1

(2.83)

The process of IET described so far is reversible. However the “backwards” process kB

IET , from an
excited bOH to a dOH, is not energetically favorable and therefore slower compared to the forward
process. Using the Arrhenius equation one can establish the ratio between kF

IET , assuming
that the system is in equilibrium with respect to IET and that there is there is no energy barrier.[17]

IET /kB

This means that the energy gap between dOH and bOH can be used for the calculation:

(cid:18) ∆E

(cid:19)

kBT

kF
IET
kB
IET

= exp

(cid:32)

= exp

hc(3680− 3500)cm−1

kBT

= 2.42

(2.84)

(cid:33)

where T = 293K has been used. Hence, using the ratio RIET we can express kB

IET in terms of kF

IET :

IET = RIET × kF
kB

IET =

1

2.42

× kF

IET = 0.41× kF
IET

(2.85)

RIET is expected to change with isotopic dilution because the bOD feature is located at ≈
2500cm−1.[33] Using equation 2.84 leads to ˜RIET = 0.003.

2.4.4 Dipole-dipole interactions - ki− j
p−p

At the interface, dipole-dipole interactions can couple an excited dOH stretch vibration with dOHs

or bOHs of other molecules. The interaction with another dOH is described by the rate constant
kd−d
p−p. The coupling with a bOH is described by kd→b
and by kb→d
an excited bOH with a bOH of another molecule with the rate constant kb−b

p−p to indicate a “forward” process
p−p to describe a “backward” process. Dipole-dipole interactions also couple
p−p. The rates kd−d
p−p,

p−p or kF

p−p or kB

32

p−p and kb−b

p−p in general differ from each other. Cowan et al. determined kb−b,bulk
kF
p−p, kB
(0.075ps)−1 for bulk H2O.[29] Since we are only sampling water molecule at the interface, the rate
is potentially further decreased by a factor of 1/2 due to the absence of OH groups in half of the
solid angle: kb−b
p−p ≥ (0.150ps)−1. Upon isotopic dilution, this rate decreases according to Förster’s
theory:[34, 13]

p−p

=

(cid:32)

k =

(cid:33)2 r6

0
T1

4π3/2

3

[OH]

∝ [OH]2 ∝ 1
r6

(2.86)

where [OH] is the number concentration of OH groups, r0 is the Förster radius, T1 is the lifetime
of the excited state without the Förster process and r is the average distance between OH groups.

p−p = 1/9× kb−b

Isotopic dilution according to the ratio 1:2 H2O:D2O lowers the OH concentration [OH] by a factor
of 2/3, resulting in a reduction of the undiluted rate by a factor of (1/3)2: ˜kb−b
p−p ≥
(1.35ps)−1. Moreover, non-Condon effects (the strong dependence of the dipole derivative of an
OH stretch on its hydrogen-bonding environment) reduce the dOH dipole transition matrix element
by a factor of 1/3 relative to that of bOH, so that in combination with the dipole-dipole r−6 distance
dependence and the small spectral overlap between dOH and bOH stretch excitations, it is possible
to neglect kd−d
p−p.[35, 36, 37] In our experiment, dipole-dipole interactions between
bOH groups act as an irreversible process: An initially excited dOH transfers its energy via IET
to the bOH of the same molecule, from here following the fastest processes are IVR and the kb−b
p−p
dipole-dipole interactions. While IVR leads to vibrations red shifted from our detection window,
kb−b
p−p effectively channels the excited dOH stretch population away from the OH groups at the
surface and dissipates the energy into the bulk and can therefore be regarded as irreversible for our

p−p and kB

p−p, kF

purposes.

33

Figure 2.17: Overview from Xiao et al.[4] of the processes involved in the decay of the excited
dangling OH stretch vibration.

H2O

term value
kb
IV R
kb−b
p−p ≥ (0.15ps)−1
kF
IET
RIET
R

(0.2ps)−1
(13ps)−1
0.37
1/13

source
[27, 28, 29]
[29]
[30, 32]
[17]
[10]

H2O:D2O::1:2

value
(0.75ps)−1
≥ (1.6ps)−1
(cid:28) (13ps)−1
0.0025
1/13

source
[30]
[13]
[30, 32]
[17]
[10]

Table 2.1: Summary of the rate constants of the dominant processes involved in relaxation of the
pump-probe SFG signal.

34

2.5 Isotopic dilution

Isotopic dilution of H2O with D2O can be of great aid in suppressing certain decay mechanisms
and better identifying the dominant mechanisms involved in the decay of the pump-probe SFG

signal. The introduced D2O leads through dissociation of H2O and D2O and exchange of H and D
to the formation of HDO molecules. Sufficiently diluted with D2O, the heavy water molecules and
OD groups effectively isolate the excited dOH spatially and spectrally. The vibrational excitations of

OD groups are at a significantly lower frequency than the excitations of the OH groups. If we focus

on an HDO molecule with an excited dOH group, the lack of spectral overlap between dOH and

bOD will effectively eliminate IET from dOH to bOD. In addition, compared to the case of H2O the
overtone of the bending modes of D2O and HDO have less overlap with the dOH and bOH stretch
vibration, slowing the IVR rate considerably. Reducing the density of OH groups also reduces the
dipole-dipole interactions (dOH→dOH, dOH→bOH, bOH→bOH) due to the increased distance
between OH groups. The reorientation time is not expected to be affected substantially by isotopic

dilution. However, decreasing the H2O:D2O ratio decreases the SFG intensity proportionally to
N2
dOH, the square of the number density of dOH groups at the interface. We use a molar ratio
1:2 of H2O:D2O as a compromise between loss of signal and having the majority of dOH groups
associated with HDO molecules.

35

CHAPTER 3

EXPERIMENTAL SETUP

In order to perform SFG we require an IR source that can be tuned to be resonant with the dOH
stretch vibration at 3680cm−1. Since the SFG signal is relatively weak, there is need for substantial
IR power too. Moreover, to resolve the fast processes presented in the previous chapter, it is

necessary that those IR pulses are short enough. At the same time, it is of advantage to generate the

SF with a narrowband and long VIS pulses to provide the required spectral resolution. This chapter

elaborates on the setup of such a system and the signal processing that leads to the data presented in

the next chapter.

3.1 Laser source

Our primary laser source is the Spectra Physics Spitfire PRO-XP, which uses a mode-locked

Ti:sapphire Mai Tai oscillator as the seed laser and a pulsed green solid state Spectra Physics
Empower Nd:YLF laser as the pump laser. The system is designed to produce ≈ 4mJ/pulse at

Figure 3.1: Schematic illustration of the Spectra Physics Spitfire PRO-XP system.

36

Figure 3.2: Schematic overview of the experimental setup used to investigate the dangling OH.

800nm with a full-width half-maxiumum bandwidth of 13nm at a repetition rate of 1kHz and

< 120fs pulses. About 2mJ/pulse are used for the water experiments.

3.2 Optical parametric amplifier (OPA)

In order to create infrared beams that are resonant with the dangling and bonded vibrational

OH stretch modes there are several stages involved, the first one being an OPA based on the design

of Kaindl et al.[38] 1mJ/pulse is used to power a homemade OPA with two pump stages. A beam

splitter (BS) splits the 800nm beam into the seed and pump beam. The seed 800nm intensity is

regulated with the combination of a thin film polarizer and a λ /2 waveplate and is then focused

by a f = 75mm lens into a 1mm c-cut sapphire crystal for white light supercontinuum generation.

A spherical mirror, with f = 25mm, placed approximately one focal length from the sapphire is

used to focus the white light into the nonlinear optical crystal. The 800nm pump beam is focused

with a f = 500mm lens and is split with a beam splitter into pumps for stage 1 and stage 2. The
stage 1 pump is sent co-linearly with the white light into a β -BaB2O4 crystal (“BBO”, 5x6x4mm3,
θ = 28◦, φ = 0◦, P/P) for type II phase-matching using a dichroic mirror (Side 1: R ≥ 99.0% for
775− 825nm s polarization, Tavg ≥ 96% for 950− 3200nm p polarization; side 2: R ≤ 0.5% for
950− 1600nm p polarization). The created signal and idler are retro-reflected at a small non-zero

37

pulseshaper800nm1kHz100fs1.13μm2.72μmprobepumpdelaynarrow800nmdelaydelayOPADFGdoubling2.26μmidlersamplepurged with dry N2 Figure 3.3: Schematic overview of the homemade OPA. The red beam represents the 800nm, the
green beam is the supercontinuum white light, the orange beam is the signal and idler after the first
pass through the BBO crystal, the brown beam is its reflection from the spherical mirror and the
cyan beam is the signal and idler after the second pass through the BBO crystal.

angle into the BBO by a f = 250mm spherical mirror placed about one focal length away from the
BBO crystal. The 800nm for stage 2 is re-collimated by a f= −150mm lens and then goes through
the BBO co-linearly with the retro-reflected signal and idler to pump the second pass.

3.3 IR resonant with the dangling OH stretch vibration

The second stage generates the IR that is resonant with the dangling OH stretch vibration
(3680cm−1) via difference frequency generation (DFG) between the 800nm (12500cm−1) and the
doubled OPA idler (2× 4410cm−1 = 8820cm−1). The OPA idler is doubled with a BBO crystal,
its beam diameter expanded by a telescope, and then co-linearly sent through a KTiOPO4 (“KTP”,
10x10x6mm3, θ = 42◦) crystal along with the 800nm beam. Just before the KTP crystal, the 800nm
and doubled idler beam enters a housing through a CaF2 window. The housing is purged with
dry nitrogen gas which allows the generated IR to propagate without getting absorbed by ambient

38

Pλ/4BBOsapphiref=25mmf=500mmf=-150mmf=75mmf=250mmBSBSDMDMwater vapor. A 2.4 µm longpass filter after the DFG crystal is used to block residual light from the
doubled idler and the 800nm beam. The bandwidth of the IR was measured to be ≈ 45cm−1 at the
full-width half maximum.

3.4 Sum frequency generation with IR pump/probe and VIS 800nm

For SFG, the IR and VIS pulses are overlapped spatially and temporally at the sample. The
IR = 69.8◦, and

beam angles in the experiment and shown in figure 3.4 are θ pump
θVIS = 76.5◦. Further details follow in this section.

IR = 52.0◦, θ probe

3.4.1

IR pump and probe beams

After the vibrationally resonant IR is generated, the beam is split into pump and probe beams with a

ZnSe window. The reflected beam is the probe and the transmitted beam is the pump, which is sent

over a LabView-controlled mechanical delay stage, which determines the delay between pump and

Figure 3.4: Experimental schematic of the beam geometry and sample. The fused FS prism is
coated with an OTS SAM. The dotted blue line is the pump SFG (pump+VIS) which is discarded
using an iris during the experiment but is of importance for alignment purposes.

39

θSFGVISpumpθIRθVISprobeSFGprobespectrometer+CCDpump onpump offzyxTeflon cellOTS SAMFS prism250HzFigure 3.5: Illustration of the spectral narrowing by the pulse shaper.

probe beams during the experiment. On the way to the sample, the polarization of both beams is
controlled with CdGa2S4 λ /2 (Altechna, anti-reflection coated, zero order, 3000nm) waveplates.
Furthermore, the pump intensity is regulated by using a polarizer in conjunction with the waveplate.

When the IR is tuned to the dOH stretch vibration, the IR beams leave the housing purged with dry

nitrogen through a 2.4 µm longpass filter about 125mm before the sample. Pump and probe beams

are then focused with CaF lenses, f = 100mm and f = 80mm respectively, onto the sample.

3.4.2 VIS 800nm

For generating the sum-frequency at the sample, we use a 4f pulse shaper to spectrally narrow the
800nm beam from a full-width half-maxiumum (FWHM) of 12.5nm to 0.5nm (8cm−1) as depicted
in figure 3.5. A narrowband 800nm in combination with a broadband IR has the advantage that there
is effectively a one-to-one correspondence between the SFG and IR spectra: ωIR = ωSFG − ωVIS.
The pulse shaper, displayed in figure 3.6, consists of a grating with 1200 grooves/mm, a f = 500mm

spherical mirror and a mirror with an adjustable slit. The spherical mirror is placed one focal length

away from the grating to compensate the angular dispersion caused by the grating and to focus the

different wavelength components at different lateral positions on the the slit/mirror, which is placed

one focal length away from the spherical mirror. Translating the narrowed slit perpendicular to the

incoming beams selects for a narrow band. The flat mirror is tilted so that the spectrally narrowed

40

Figure 3.6: Schematic overview of the 4f pulse shaper. To illustrate the pulse shaper’s operation,
the figure uses three distinct colors (red, green, blue) which would correspond to the red tail, the
center and blue tail of the 800nm spectrum.

beam comes out slightly below the original optical plane and can, therefore, be redirected with a

mirror. Next, the spectrally narrowed beam goes through a delay stage, a λ /2 waveplate, and a thin

film polarizer which enables us to control the delay with respect to the pump/probe IR beams, its

power, and polarization. Finally, a f = 500mm lens is used to focus the beam onto the sample.

3.5 Signal detection

At the sample, the VIS and both IR beams are overlapped spatially and temporally to create

two SFG signals: One from VIS + pump IR and one from VIS + probe IR. The data in this

text has been detected in upp configuration (SFG signal unanalyzed, probe and VIS p-polarized).

voltage
standby current
operation current

slave master
23.6V 23.6V
0.18A 0.21A
0.31A 0.35A

Table 3.1: Galvanometer power supply (Topward Dual-Tracking DC power supply 63061) settings

41

Figure 3.7: Schematic of the galvanometer-driven separation of “pump-on” and “pump-off” beams
where α(t) is the galvanometer mirror angle at time t. The gray (red) laser pulses indicate the probe
SFG signals corresponding to by the chopper blocked (unblocked) IR pump pulses.

Note: There are other signals from different orders of photon interaction but they are at different

wavelengths and propagate into different directions than the beams of interest. Figure 3.4 gives

a simple overview of the relevant beams close to the sample. The signals are collimated by a

f=100mm lens, placed one focal length away from the sample. Just after the lens, the residual

800nm VIS and pump SFG beam are blocked by an iris that allows only the probe SFG beam to

pass. As described by Ghosh et al.,[39] a galvanometer (Cambridge Technology 6230H) and an

optical chopper (Thorlabs), both synchronized with the laser, are used to separate the signal into

two beam trajectories: One corresponds to a “pump-off” measurement where the pump is blocked

by the chopper wheel so it does not hit the sample. The probe SFG signal in this case serves as

a reference measurement. The other one corresponds to the “pump-on” measurement where the

pump pulse clears the chopper wheel and excites the sample. More specifically, the chopper and

the galvanometer frequency and phase are set so that two consecutive pump laser pulses clear the

chopper and the two corresponding probe SFG signals hit the galvanometer when the mirror is at

the “pump-on” angle. The next two pump laser pulses are then blocked by the chopper, and their

corresponding probe SFG signals hit the galvanometer at the “pump-off” angle. Figure 3.7 provides

an illustration of the galvanometric scheme, and table 3.1 shows the galvanometer power supply

settings (Topward Dual-Tracking DC Power Supply 6306D) used during the experiment. Finally,

42

a 150mm lens is used to focus the beam into the spectrometer slit. To further reduce background

light, unwanted wavelengths are blocked by a bandpass filter with a center wavelength of 620nm

and FWHM 60nm. The signal enters a Princeton Instruments Acton SP 2300 spectrometer with a

1200lines/mm grating and is collected by a liquid nitrogen cooled, two-dimensional CCD camera

from Princeton Instruments (Spec-10:400B/LN). The spectrometer is calibrated with a mercury

lamp, in particular by using the yellow-orange line pair at 576.959nm and 579.065nm. According
to the manufacturer, the spectrometer and camera together have a resolution of 3.46cm−1 at the
sum-frequency (≈ 618nm). This in combination with the VIS bandwidth of 8cm−1 yields a
resolution for the detected sum-frequency spectra of

(cid:113)
(3.46cm−1)2 + (8cm−1)2 ≈ 9cm−1.

(3.1)

The y-axis of the chip allows separation of pump-on and pump-off spectra, whereas the x-axis

represents the wavelength λ of the detected photons. Before the experiment, we choose a set of
pump-probe delays {t1, . . . ,tN}. Then, at a given pump-probe delay tk ∈ {t1, . . . ,tN}, the galvanome-
ter creates two vertically separated horizontal regions on the CCD with the intensity Ion/off(λ ,y;tk).
The CCD exposure ranges typically from a few seconds up to 24s. Vertical integration of the CCD

signal in each of the two regions of interest (ROI) yields the spectra of the pump-off and pump-on
signals of the form Ion/off(λ ;tk) as illustrated in figure 2.7. At this point, the two spectra are backed
up on the computer hard disk drive and the pump-probe delay altered to the next predetermined time.

Once one sweep through the set of delays is complete, the next sweep is initialized automatically
until the program is stopped after M scans. The labeling of our data is now Ion/off
(λ ;tk), where
l ∈ {1, . . . ,M}. In some cases, we had to intervene here to correct for distortions by cosmic rays. A
detailed description of this procedure can be found in section 3.5.1. The spectra corresponding to

l

the same delays are then averaged across the sweeps:

¯Ion/off(λ ;tk) =

1
M

M
∑
l=1

Ion/off
l

(λ ;tk)

(3.2)

43

Figure 3.8: Schematic illustration of a typical normalized intensity trace Inorm(tk).

The next signal processing step is to integrate those spectra over a chosen range of wavelengths to

count the total number of detected photons for the pump-on and pump-off signals:

Ion/off(tk) =

λ2∑

λ =λ1

¯Ion/off(λ ;tk)

(3.3)

This process yields the two scalars: Ion(tk) and Ioff(tk). To display the dynamics of the system a
common way to plot the data (as seen in figure 3.8) is to compute the on/off ratio

Inorm(tk) =

Ion(tk)
Ioff(tk)

(3.4)

versus the pump-probe delay tk. A homemade LabView program is used to coordinate data collection,
the control of the delay stage and the calculation of Inorm(tk). Probing the sample before the pump
interaction at time t = 0, yields Inorm(t < 0) ≈ 1. During and after the pump interaction, a dip in
Inorm(t ≈ 0) < 0 indicates bleaching, which recovers at long delays. For further investigation, the
bleach is calculated (figure 3.9):

β (tk) ≡ 1− Inorm(tk)

(3.5)

44

0510152025300.850.900.951.00delay[ps]NormalizedintensityInorm(t)Figure 3.9: Schematic illustration of a typical trace of the pump-induced bleach.

3.5.1 Cosmic ray removal

The CCD chip is sensitive to cosmic rays which result in spikes in the collected spectra. To

achieve the best possible signal-to-noise ratio the data is scanned by a Matlab script which removes
the cosmic ray spike. An algorithm determines if the derivative of the spectra ∆I
∆λ exceeds a
predetermined value. If such an abnormality is found, for example in the spectrum corresponding to

scan l at delay tk, it is replaced by the average of the preceding and succeeding scan as illustrated in
figure 3.10:

Il(λ ;tk) =

Il−1(λ ;tk) + Il+1(λ ;tk)

2

.

(3.6)

The basis for the algorithm is the assumption that it is unlikely that cosmic rays interact with the

same detector chip pixel column (which determine the wavelength) during consecutive scans. This

procedure is performed for pump-on and pump-off spectra.

45

0510152025300.000.050.100.15delay[ps]Pumpinducedbleachβ(t)Figure 3.10: Schematic of the data collection and cosmic ray removal process. Each box represents
a spectrum collected during scan number l at pump-probe delay tk.

3.6 General procedure

After overlapping the IR and VIS beams with a pinhole, a BBO crystal is put in place to roughly

find the temporal overlap of the beams by maximizing the intensity of the VIS+probe and VIS+pump

SFG signals by monitoring the intensity on a white business card by eye while manually moving

the delay stages. Usually, after the temporal overlap has been found, small adjustments of the IR

beam mirrors are made to increase the spatial overlap. Next, an Ag mirror is positioned at the

location of the overlap. The mirror is used to further fine tune both the spatial and temporal overlap

by monitoring the pump and probe SFG signals through the monochromator. At this point, it is
advisable to confirm that the SFG signal occurs at the intended wavelength (ωSFG = ωIR + ωV IS).
To accomplish this, an accurate measurement of the VIS spectrum ωV IS has to be performed. Only

46

delayscan number....................................................................................Cosmic ray eventt1tk-1tktk+1tN-1tN12MM-1l+1ll-1t2Il(λ;tk)IM(λ;tN)I1(λ;tN)I1(λ;t1)I(λ;t1)I(λ;t2)I(λ;tk-1)I(λ;tk)I(λ;tk+1)I(λ;tN-1)I(λ;tN)Il-1(λ;tk)Il+1(λ;tk)then can one proceed with calculating the IR spectrum with confidence. Since the VIS is spectrally
very narrow it is sufficient to use the peak frequency ω∗
V IS of its spectrum for the calculation.
Furthermore, the IR range that we are interested in is far away from any resonances of Ag.[40]
This means the SFG signal from the Ag mirror can be directly converted (ωIR = ωSFG − ω∗
V IS)
into the spectrum of the IR as plotted in figure 4.1. Should the IR not be in the desired spectral

range, adjustments upstream are necessary. Otherwise one can put the sample in place. Next, it

is usually only a matter of translating the sample, while making sure that the reflected VIS beam

travels to the monochromator. In our case, the VIS is an appropriate guide because the SFG signal

will be close enough to the VIS beam due to the collection lens placed one focal length from the

sample. One should be able to see the SFG signal from the water-OTS interface on the CCD after

patiently translating the sample in the optical plane. However, after having found the SFG signal,

further tweaking of spatial and temporal overlap is usually needed as well as adjustment of the focal

points of the IR beams because of refraction of the beams upon entering the fused silica prism.

Since the Rayleigh length of the VIS beam is larger than those of the IR beams, the VIS beam lens

rarely needs adjustment at this point. In contrast, the SFG signal is more sensitive to the IR lens

position. Translating the pump and probe lenses parallel to the beams by only a few millimeters

can increase the SFG signal significantly if the IR beams were not focused properly at the sample.

After maximizing the pump and probe SFG signals, the third-order signal originating from the

mixing of VIS, IR probe, and IR pump is maximized to ensure that not only VIS + probe and

VIS + pump are individually overlapped but that all three beams are optimally overlapped. Since
the three third-order signals, ωV IS + 2ωpump, ωV IS + 2ωprobe, ωV IS + ωpump + ωprobe, emerge at
the same wavelength one needs to take advantage of their spatial separation and physically block
ωV IS + 2ωpump and ωV IS + 2ωprobe to prevent them from getting to the detector. The remaining
signal can now be used to establish the pump-probe delay t = 0 and to determine the instrument

response function as seen in figures 4.2 and 4.3.

47

Figure 3.11: Chemical structure and space-filling-model of octadecyltrichlorosilane, CH3 (CH2)17
SiCl3. Colour code: Carbon, C: grey-black; Hydrogen, H: white; Chlorine, Cl: green; Silicon, Si:
turqouise-grey. [5]

3.7 Sample preparation

For all experiments deionized H2O (Millipore, ≥ 18MΩ) was used. The D2O for isotopic
dilution was obtained from Sigma-Aldrich (99.9%). A homemade Teflon cell was machined and
covered with an IR grade fused silica isosceles trapezoidal prism with a base angle of 60◦. The
polar silanol groups terminating the fused silica prism provide a hydrophilic surface with which

water molecules can form hydrogen bonds. In order to investigate dangling, non-hydrogen-bonded

OH groups the fused silica prism is made hydrophobic by growing a self-assembled monolayer

of octadecyltrichlorosilane (OTS), following a variation of the procedure outlined in the work

by Sagiv.[41] The OTS molecule has a hydrophobic part consisting of a long alkyl group and a

hydrophilic polar head group. Storage of OTS, therefore, needs to be in a dry atmosphere to prevent

the OTS from reacting with any water. Furthermore, it is important that during the deposition

process of the self-assembled monolayer, all beakers need to be pretreated and baked to get rid of

excess water molecules.

48

Figure 3.12: OTS coated prism. A circular fit yields a contact-angle of θC = 100.4± 1.4◦. An
elliptic fit yields a contact-angle of θE = 104.8± 0.94◦.

3.7.1 Growing the octadecyltrichlorosilane self-assembled monolayer (OTS SAM)

To create the SAM on the fused silica prism the following protocol was used. First, the glassware is
prepared by washing it with acetone, then methanol, followed by baking it at 150◦C for 1 hour. The
fused silica prism is pre-cleaned by sonicating it in acetone, then methanol and finally de-ionized

water, each for 15 minutes. The prism is then placed into Nochromix oxidizer (Godax Laboratories)

for several hours. In a separate beaker, 150mL of solvent mixture is prepared by combining 120mL

n-hexadecane (Fisher Scientific, 99% ), 18mL carbon tetrachloride (Sigma-Aldrich, 99.5%) and

12mL chloroform (DriSolv, 99%). The solution is stirred for 10 to 15 minutes. The prism is then

taken out of the Nochromix, rinsed with deionized water and blown dry with dry nitrogen gas. Next,

0.15mL OTS (VWR, 97%) is added to the solvent mixture and stirred until homogenized. Finally,

the prism is placed into the OTS solution. The deposition is performed for 30 minutes while stirring.
Afterward, the prism is rinsed in chloroform and dried for 1 hour at 50◦C.

49

Figure 3.13: Setup by Lamour et al.[6] for the contact-angle measurement optimized for analysis
with “ImageJ”.

3.7.2 Contact-angle measurement

To determine the quality of the OTS monolayer, a contact-angle measurement is performed following

the procedure outlined by Lamour and colleagues.[6] A drawing of the setup is depicted in figure

3.13. In short, a Nikon COOLPIX AW100 digital camera was mounted on the optics table just in

front of a f = 50mm lens. One focal length behind the lens a podium on a delay stage was set

up that can support and translate the sample. In order to provide good contrast a light bulb and a

diffuser was mounted behind the sample. A simple card board enclosure blocks the ambient light.

The pictures were transfered to a lab computer where the contact-angles were measured with the

ImageJ software [42] in combination with the contact angle plugin[43].

3.7.3 Removing the OTS SAM

To remove the OTS SAM, the prism is pre-cleaned, like previously described, with acetone, methanol

and deionized water. The removal itself is done by 10min of slow O2 plasma etching with a pressure
of 500mTorr O2 at 100W, which leaves extremely hydrophobic fused silica behind.

50

Figure 3.14: Apparatus equipped with a Whatman Carbon Cap 75 to filter salt solutions.

3.7.4 Preparation of salt solutions

Hua et al. have investigated the influence of contaminated salt solutions on SFG spectra.[44] Their

recommendation with regard to eliminating organic contaminants can be summarized as follows.

Metal-halide salts with high melting temperatures should be baked, then filtered. However, just

baking and just filtering individually also yield SFG spectra indistinguishable from those obtained

by baking and then filtering. Molecular ion based salts should only be baked if their melting point is
high enough (> 500◦C) and then filtered. Hua et al. do not recommend baking only for molecular
ion based salts with high melting temperatures and suggest to filter whether the melting temperature

allows for baking or not.

51

3.7.4.1 Sodium chloride

Based on the research of Hua et al., sodium chloride (NaCl, Columbus Chemical Industries, ACS
grade, ≥ 99.0) was baked for about 12h at ≈ 650◦C (melting temperature Tm = 801◦C), in a
Thermolyne 48000 furnace. A 3M solution was then prepared by dissolving 0.3moles of NaCl
(uNaI = 58.44 g
mol), equivalent to 17.5g in 100mL filtered water from a Millipore Milli-Q water
purification system. The solution was then filtered four times using a thoroughly rinsed Whatman

Carbon Cap 75 filter with activated carbon from GE Healthcare Life Sciences shown in figure 3.14.

3.7.4.2 Sodium iodide

Similarly to sodium chloride, the sodium iodide (NaI, Sigma-Aldrich, ACS reagent, ≥ 99.5%) was
baked for about 10h at ≈ 600◦C, approximately 60◦C below its melting temperature (Tm = 661◦C),
in a Thermolyne 48000 furnace. A 3M solution was then prepared by dissolving 0.3moles of
NaI (uNaI = 149.89 g
mol), equivalent to 45g, in 100mL filtered water from a Millipore Milli-Q
water purification system. To prevent the oxidation reaction of the iodide ions, 50mM of sodium
thiosulfate pentahydrate (Na2S2O2·5H2O, u = 248.18 g
mol, Sigma-Aldrich, ACS reagent, 99.5%)
was used by adding 1.24g to the previous 3M NaI solution. Furthermore, the solution was then

filtered four times using the filter described in the previous section 3.7.4.1.

3.7.4.3 Magnesium sulfate

The magnesium sulfate (MgSO4) solution was prepared with magnesium sulfate heptahydrate
mol, Sigma-Aldrich, ACS reagent, ≥ 99.0%) which starts to decompose
(MgSO4·7H2O, u = 246.47 g
at ≈ 150◦C. Therefore, it was not baked. A three molar solution was prepared by dissolving 73.9g
of magnesium sulfate heptahydrate in 100mL from the Millipore Milli-Q. Subsequently, the solution

was filtered four times with the filter setup described in the previous section 3.7.4.1.

52

CHAPTER 4

ORIENTATIONAL DYNAMICS OF WATER AT AN EXTENDED HYDROPHOBIC

INTERFACE

To investigate the orientational dynamics of water at an extended hydrophobic interface, we use

the setup described in detail in Chapter 3. Specifically, figure 3.4 gives a good overview of the

setup. In short, a fused silica prism is coated with an OTS self-assembled monolayer in order to

produce an extended hydrophobic interface. VIS, pump and probe beams are overlapped at the

water/OTS interface. The IR pulses of pump and probe are resonant with the dOH stretch mode

(figure 4.1). By detecting the SFG signal of the probe and VIS from the water/OTS interface, the

pump-induced bleach versus the pump-probe delay is measured. Moreover, to obtain further insight

into the contribution of non-orientational mechanisms, isotopic dilution is used.

Figure 4.1: Red: IR spectrum with a ≈ 45cm−1 full-width half-maximum determined from the
probe SFG spectrum from a Ag mirror. Black: SFG spectrum of the dOH stretch at the
fused-silica/ODS/H2O interface.

53

4.1 Neat water

The normalized SFG probe intensity traces shown in figure 4.2 are fit with a biexponential

decay:

Inorm(t) ≡

I(t)

I(t < −1.2ps)

Equivalently the bleach is fit by:

(cid:19)

(cid:18)
−t − 0.6ps
(cid:19)

= 1− A1 exp
(cid:18)
−t − 0.6ps

τ1

(cid:18)
−t − 0.6ps
(cid:19)

− A2 exp
(cid:18)
−t − 0.6ps

τ2

(cid:19)

.

(4.1)

(4.2)

τ2

β (t) = 1− Inorm(t) = A1 exp

+ A2 exp

τ1

Throughout this analysis we only fit to data points with a pump-probe delay of t ≥ 0.6ps, by which
time the pump pulse has almost completely passed. For p-polarized pump, the result τ1 = 1.34±
0.03ps for the fast decay time is consistent with earlier measurements at the water/OTS interface[27]

but is substantially slower than the result (0.84ps) for dOH at the water/air interface.[17] Table

4.1 provides an overview of the time constants obtained by the fitting procedure. Since at long

delays the orientational anisotropy has decayed, p- and s-pump signals should be governed by the

same slow dynamics associated with transient heating which is described in more detail below.
Hence, τ2, the time constant describing the slower dynamics is used globally to fit the data at long
delays. The same fit can be done with isotopically diluted water. Here, the same slow dynamics
as in the neat case is observed. However, the short dynamics are characterized by a 1.13± 0.04ps
decay constant for p-pump and a 1.38± 0.08ps constant for s-pump. To further analyze the data we
have to take into account that at long delays the bleach is proportional to the pump pulse energy

absorbed at the surface, which depends on pump polarization and dOH polar distribution. The

incomplete recovery on the long (50ps) time scale can be explained by the aforementioned small,

p-pump
1.34± 0.03

s-pump
1.55± 0.04

H2O

15± 2

1:2 H2O:D2O

p-pump
1.13± 0.04

s-pump
1.38± 0.08

15± 2

[ps]
τ1
τ2

Table 4.1: Fast and slow time constants extracted from a fit to the data shown in figure 4.2

54

Figure 4.2: Time-resolved sum-frequency signal of the dOH stretch at the ODS/water interface
normalized to the SFG signal at t < −1.2ps for pure H2O and H2O:D2O. The blue shading
highlights the difference between p- and scaled s-pump data. The black curve is the third-order
cross-correlation at ω = ωVIS + ωpump + ωprobe recorded at the same configuration as the
measurements to indicate t = 0 and the temporal resolution of our instrument. The bright green
lines show biexponential fits to the data for t ≥ 0.6ps.

transient rise in temperature. This leads to a weakening of hydrogen bonds close to the interface,

causing a blueshift of the bOH spectrum. Phase resolved spectra by Tian et al. reveal the imaginary
part of χ(2) associated with the blue part of the bOH spectrum to be of opposite sign to the dOH
contribution to χ(2)[2]. The blue shift of the bOH spectrum is, therefore, expected to reduce the

magnitude of the SFG intensity from the dOH region with respect to the un-pumped sample and

therefore prevents a complete recovery to a pre-pump value during the time scale depicted. With

this knowledge we proceed by scaling the s-pump trace to match the p-pump trace at long delays:

s,scaled(t) = 1− βs,scaled(t)
Inorm

(cid:10)βp(t > 15ps)(cid:11)

(cid:104)βs(t > 15ps)(cid:105) βs(t).

= 1−

55

(4.3)

(4.4)

Indeed, as one can see in figure 4.2 the dynamics of the p- and the scaled s-pump data at long delays

are identical, while displaying significant differences at short delays.

In some samples, anomalies were observed: The SFG spectrum would exhibit a small second feature
blueshifted by 10−20cm−1 from the main feature at 3680cm−1. In these cases, a rapidly decaying
beating of the signal characterized by a frequency of ≈ 20cm−1 and a decay constant ≈ 1ps was
visible in the time-resolved traces. This phenomenon has been described by Ye et al.[45] and in the

work of Eftekhari-Bafrooei[46]. It is caused by water molecules that are trapped between the OTS

monolayer and the fused silica. These water molecules are essentially decoupled from the water at

the water/OTS interface or the bulk. To minimize the contribution from the trapped water molecules

we focus on those samples that did not display any of the signs just mentioned.

We focus on the fast dynamics by calculating the difference between the p- and scaled s-pump

bleach traces:

∆β (t) = βp(t)− βs,scaled(t)

and the average of the p- and scaled s-pump bleach traces:

(cid:2)βp(t) + βs,scaled(t)(cid:3) .

¯β (t) =

1
2

(4.5)

(4.6)

Figure 4.3 shows ∆β (t) and ¯β(cid:48)(t), which is the average bleach ¯β (t) minus its long-delay value:

¯β(cid:48)(t) = ¯β (t)− ¯β (t > 25ps).

(4.7)

According to equation 2.73, ∆β (t) is fit with a single exponential decay starting at t0 = 0.6ps, just
after the pump pulse has passed:

(4.8)

(4.9)

(cid:33)

(cid:32)
−t −t0
τ∆β

∆β (t;t0) = Aexp

.

¯β(cid:48)(t) is fit according to equation 2.74 with a double exponential decay

¯β(cid:48)(t;t0) = B1 exp

(cid:19)

(cid:18)
−t −t0
τ2

.

+ B2 exp

(cid:19)

(cid:18)
−t −t0
T1

56

¯β(cid:48)(t;t0 = 0.6ps)
¯β(cid:48)(t;t0 = 2.0ps)
T1 = 1.42± 0.02ps T1 = 1.46± 0.06ps

τβ slow = 15ps *

∆β (t;t0 = 0.6ps)
τ∆β = 0.74± 0.05ps

Table 4.2: Time constants extracted from fits to ∆β and ¯β(cid:48) shown in figure 4.3. * = fixed parameter.

The fit is performed twice, starting at t0 = 0.6ps and at t0 = 2.0ps and the time constants from
the fits are listed in table 4.2. The compatibility of the fast time constants obtained by the two
biexponential fits of ¯β(cid:48)(t) starting at 0.6ps and 2.0ps indicates that the anisotropy underlying
∆β (t) does not contribute significantly to ¯β(cid:48)(t). This is consistent with the earlier discussion of
equation 2.74, where it was established that ¯β(cid:48)(t) is only expected to have a weak dependence
on orientational dynamics. To rule out small contributions of orientational dynamics, the rate
= (1.46 ± 0.06ps)−1 from the fit starting at 2.0ps is used for further analysis. To
kdOH = 1
T1
calculate the rate kor associated with the pump-induced orientational anisotropy, equation 2.73 is

used:

The uncertainty in kor can be calculated using

δ k∆β
k∆β

=

δτ∆β
τ∆β

and δ kdOH
kdOH

=

δ T1
T1

:

kor = k∆β − kdOH

= (0.74ps)−1 − (1.46ps)−1
= 0.67ps−1 = (1.50ps)−1

δ kor =

=

δ k2

∆β + δ k2

(cid:113)
(cid:118)(cid:117)(cid:117)(cid:116)(cid:32)δτ∆β
(cid:115)(cid:18)0.05ps

τ∆β

dOH

(cid:33)2

k∆β

+

× 1

0.74ps

=
0.74ps
= 0.096ps−1

(4.10)

(4.11)

(4.12)

(4.13)

(4.14)

(4.15)

(4.16)

(cid:18)δ T1
(cid:19)2

T1

(cid:19)2
(cid:18)0.06ps

kdOH

+

1.46ps

(cid:19)2

× 1

1.46ps

57

The uncertainty of τor = 1

kor = 1.50ps is

δτor =

=

τor

δ kor
kor
0.096ps−1
0.67ps−1 ×

= 0.22ps

1

0.67ps−1

To conclude:

kor = 0.67± 0.10ps−1 = (1.50± 0.22ps)−1

(4.17)

(4.18)

(4.19)

(4.20)

Notably, kor is within error the same as kdOH. This is not a coincidence. Rather, since the
experimental setup is mainly probing χ(2)
zzz , it is sensitive to molecular orientational motions that
involve the change in the polar angle θ. Such motion leads to the decay of the pump-induced

anisotropy, but it also contributes to population decay by reconfiguring the dOH to a bOH. Keeping

in mind that the stretch frequency of a bOH red shifts out of the detection region, this reorientation

process appears as part of the population decay.

At this point, it is possible that kor is not purely determined by molecular reorientation. Other

processes, like dipole-dipole coupling, intra-molecular energy transfer or intra-molecular vibrational

relaxation, described in section 2.4, might alter the angular distribution of excited and ground-state

dOH stretch populations without changing the actual orientation of the dOH groups that are involved.

Sufficient isotopic dilution of the H2O sample reduces the latter pathways and helps establish a
more reliable rate for the molecular reorientational motion.

58

Figure 4.3: dOH stretch SFG dynamics at the ODS/H2O and ODS/H2O:D2O interfaces showing
the decay of the average bleach ¯β(cid:48)(t) (open blue circles) relative to the long-decay bleach and the
difference ∆β (t) (solid red circles) between the p-pump and scaled s-pump data. Solid curves show
biexponential fits (cyan) for t ≥ 2.0ps and single-exponential fits (orange) for t ≥ 0.6ps. The
third-order cross correlation is shown in solid black.

4.2 Isotopically Diluted Water

As elaborated in section 2.5, a sample with isotopically diluted water can give further insight into

how much non-orientational mechanisms, such as IET, IVR or dipole-dipole interactions, contribute

to the decay of pump-induced anisotropy. If they indeed play a significant role it is expected that

isotopic dilution will slow the observed dynamics due to the reduction of their respective rates.

Here, we use a molar ratio of 1:2 of H2O:D2O, which after equilibration leads to the ratio 1:4:4
of H2O:HDO:D2O. This means that there are two species of dOH: Those associated with an H2O
molecule and those associated with an HDO molecule. Under the assumption that intermolecular

dynamics remain unchanged for the H2O molecules in the isotopic mixture, it is assumed that 1/3
of the signal decays following the neat water dynamics while the remaining 2/3 reflects the new

59

dynamics of the HDO molecules. In reality intermolecular dynamics of H2O molecules are slowed
by isotopic dilution. Therefore, any actual slowing of H2O dynamics will manifest in the HDO time
constant and yield a slower-than-actual time constant for the HDO dynamics. The fit of ¯β(cid:48)(t) is
performed with a triple exponential decay consisting of the double exponential decay of neat H2O
(cid:19)
plus one additional exponential decay with a fast time constant to capture the HDO dynamics
¯β(cid:48)HDO(t) = B3

(cid:18)
−t − 2.0ps

(cid:18)
−t − 2.0ps

(cid:33)(cid:35)

+ B4 exp

× exp

× exp

(cid:19)

(cid:34)

+ y0

+

(cid:32)
−t − 2.0ps
τHDO
dOH

1.46ps

15ps

1
3

2
3

and yields

kHDO
dOH =

1

τHDO
dOH

= (1.41± 0.21ps)−1.

(4.21)

(4.22)

This is statistically indistinguishable from the population decay in neat H2O determined in the
previous section kdOH = (1.46 ± 0.06ps)−1. This indicates that population dynamics are not
significantly affected by isotopic dilution. With the further assumption that orientational motion
is identical for H2O and D2O, one can fit ∆β HDO(t) obtained from the isotopic dilution with a
∆β = (1.02± 0.08ps)−1. Analogous to the neat water
single-exponential decay, which yields kHDO

case we subtract the population decay to obtain the reorientation time

kHDO
or = kHDO

∆β − kHDO

dOH

(cid:16)

= (1.02ps)−1 − (1.41ps)−1
= 0.27ps−1 = (3.7ps)−1
(cid:114)(cid:16)
(cid:17)2
(cid:118)(cid:117)(cid:117)(cid:117)(cid:116)δτHDO
(cid:115)(cid:18)0.08ps

(cid:17)2
(cid:32)δ T HDO
(cid:18)0.21ps

2
(cid:19)2

∆β
τHDO
∆β

δ kHDO
dOH

T HDO
1

kHDO
∆β

δ kHDO

∆β

+

+

(cid:33)2

1

kHDO
dOH

(4.23)

(4.24)

(4.25)

(4.26)

(4.27)

(4.28)

(4.29)

(cid:19)2

and its uncertainty

δ kHDO

or =

=

× 1

1.02ps

+

1.41ps

× 1

1.41ps

=
1.02ps
= 0.13ps−1

60

or

δτHDO

or =

=

δ kHDO
or
τHDO
or
kHDO
or
0.13ps−1
0.27ps−1 ×

1

0.27ps−1

= 1.8ps.

(4.30)

(4.31)

(4.32)

It seems from kHDO

or = (3.7± 1.8ps)−1 that the reorientation has slowed upon isotopic dilution.
However, the reduced signal-to-noise ratio that comes hand in hand with isotopic dilution, results

in a less accurate scaling of the s-pump data than in the case of neat water. In addition, k∆β
is determined from a difference of the p- and scaled s-pump traces and so is more susceptible

to inaccuracies than the average kdOH. This leads to the large uncertainty that allows for any
reasonable value within two standard deviations.

4.3 Experimental rates and uncertainties

After examining the decay mechanisms of the dOH stretch vibration summarized in table 2.1,

it is apparent that for pure water kb
population decay described by kdOH = 1/T1 of dOH to bOH is governed by kF
orientational dynamics include out-of-plane (kF

IET ,kor. Therefore, the effectively irreversible
IET and kor. The

jump) and in-plane (kip) reorientational dynamics

IV R,kb→b

p−p (cid:29) kF

kor = kF

jump + kip,

(4.33)

whereas, as discussed before in section 2.3.1, of the two orientational processes only the out-of-plane

reorientation contributes to population decay which leaves us with the following rate equation:

kdOH = kF

jump + kF

IET .

(4.34)

The average bleach ¯β(cid:48) which is accessed experimentally is representative of population decay, and
hence we can write:

k ¯β(cid:48) ≡ kdOH = kF

jump + kF

IET .

(4.35)

61

In addition, the measurement of the isotopically diluted water, where kF

IET is essentially turned off,

the above equation can be simplified to

¯β(cid:48) ≡ kHDO
kHDO

dOH ≈ kF

jump.

(4.36)

The third experimentally measured quantity is the difference between p- and s-pump bleach ∆β .

There, the contributions are population decay and orientational dynamics. Using equation 4.33 and

4.34 we arrive at

k∆β = kor + kdOH = 2kF

jump + kip + kF

IET .

(4.37)

To summarize, we have three equations (4.35, 4.36, and 4.37) and in principle three unknowns
(kF

IET can be calculated as described in section 2.4.3, its uncertainty

IET ). Although, kF

jump, kip, and kF
is still unknown.

4.3.1 Determination of the uncertainty of kF

IET

An estimation of the uncertainty of kF

IET can be obtained from our measurements by subtracting

equation 4.36 from 4.35:

and

IET ≈ k ¯β(cid:48) − kHDO
kF

1.41ps

1.46ps

¯β(cid:48)
≈ 1
− 1
≈ −0.024ps−1
(cid:18)
(cid:16)

(cid:17)2

σ

(cid:115)
(cid:113)
=
= 0.109ps−1.

k ¯β(cid:48)

(cid:19)2

σ (kF

IET ) =

+ σ

kHDO
¯β(cid:48)

(0.028ps−1)2 + (0.106ps−1)2

(4.38)

(4.39)

(4.40)

(4.41)

(4.42)

(4.43)

IET = −0.024± 0.109ps−1, the conclusion is that the upper limit of kF
From this result, kF
IET is
IET,max = (−0.024 + 0.109)ps−1 = 0.085ps−1. Taking the difference with the value for kF
kF

IET

62

determined in section 2.4.3, 0.077ps−1, the upper uncertainty is now evaluated to be σ +(cid:0)kF

IET

(cid:1) =

0.007ps). On the lower side, the experimentally obtained uncertainty would encompass unphysical
values. It is therefore reasonable to limit the uncertainty to the physical range: kF

IET,min = 0. To

conclude:

kF
IET = 0.077+0.007

−0.077 ps−1 =

(cid:16)

13+∞−1.24 ps

(cid:17)−1

(4.44)

(cid:12)(cid:12)(cid:12) γ

δ

(cid:12)(cid:12)(cid:12)2

Note, that the experimental data does not provide an upper limit for the IET timescale.

4.3.1.1 Additional constrains for kF

IET based on the research literature

It is possible to determine, not just the value of kF
uncertainty based on the literature. The way kF

IET as outlined earlier in section 2.4.3, but also the
IET was calculated involves the coupling between the
bOH and dOH stretch modes, the frequency difference between the uncoupled modes, and the rate

describing the intramolecular vibrational relaxation:

kF
IET =

kb
IV R.

(4.45)

In order to calculate a reasonable uncertainty for kF

IET , the uncertainties of γ, δ , and kb

IV R are

needed.

Uncertainty of the coupling γ The uncertainty of the coupling between the bOH and dOH stretch

modes on the same molecule can be inferred from the work of Stiopkin et al.[32] There, they describe
the coupling between a dOD and the bOD on the same D2O molecule to be γD2O = −48cm−1 with
a FWHM of ≈ 12cm−1. This corresponds to a standard deviation of σ (γD2O) = 5.1cm−1. Since
the coupling in the case of H2O is just

γ =

−50
−60

× γD2O,

the uncertainty of γ is simply calculated as follows:

To conclude: γ = −40± 4.25cm−1.

σ (γ) =

−50
−60

× σ (γD2O).

63

(4.46)

(4.47)

Uncertainty of the frequency difference δ

It is known that a bOH group, associated with a

water molecule at the surface that exhibits a dOH has stretch frequency on the blue end of the

broad bOH spectrum that includes other bOH groups. Ni et al. calculate the spectrum of these
particular bOH groups to be at 3480cm−1 with a FWHM of about 239cm−1. The spectrum of dOH
groups at 3680cm−1 is very narrow compared to this. This corresponds to a standard deviation
of σ (δ ) = 101cm−1. As it turns out, the uncertainty term resulting from this is three magnitudes
larger than the other contributions and not likely to be representative of the actual uncertainty of the
frequency difference δ = 157cm−1. In reality, the uncertainty should be significantly smaller. This
discussion continues in paragraph 4.3.1.1.

Uncertainty of kb

IV R The source of the value for kb

IV R specifies the uncertainty σ (T b

IV R) =

0.05ps.[30], therefore

IV R = (0.75± 0.05ps)−1 = 1.33± 0.09ps−1.
kb

(4.48)

Uncertainty of kF
kF
IET can be calculated with the following formula:

IET Having determined the uncertainties of γ, δ , and kb

IV R the uncertainty of

(cid:115)(cid:18)

(cid:19)2

(cid:18)

σ (kF

IET ) =

2

γ
δ 2 kb

IV Rσ (γ)

+

−2

γ2
δ 3 kb

IV Rσ (δ )

+

(cid:19)2

(cid:18) γ2
δ 2 σ (kb

(cid:19)2

IV R)

.

(4.49)

As indicated, the uncertainty of δ is not exactly known. Ignoring it (σ (δ ) = 0) or using the standard

deviation of δ as an upper limit are the extreme ends of this estimate. At this point, the real quantity

of the uncertainty is not know. Rather than ignoring the uncertainty of δ or randomly using a value
in between, this dissertation uses the standard deviation of δ (σδ = 101cm−1) to continue. This
will yield larger uncertainties, however, one can be sure that it will encompass the real values.
Inserting the values obtained in the earlier sections yields and remembering the value of kF

IET

obtained in section 2.4.3:

IET = 0.077± 0.114ps−1 = (13± 19ps)−1 .
kF

(4.50)

64

The uncertainty obtained above includes negative, and therefore unphysical, values of kF
other hand, it provides an upper limit for kF

IET :

IET . On the

IET < 0.190ps−1 = (5.25ps)−1 .
kF

(4.51)

Combining the constraints of kF

IET Unfortunately, with the lack of knowledge about a reasonable
value for the uncertainty of δ , using the research literature is not helpful to further constrain the
experimentally obtained uncertainty of kF
a more reasonable value for σ (kF

IET . The approach laid out, could be helpful to determine
IET ) in the future once the uncertainty of δ has been determined.

Continuing from here, solely the experimentally obtained uncertainties are used:

kF
IET = 0.077+0.007

−0.077 ps−1 =

13+∞−1.24 ps

(cid:16)

(cid:17)−1

4.3.2 Determination of the uncertainty of kF

jump

Finally, the forward out-of-plane jump rate is calculated through equation 4.35:

(4.52)

(4.53)

IET .
IET has asymmetric uncertainties, the uncertainty of kF

jump = k ¯β(cid:48) − kF
kF

Since kF
since in the process of calculating kF
determine the lower limit of kF

jump:

σ±(cid:16)

(cid:17)

(cid:114)(cid:16)

kF
jump

=

σ (k ¯β(cid:48))

The result is:

jump the rates are subtracted, the upper limit of kF

jump will also be asymmetric. However,
IET will

(cid:17)2
(cid:16)

+(cid:0)σ∓(kF

.

IET )(cid:1)2
(cid:17)−1

(4.54)

(4.55)

.

kF
jump = 0.608+0.082

−0.029 ps−1 =

1.64+0.08−0.22 ps

4.3.3 Determination of the in-plane reorientation

Using equation 4.37 derived in the previous section the rate for in-plane reorientational dynamics

can be determined:

kip = k∆β − 2kF

jump − kF

IET .

(4.56)

65

Again, since the quantities kF

jump and kF

IET , their upper uncertainties will contribute to the lower

uncertainty of kip:

(cid:114)(cid:16)

(cid:17)2

(cid:16)

2σ∓(kF

(cid:17)2

+(cid:0)σ∓(kF

IET )(cid:1)2

σ (kip)± =

σ (k∆β )

(4.57)
This results in σ (kip)− = 0.19ps−1 and σ (kip)+ = 0.13ps−1. Since the lower uncertainty includes
negative unphysical values the uncertainty is truncated to match kip,min = 0:

jump)

+

.

kip = 0.06+0.13−0.06 ps−1 =
Similarly to the determination of the uncertainty of kF
upper limit of the in-plane reorientation rate: kip < 0.19ps−1 ≈ (5ps)−1.
A hand-waving argument can be made as to why it is expected that in-plane reorientation is slower

IET , the experiment only yield a meaningful

17+∞−12 ps

(4.58)

.

(cid:16)

(cid:17)−1

than out-of-plane reorientation: In contrast to out-of-plane reorientation, in-plane reorientation

requires the breaking of an hydrogen bond and consequently the formation of a new one. In bulk, the

process of breaking, reorientation of the molecule, and reforming of an hydrogen bond is associated
with a time scale of kbulk = (3.3ps)−1.[10] However, in bulk, this process is the result of either the
rotation about the x, y, or z axis. All three rotations are equally likely in bulk (kφ = kθ ), whereas at
the interface (here the z-axis is fixed to be normal to the interface), due to the broken symmetry, the

situation is different. Imagining that the pump and probe pulses are polarized in z-direction, only

in-plane reorientation, described by rotations about the z-axis, do not appear like reorientation. In

contrast, reorientation about the x- or y-axis look like a forward jump. Therefore, this hand-waving

argument is described by the following rate equation:

kbulk = kx + ky + kz

= 3kip,

(4.59)

(4.60)

where kip is the in-plane reorientation at the interface. This implies that we expect kip = kbulk/3 =
(9.9ps)−1, which is certainly compatible with the result derived from our data. In addition to
this simple argument, Stirnemann et al. arrive at 9.6ps in their MD simulations for in-plane

reorientation.[10]

66

H2O

k [ps−1]

τ [ps]

H2O::D2O

k [ps−1]

τ [ps]

experimental values

¯β(cid:48)
∆β

calculated values

IET
jumpF

ip

0.685± 0.028
1.351± 0.091

1.46± 0.06
0.74± 0.05

0.709± 0.106
0.98± 0.08

1.41± 0.21
1.02± 0.08

0.077+0.007
−0.077
0.608+0.082
−0.029
0.06+0.13−0.06

13.0+∞−1.2
1.64+0.08−0.22
17+∞−12

Table 4.3: Summary table of experimentally obtained and calculated values.

4.4 Summary and interpretation by MD simulations

Using high sensitivity time-resolved pump-probe SFG spectroscopy in TIR geometry, the orien-

tational dynamics for the out-of-plane orientation of dangling OH bonds at the water/hydrophobic
interface occurs on a 1.64+0.08−0.22 ps timescale.
The out-of-plane reorientation time obtained in this dissertation is twice as fast as the orientational

motion of water in bulk and about 50% slower than the out-of-plane orientational motion at the

water/air interface.

Furthermore, the in-plane orientational dynamics have been examined. In this case, the scope of

this dissertation only yields a lower boundary of kip > 5ps. This is compatible with expectations
and values found in the research literature.

4.4.1 MD simulations and interpretation

Our collaborators, Guillaume Stirnemann and Damien Laage, conducted MD simulations compar-

ing the water/air and water/hydrophobic interfaces.[4] They replicated and refined the results of

simulations that have been conducted individually on water/air and water/hydrophobic interfaces.

The key points of their work were the simulation of both interfaces with the same model and to

implement the instantaneous surface, a more reliable description of the water surface. Previously,

67

simulations would consider the flat average Gibbs interface which ignores ripples at the interface

(figure 4.4 a, b). This detail is of importance because the dOH groups are determined by geometric

properties that use the surface as reference system. Modeling the surface more accurately should

therefore lead to more accurate results. For the reorientation time of the dOH they obtain 1.4ps at

the water/air interface and 1.7ps at the water/hydrophobic interface (figure 4.4 c). Independent of

the approach used in this dissertation, the uncertainties do not allow a clear experimental distinction
between the values obtained from the MD simulations. If the uncertainty of kF
using our experimental data, there is a slight bias towards kF

IET is determined
jump being compatible with the MD

result for the water/hydrophobic interface.

The MD result for the water/air interface is slightly slower than the 1.1ps previously reported

experimentally by Hsieh et al. and backed by 0.8ps from MD simulations by Verde et al. [17, 16].

The difference in the MD simulation results are sought to be the result of the improved modeling of

the surface. On the experimental side, the signal-to-noise ratio in the data of Hsieh et al. makes it

difficult to discriminate experimentally between the different simulation results.

By investigating the free energy difference between the two interfaces (figure 4.4 d), Stirnemann

and Laage found that the dOH configuration experiences a stabilization at the water/hydrophobic

interface compared to the water/air interface which corresponds to a slowdown of dOH jump
dynamics by exp(∆FdOH /kBT ) ≈ 1.2 and agrees with the ratio they obtained: 1.7ps/1.4ps = 1.2.

68

Figure 4.4: MD simulations by Stirnemann and Laage: a) MD snapshot of the water/air interface
showing the instantaneous interface as a red mesh. b) Definition of the OH group angle θ and the
distance to the instantaneous interface z. c) Population decay of dOH→bOH for water/air and
water/hydrophobic interfaces. d) Free energy difference ∆FdOH of the water/air and
water/hydrophobic interface.[4]

69

CHAPTER 5

SALT SOLUTIONS

A well documented macroscopic phenomenon is that aqueous salt solutions have a different

viscosity[47] and surface tension than neat water.[48] For a long time the change in viscosity was

assumed to be largely related to the strong effect of the ions on the hydrogen bond network.[49]

Some ions seem to enhance the hydrogen bond network (“structure makers”) while others seem to

disrupt it (“structure breakers”). Since 1934 until the 1990s, it was understood that the change in

surface tension was caused by a surface ion concentration that is lower than in the bulk[50], which

was understood in terms of repulsion of ions from the interface by electrostatic image forces.[51]

5.1 Previous research on bulk salt solutions

Due to the difficulty in distinguishing hydrogen bonds formed by water molecules surround-

ing ions from the hydrogen bonds of bulk water, molecular dynamic simulations have preceded

Figure 5.1: Solvated anions, like Cl−, form hydrogen bonds with surrounding water molecules
(OH···Cl−, orange). These hydrogen bonds can have a different absorption spectrum than bOH in
bulk water (OH···O, blue).

70

...Cl-.....................Figure 5.2: Solvated cations, like Na+, do not form hydrogen bonds with the surrounding water
molecules. Due to the positive charge of the cation, the water molecules orient themselves so that
the partially positive hydrogen atoms face away from the cation. The hydrogen bonds formed by
water molecules in the first solvation shell (OH···O, blue) already exhibit the same frequencies as
bOH in bulk.

Figure 5.3: Tielrooij et al.[7] show that in a semi-rigid hydration shell water molecules are only
locked in one direction (A,B) and the cooperative interaction of cation and anion that locks water
molecules in two directions (C) while involving water molecules beyond the first solvation shell.

71

......Na+experiments in answering questions related to the presence of ions at water surfaces. X-ray and

neutron diffraction have been used to examine structural properties in salt solutions like coordi-

nation numbers and ion-water distances. However, they lack the ability to provide information

about dynamics.[52, 53, 54] Nuclear magnetic resonance is able to measure the reorientation time

of water molecules, but lacks the specificity required to distinguish bulk water hydrogen bonds

from hydrogen bonds in the hydration shell around the ions.[55, 56] The group led by Bakker,

reported experimental results using mid-IR time-resolved pump-probe spectroscopy. In their experi-

ments, they were able to isolate the contributions introduced through ions by temporally[57, 58] or

spectrally[49] resolving the signals. Their main results can be summarized with the help of Figures

5.1 and 5.2. By performing pump-probe spectroscopy at the magic angle, the vibrational lifetime of

the OH stretch vibrations in isotopically diluted (HDO:D2O) salt solutions was found to consist
of two time constants. The fast time constant is similar to what is found in HDO:D2O bulk water
(T1 = 800fs [30]) and is independent of ion-type and ion concentration. It must therefore correspond
to OH groups that form a hydrogen bond with another water molecule OH···O, which includes
water molecules solvating a cation like in Figure 5.2. The second, slow component is dependent on

anion-type and anion concentration, yet independent of cations. This leads to the conclusion that it is
associated with OH groups that solvate the anion OH···Y− like in Figure 5.1. Bakker et al. isolated
the orientational correlation time and found like in their previous experiments two distinct species:

One that is independent of the salt and its concentration, and has an orientational correlation time

that one would expect to find in the bulk (2.5ps - bOD in HDO:H2O) and one with a much slower
orientational correlation time (7.6ps) that belongs to the first solvation shell around the anion.[49]

It appears that the first solvation shell around an anion is “rigid” on a picosecond scale. Modeling

this system with the Stokes-Einstein equation, which relates the diffusion of spherical particles

through a liquid to the liquid’s viscosity, explains the observed increased viscosity of aqueous salt

solutions.[59]

The MD simulations based on the molecular jump mechanism by Laage and Hynes show how

water molecules can depart from a semi-rigid solvation shell.[8] In their research they studied water

72

Figure 5.4: Schematic representation by Laage et al.[8] of the molecular jump mechanism in a
solvation shell around a chloride atom showing the “departure” of a water molecule from the
solvation shell. (Blue sticks: H-bonds involving molecules in the surrounding solution. Green
sticks: H-bonds directly involved in reorientation.) The sequence starts with overcoordination of
the chloride ion and an undercoordinated water molecule in the second solvation shell. This is
followed by a bifurcation and abrupt switch of the hydrogen bond, followed by the departure of the
water molecules from the first solvation shell.

molecules surrounding a Cl− ion, pictured in figure 5.4. Fluctuations in the hydrogen-bond network
preceding the departure of a water molecule from the anion lead to an overcoordinated anion and an

undercoordinated water molecule in the second hydration shell. These initial conditions facilitate an

angular jump motion of the OH bond originally interacting with the anion to the undercoordinated

second shell water molecule; bifurcating briefly like in the case of bulk neat water. Furthermore,

certain ion combinations have an effect on the structure and dynamics of water beyond the first

solvation shell[60, 61, 62]. Bakker’s group found that this is the result of strong cooperativity

of water molecules located between anions and cations[7]. This effect seems to only arise if the

involved anions and cations are strongly hydrated (Figure 5.3-C), for example in the case of MgSO4
resulting from the double charge of the individual solvated ions (Mg2+ and SO2−
4 ). If one of them
is weakly hydrated, the picture of a semi-rigid hydration shell around the strongly hydrated ion

seems to be a more accurate picture.

5.2 Previous research on interfaces of salt solutions

At the water/air interface, research experienced a large step forward in recent times. Macro-

scopically measuring and modeling the surface tension was a cornerstone of investigating the

composition of interfaces for a long time. Using the “Gibbs-Duhem” or “Gibbs absorption” equa-

tion in combination with surface tension measurements, it was thought for most of the 20th century

73

Figure 5.5: MD simulations from Jungwirth and Tobias show number densities of water, oxygen
atoms, and ions, normalized by the bulk water density ρb, for several 1.2M salt solutions.[9]

that the interfacial region is depleted of inorganic ions. This method involved describing ions as

point particles which can be a good approximation for small, non-polarizable ions such as fluoride.

However, these simplifications only hold for a very limited range of solutes and do not reflect the

microscopic polarizability of ions or water molecules. Wilson et al. and Benjamin pioneered the

molecular dynamics simulation of single anions in a water slab of several hundred water molecules,

approximating the water/air interface. Their results, after 21 days of CPU time, pictured ions to

be more energetically favored in the bulk which would support the hypothesis of a surface devoid

of ions.[63] Nonetheless, the researchers realized that their simulations “do not provide a faithful

74

Figure 5.6: Schematic illustration of dangling OH bonds at the water/air interface showing dOH
groups within the first (illustrated by the dashed circle) and second solvation shell of a iodine ion.

description of ions at the water interface” but should only be helpful to create better, more physical

models.[64] Only recently, more accurate simulations and better spectroscopic methods have chal-

lenged the view that the surface is devoid of ions by collecting evidence that some inorganic ions

exhibit a surface propensity.
In the early 1990s, photoelectron spectroscopy measurements by Markovich et al.[65] of I−· (CO2)n
(n = 1, . . . ,7) clusters attracted the attention of researchers and motivated a series of molecular

dynamics studies of clusters. Experimental limitations still prevented a thorough investigation

of water clusters, but a few years later Markovich et al. expanded their initial study to water
clusters with Cl−, Br−, and I−.[66] At the time, molecular dynamics simulations were limited
to studying clusters involving single anions and initially few, later several hundreds of, water

molecules.[67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78] Only molecular dynamic simulations that

used polarizable force fields were able to reproduce the electron binding energies experimentally

obtained with photoelectron spectroscopy. These simulations contradicted the initial hypothesis of

a fully solvated ion and showed the larger anions with a stable coordination number of 3-5 water

molecules at the surface of water clusters even with hundreds of water molecules. They established

75

I-I-the consensus for ions in water clusters: Hard, non-polarizable ions such as F− or Na+ that can be
approximated as point charges are, within the classic Onsager-Samaras model[51], repelled from the

surface by image charges. The comparison of polarizable and non-polarizable force fields suggests

that the polarizability of ions is the driving force for the surface propensity of polarizable anions
like Cl−, Br−, and I−. In addition, the increasing polarizability of Cl−, Br−, and I− manifests
in an increasing surface propensity. The first indirect experimental evidence that bromide has a

larger surface propensity than chloride was obtained by exposing a NaCl crystal doped with NaBr

to water vapor. After redrying the surface was examined with X-ray photoelectron spectroscopy and
scanning electron microscopy revealing substantial surface enrichment of Br−.[79] MD simulations
in aqueous solutions supported this conclusion.[80]

The question if this consensus would hold for larger clusters or for extended interfaces is unavoid-

able. Rather surprisingly, experimental evidence for the surface propensity of anions at extended

interfaces arose by investigation of the phenomenon of ozone depletion at polar sunrise: A rapid

drop in ozone levels and rise of chlorine gas levels is observed in the Arctic (Antarctic) when the

sun rises for the first time at the spring (fall) equinox after months of darkness. Kinetic models of

this phenomenon only considering bulk-phase reactions could not reproduce the recorded on-site

measurements. They could only be reproduced satisfactorily by including anions at the water

surface, which enables direct reactions at the interface. Since the surface propensity emerges upon

including polarizability of ions and water molecules during simulations of clusters up to quasi-bulk

dimensions, this suggests that bulk water might be similar to water clusters in terms of ion surface

propensity.[81, 82]
Refined MD simulations paint an increasingly more detailed picture: Br− and I− anion concen-
trations at the interface are found to exceed their average bulk concentrations. For F−, which is
repelled from the surface, surface OH bonds prefer to point towards the F− ions in the bulk. The
opposite is true for NaBr and NaI solutions, where OH groups point to towards the ions which

are enriched at the surface; the OH groups are therefore parallel to the surface. In the case of a

NaCl solution, intermediate angles are observed.[9] If an aqueous solution contains two anions

76

that are attracted to the interface, at low bulk ion concentration both anions can be found at the

interface. If the bulk concentration is high enough, the ion that experiences the higher surface

propensity will take over the surface and replace “weaker” competing ion. In the case of a solution
with high concentrations of Cl and Br anions, the Br− will replace almost all Cl anions at the
surface.[80] Another effect that manifests at high bulk salt concentrations is the increased ion pair

formation in the bulk and at the interface. Ion pair formation increases from merely 5% at low

concentration to 75% in the bulk and to 55% at the interface of a saturated NaCl solution.[83] This

mechanism explains the 20% drop in hydration numbers at high salt concentrations compared to

low salt concentrations.

With the establishment of SFG spectroscopy as a practical technique for surface studies, experi-

ments could finally move forward. Using 1.7M salt solutions and static SFG, the interfacial region

was probed for spectral changes that indicated structural differences compared to neat water.[84]
Fully coordinated interfacial bOH groups at 3325cm−1 experience a red shift by 40cm−1 due to
engaging in stronger hydrogen bonds with small non-polarizable ions (e.g. F−) while a blue shift of
35cm−1 is observed due to fully coordinated interfacial bOH groups engaging in weaker hydrogen
bonds upon addition of large polarizable ions (Cl−, Br−, I−). These shifts are indicative of the
presence of ions in the vicinity of the surface. The same shifts are well documented for bOH in bulk
solutions[85, 86, 87]. At the same salt concentration (1.7M), only I−, the halide with the largest
surface propensity, affects tangential OH groups associated with water molecules with a dangling
OH which are directly at the interface. The blue shift from 3450cm−1 by up to 50cm−1 of the
feature associated with the tangential OH is experimental evidence that iodide anions are directly at

the interface. At higher salt concentrations, closer to saturation, other studies have observed a blue

shift for all the halides exhibiting a surface propensity.[86, 58, 88, 89] A similar SFG study used

higher salt concentrations and was able to detect signs of a weakened hydrogen bond network in the

interfacial region for NaBr and NaI solutions. No significant difference was detected while using

NaF and NaCl solutions. These differences did not manifest themselves as blue shifts, but rather

as increased intensity of the tangential bOH feature. The increase in intensity at that frequency

77

was attributed to an increase in depth of the interfacial region, and a corresponding increase in the

number of water molecules contributing to the SFG signal.[90]

In 2005, Archontis et al. used MD simulations to answer the question of what exact interplay of

interactions causes an ion to be attracted to the surface.[91] The Coulomb interactions between the
permanent charges of I− ions and the permanent dipole moment of H2O would favor iodide ions in
the bulk. However, the interaction of induced charges stabilizes the iodide ions at the interface. The

most relevant contribution to this surface stabilization originates from dipole-dipole interactions

between the induced dipole in the iodide and the permanent dipole of water molecules. The shorter

range of dipole-dipole interactions supports the conclusion that they mainly originate among the

ion and its first solvation shell. Archontis et al. also added to the accumulating research showing

evidence, theoretically[92, 9, 80, 93] and experimentally[90, 94], for the existence of a double layer

at the surface: The high density of ions with strong surface propensity causes ions of the opposite

charge to accumulate just below them.

Further advances were made more recently by using phase-sensitive SFG spectroscopy. The relative

surface propensity follows a Hofmeister series from the ion expressing the most surface propensity to
the one expressing the least surface propensity: I−, NO−
4 .[95, 96]
In bulk, introducing salt into neat water leads to structural changes of the hydrogen bond network

4 , Cl−, K+, Na+, and SO2−

3 , NH+

caused by solvation shells or ion cooperativity. These interruptions also bring changes in dynamics.

Since structural changes are also documented at the interface it is a reasonable assumption that the

dynamics might differ from the neat water interface as well. To investigate this hypothesis we used

three different salts: NaCl, NaI, and MgSO4. NaCl has been identified as a salt introducing little
change in structure and dynamics compared to other salts. Therefore, an experiment using NaCl is
expected to yield results that are closest to neat water. With the strong surface propensity of I−, NaI
should be an ideal candidate to observe changes of dynamics at the interface that might be caused

by the formation of solvation shells and the presence of iodide anions directly at the surface. To

investigate the influence of cooperativity effects that are reaching further than the first solvation

shell on the surface dynamics an aqueous solution with MgSO4 is used.

78

τfast

τslow
¯β(cid:48)

∆β

[ps]

p-pump
s-pump
shared

H2O

1.34± 0.03
1.55± 0.04

15± 2

3M NaCl
1.41± 0.01
1.83± 0.03

31± 3

3M NaI
1.10± 0.04
1.76± 0.08

11± 1

3M MgSO4
1.35± 0.03
1.56± 0.03

20± 1

T1(t0 = 2.0) 1.46± 0.06
T1(t0 = 0.6) 1.42± 0.02

1.43± 0.03
1.41± 0.01

1.60± 0.08
1.32± 0.03

1.54± 0.07
1.45± 0.02

τ ¯β slow
τ∆β

τ∆β slow

15*

31*

11*

20*

0.74± 0.05

0.84± 0.03

0.56± 0.05

0.60± 0.08

-

-

11*

-

Table 5.1: Summary of all time constants. * = fixed parameter.

5.3 Rotational dynamics of aqueous salt solutions at an extended hydropho-

bic interface

For the experiments using salt solutions the same procedure and analysis, as in the neat water

case, were used. The calculation of rates/decay times is analogous to the neat water case and is not

explicitly repeated in this section. The results of the measurements and fits are summarized in table

5.1. A more detailed discussion follows in the next section.

5.3.1

3M NaCl solution

NaCl is the salt where the least difference on rotational dynamics at the interface is expected due

to it being the one having the least surface propensity. Like in the neat water case, a global fit

with a double exponential (equation 4.1) is performed on the normalized p- and s-pump intensity

traces with individual fast decay constants and a global slow decay constant. Significant deviations

from the neat water case are observed in both fast time constants and the global slow constant and

are listed in table 5.1. Yet, when continuing this analysis like in the neat water case the results
(table 5.1) are similar: The fast time constants obtained from fitting ¯β(cid:48) starting at 0.7ps and 2.1ps
are compatible with each other and are compatible with their equivalents in the neat water case
which indicates that there is little leaking of orientational dynamics into ¯β(cid:48). Proceeding with the

79

Figure 5.7: NaCl normalized SFG intensities

analysis like in the neat water case, the single exponential fit of ∆β yields an observed time constant
of 0.84± 0.03ps, which is slightly slower than in neat water (0.74± 0.05ps) and, although not
compatible within one standard deviation, is well within two standard deviations of the result for
neat water. The final forward jump time for the dOH groups of a 3M NaCl solution is 1.61+0.04−0.20 ps,
not significantly different from that of neat water. For the in-plane orientational time scale, the

limits of our experiment become apparent: Presumably, limitations caused by signal-to-noise and

the double-differential measurement lead to an unphysical value for kip in the case of NaCl.

5.3.2

3M NaI solution

Iodide has been shown to have one of the strongest surface propensities among common ions and

should be a good candidate for producing solvation shells at the surface. The 3M NaI solution is, in

fact, the case where the largest deviation from neat water is measured. Figure 5.9 show the fit of the

p- and s-pump traces with a double exponential and shared slow time constant (table 5.1). It reveals

80

Figure 5.8: NaCl difference and average of bleach

fit parameters

experimental values

¯β(cid:48)
∆β

IET (H2O)

calculated values

jumpF

ip

3M NaCl

k [ps−1]

τ [ps]

0.699± 0.015
1.190± 0.043
0.077+0.007
−0.077

1.43± 0.03
0.84± 0.03
13.0+∞−1.2

0.622+0.078
−0.016
−0.13+0.09−0.16

1.61+0.04−0.20
−7.6+9.45−5.45

Table 5.2: Summary table of experimentally obtained and calculated values for NaCl.

81

Figure 5.9: NaI normalized SFG intensities

differences compared to neat water. The fast time constant of the p-pump trace (1.10± 0.04ps) is
significantly faster than the one measured for neat water (1.34± 0.03ps). However, the fast time
constant of the s-pump trace (1.76± 0.08ps) is compatible with neat water (1.55± 0.04ps) within
two standard deviations. Furthermore, the slow time constant (11.1± 1.4ps) is 25% faster than but
still within two standard deviations of the neat water case (15± 2ps). The fitting of ¯β(cid:48) in figure 5.10
yields time constants listed in table 5.1 that are compatible with neat water to within two standard
deviations. More importantly, the two fast time constants from starting the fit of ¯β(cid:48) at t0 = 0.6ps
and t0 = 2.0ps are only compatible with each other within three standard deviations, indicating
that a significant amount of orientational dynamics might have “leaked” into ¯β(cid:48). Looking at ∆β it
also becomes apparent that there is “leaking” of the slow time constant from ¯β(cid:48). A fit with a single
exponential is therefore unreasonable. Instead, a fit with a double exponential including the slow
time constant from ¯β(cid:48) produces the fastest time constant (0.56± 0.05ps) of the salts used, but is
still within two standard deviations of neat water (0.74± 0.05ps). An acceleration of dynamics by

82

Figure 5.10: NaI difference and average of bleach

anions is counterintuitive because in bulk a slow down of dynamics is observed. The dangling OH
forward jump time is, at 1.82+0.11−0.28 ps, still compatible with the neat water case.
Because of the leakage of dynamics among ¯β(cid:48) and ∆β , it was reasonable to assume that ¯β(cid:48) might
follow a triple exponential decay instead of just a double exponential. The new third exponential

would correspond to the short dynamics from k∆β that leaked in. Nonetheless, the fit came back
with a negligible amplitude associated with k∆β and the idea of using a triple exponential was
discarded. The in-plane orientational time scale is with 1.63+0.61−0.50 ps shorter than expected. It is
important to note though, that the reliability of this result might be affected by the leaking of time

constants as discussed above.

5.3.3 MgSO4 solution

MgSO4 is expected to display a cooperativity effect between the ions which, in bulk, leads to a slow
down of reorientational dynamics affecting water molecules beyond the first solvation shell: The

83

experimental values

¯β(cid:48)
∆β

IET (H2O)

calculated values

jumpF

ip

3M NaI

k [ps−1]

τ [ps]

0.625± 0.031
1.79± 0.16
0.077+0.007
−0.077

1.60± 0.08
0.56± 0.05
13.0+∞−1.2

0.548+0.083
−0.032
0.61+0.19−0.23

1.82+0.11−0.28
1.63+0.61−0.50

Table 5.3: Summary table of experimentally obtained and calculated values for NaI.

Figure 5.11: MgSO4 normalized SFG intensities

84

Figure 5.12: MgSO4 difference and average of bleach

hydrogen-bond network between the ions is locked in multiple directions. However, the surface
propensities of SO2−
and Mg2+ are expected to be lower compared to the previous salts. Figure 5.11
4
shows the p- and s-pump traces that look qualitatively very similar to the neat water case. Double

exponential fits with a common slow time constant yield fast time constants that are compatible
with the neat water fit (table 5.1). The slow time constant obtained for MgSO4 (20± 1ps) is only
compatible with neat water (15± 2ps) within two standard deviations. Examining the average
of the p- and s-pump bleach ¯β(cid:48) with a double exponential fit using again the slow constant from
before, the fast time constants agree again with the neat water experiment. Finally, the fit of ∆β
with a single exponential yields a time constant (0.60± 0.08ps) that is just barely incompatible
within one standard deviation compared to neat water (0.74± 0.05ps), but well compatible within
two. The dangling OH forward jump time is, at 1.75+0.09−0.25 ps, compatible with the neat water
case. Interestingly, the in-plane orientational time scale is with 2.2+1.4−1.2 ps compatible with the
bulk orientational time scale of 3.3ps.[10]. Even though the result is lower than expected, it is not

in a completely unphysical range. Unfortunately, the outcome for the in-plane orientational time

85

3M MgSO4

k [ps−1]

τ [ps]

experimental values

¯β(cid:48)
∆β

IET (H2O)

calculated values

jumpF

ip

0.649± 0.030
1.67± 0.22
0.077+0.007
−0.077

1.54± 0.07
0.60± 0.08
13.0+∞−1.2

0.572+0.082
−0.030
0.44+0.24−0.28

1.75+0.09−0.25
2.2+1.4−1.2

Table 5.4: Summary table of experimentally obtained and calculated values for MgSO4.

scales in the previous two cases involving salt solutions indicates that this result might also not be a

reliable quantity.

5.4 Summary and discussion

Within these experiments, we found no significant evidence for an effect of salts on the orien-

tational dynamics of the dOH. Table 5.5 provides a summary of the results. All three tested 3M

salt solutions (NaCl, NaI, MgSO4) yielded forward jump orientation times that are compatible with
neat water.
If these results hold true, this would indicate that anions like I− which have a high surface propensity
and are know to form solvation shells have little influence on the forward jump motion of the dOH.

This could mean that the bulk concentration of NaI was insufficient and caused only few dOH

groups to be in the range of influence of iodide ions at the interface and the experiment was not

3M NaCl

3M NaI

3M MgSO4

-

-

-

[ps]
τIET
τF
jump
τip

H2O
13+∞−1.24
1.64+0.08−0.22
17+∞−12

1.61+0.04−0.20

1.75+0.09−0.25
2.2+1.4−1.2
Table 5.5: Summary of experimentally obtained constants.

1.82+0.11−0.28
1.63+0.61−0.50

N/A

86

sensitive to their change in reorientation, or that very little dOH bonds exist in the vicinity of an

iodide ion. For example, the partially positively charged hydrogen of dOH bonds rotate toward an

anion and could form a hydrogen bond with the ion. Now being a bOH, it would therefore be out of

our detection range.

Furthermore, cooperativity effects beyond the first solvation shell from salts like MgSO4 do not
seem to influence the τF
jump either. In this case, the explanation could be that due to the relatively
low surface propensity of Mg2+ and SO2−
the "locked in" water molecules due to the cooperativity
4
effect might not extend to the interface and would therefore not be detectable.

It is notable that the slow dynamics of the salt solutions deviate from the neat water values, especially

in the case of the 3M NaCl solution. This could be due to several potential issues. First, a systematic

experimental error, such as a slight misalignment of pump path down the delay stage, could cause

effects at long delays that are not noticeable at short delays due to the higher signal-to-noise ratio.

Second, the proximity of ions at or near the surface could lead to formation of hydrogen bonds

between ions and water molecules. Due to the nature of the salts that were used, these ion-water

hydrogen bonds would be weaker than hydrogen bonds between water molecules. The ion-water

hydrogen bond would be blue-shifted and would have less spectral overlap with the surrounding

bOH which would result in a slow down of dynamics that are related to transient heating.

As discussed earlier, the reliability of the in-plane orientational time scale results is questionable.

This might simply be due to the double-differential nature of the approach in this dissertation:

Since pump-probe measurements are differential measurements, calculating the difference of such

measurements to obtain quantities like the in-plane orientational time scale has to be viewed with

caution since signal-to-noise issues are increasingly influential. The results that yield physical

values should be backed up by further simulations or measurements.

87

CHAPTER 6

SUMMARY AND OUTLOOK

6.1 Orientational dynamics of neat water at hydrophobic interfaces

At the interface between neat water and an extended hydrophobic surface, realized by a hy-

drophobic OTS SAM, the out-of-plane orientational dynamics of dangling OH groups was measured
to occur on a 1.64+0.08−0.22 ps timescale. The measurement was performed in a total internal reflection
geometry that yields high signal-to-noise and simplifies the analysis by heavily weighting the χ(2)
zzz
contribution to the second order optical susceptibility. Our result for the out-of-plane orientational

dynamics of dangling OH groups at a hydrophobic interface is twice as fast as the orientational

dynamics of water in bulk and about 50% slower than the out-of-plane orientational dynamics at

the water/air interface. Our interpretation of the pump-probe SFG data was confirmed with MD

simulations by our collaborators Guillaume Stirnemann and Damien Laage. Their MD simulations

suggest that the slower out-of-plane orientational dynamics at the hydrophobic interface is a result

of the dangling OH groups being slightly more stabilized by the hydrophobic medium. This inter-
pretation aligns with the 20cm−1 redshift observed in stativ SFG spectra of the dangling OH peak
at the water/hydrophobic surface compared to water/air.

6.2 Orientational dynamics of salt solutions at hydrophobic interfaces

Upon adding 3M of each of the salts NaCl, NaI, and MgSO4 to neat water, measurements of
the out-of-plane orientational dynamics of the dangling OH groups were repeated. Despite the fact

that these salts have shown surface propensity and evidence in the literature for a slow down in

orientational dynamics in bulk salt solutions, no significant change of the out-of-plane orientational

dynamics of the dangling OH groups was detected. The time scales for out-of-plane reorientation
for NaCl, NaI, and MgSO4 were measured to be 1.61+0.04−0.20 ps, 1.82+0.11−0.28 ps, and 1.75+0.09−0.25 ps,
respectively. In principle, the potential changes could be beyond our ability of detection and might

88

be detectable at higher salt concentrations. If the salt concentration was appropriate, for anions
like I− that display a high surface propensity and form solvation shells at the interface, this would
indicate that a solvation shell at the interface might not have any slowing effects even if water

molecules with dOH are in the vicinity or even part of the first solvation shell. In the case of MgSO4,
the surface propensity of its ions is considerably lower than for iodide. A reasonable explanation for

not being to detect a difference for MgSO4 could be that the cooperativity effects between Mg2+
and SO2−
4 might not extend to the uppermost portion of the interface. The dOH would therefore be
unaffected.

6.3 Outlook

The measurement of the dynamics of the dOH stretch vibration at the interface with a hydropho-

bic medium ensures the hightest degree of surface specificity. By tuning the infrared to the bonded

OH stretch vibration, the dynamics of bonded OH groups in the immediate vicinity of the interface

can be examined. In addition, many water interfaces involve hydrophilic substances, where no

dangling OH groups exist. In that case, a measurement where the pump and probe IR pulses are

resonant with the bonded OH stretch vibration would enable measuring the dynamics directly at the

water/hydrophilic interface. Finally, measurements resonant to the bonded OH stretch vibration

might help with the detection of changes in interfacial dynamics of salt solutions. Since water

molecules participate in solvation shells around ions, and thereby form hydrogen bonds with anions,

being able to specifically probe those hydrogen bonds would be of advantage.

89

APPENDIX

90

A.1 Calculation of error bars in figures

For a given pump-probe delay t, the CCD detects Ni j number of photons in pixel i j: Ni j with an

associated uncertainty σ (Ni j) = (cid:112)Ni j. Summation over a certain pixel range yields the intensity:

s j = ∑i Ni j with an uncertainty of

σ (s j) =

σ (N1 j)2 + (σ (N2 j)2 + . . . =

The pixel range is measured n times and averaged to I = 1

(cid:113)

(cid:113)

In figure 4.2 the normalized quantity Inorm = Ion
Io f f

is plotted where

σ (I) =

1
n

σ (s1)2 + σ (s1)2) + . . . =

1
n

σ (Inorm)
Inorm =

(cid:118)(cid:117)(cid:117)(cid:116)(cid:18)σ (Ion)

Ion

(cid:19)2
(cid:115)

+

(cid:18)σ (Io f f )

(cid:19)2

Io f f

1
Io f f

σ (Inorm) = Inorm × 1√
n

1
Ion

+

(cid:113)
N1 j + N2 j + . . . = (cid:112)s j.
(cid:114) I

j=1 s j with the uncertainty

n ∑n

√
n× I =

√
s1 + s2 + . . . =

1
n

n

=

1

n× Ion

+

1

n× Io f f

+

I2
on
I3
o f f

(cid:115)
(cid:118)(cid:117)(cid:117)(cid:116) Ion

I2
o f f

+

I2
on
I3
o f f

=

1√
n

(cid:118)(cid:117)(cid:117)(cid:116) Ion

I2
o f f

determines the size of the error bars, which are the same size for the bleach β = 1− Inorm:

σ (β ) = σ (Inorm) =

1√
n

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

The error bars of the scaled s-pump trace is simply the original s-pump trace error bars multiplied

by the scaling factor:

σ (Inorm

s,scaled) = σ (βs,scaled) =

βp(t (cid:29) k−1
or )
βs(t (cid:29) k−1
or )

× σ (βs)

Finally, the error bars for ∆β = βp − βs and ¯β = 1

2(βp + βs), plotted in figure 4.3, are:

(cid:113)
σ(cid:0)βp
(cid:113)
σ(cid:0)βp

(cid:1)2
(cid:1)2

+ σ(cid:0)βs,scaled
+ σ(cid:0)βs,scaled

(cid:1)2
(cid:1)2

σ (∆β ) =
1
σ ( ¯β ) =
2

91

BIBLIOGRAPHY

92

BIBLIOGRAPHY

[1] D. Laage and J. T. Hynes. On the molecular mechanism of water reorientation. The Journal of

Physical Chemistry B, 112:14230–14242, 2008.

[2] C. S. Tian and Y. R. Shen. Structure and charging of hydrophobic material/water inter-
faces studied by phase-sensitive sum-frequency vibrational spectroscopy. Proceedings of the
National Academy of Sciences, 106:15148–15153, 2009.

[3] H.-K. Nienhuys and M. Bonn. Measuring molecular reorientation at liquid surfaces with
time-resolved sum-frequency spectroscopy: A theoretical framework. The Journal of Physical
Chemistry B, 113:7564–7573, 2009.

[4] S. Xiao, F. Figge, G. Stirnemann, D. Laage, and J. A. McGuire. Orientational dynamics
of water at an extended hydrophobic interface. Journal of the American Chemical Society,
138:5551–5560, 2016.

[5] Wikipedia. Octadecyltrichlorosilane - wikipedia,

the free encyclopedia.

en.wikipedia.org/wiki/Octadecyltrichlorosilane, 2016.
February-2016].

https://
[Online; accessed 23-

[6] G. Lamour, A. Hamraoui, A. Buvailo, Y. Xing, S. Keuleyan, V. Prakash, A. Eftekhari-Bafrooei,
and E. Borguet. Contact angle measurements using a simplified experimental setup. Journal
of Chemical Education, 87:1403–1407, 2010.

[7] K. J. Tielrooij, N. Garcia-Araez, M. Bonn, and H. J. Bakker. Cooperativity in ion hydration.

Science, 328:1006–1009, 2010.

[8] D. Laage and J. T. Hynes. Reorientional dynamics of water molecules in anionic hydration

shells. Proceedings of the National Academy of Sciences, 104:11167–11172, 2007.

[9] P. Jungwirth and D. J. Tobias. Molecular structure of salt solutions: A new view of the
interface with implications for heterogeneous atmospheric chemistry. The Journal of Physical
Chemistry B, 105:10468–10472, 2001.

[10] G. Stirnemann, P. J. Rossky, J. T. Hynes, and D. Laage. Water reorientation, hydrogen-
bond dynamics and 2D-IR spectroscopy next to an extended hydrophobic surface. Faraday
Discussions, 146:263–281, 2010.

[11] D. Laage and J. T. Hynes. A molecular jump mechanism of water reorientation. Science,

311:832–835, 2006.

[12] H. Graener, G. Seifert, and A. Laubereau. Direct observation of rotational relaxation times by

time-resolved infrared spectroscopy. Chemical Physics Letters, 172:435–439, 1990.

[13] S. Woutersen and H. J. Bakker. Resonant intermolecular transfer of vibrational energy in

liquid water. Nature, 402:507–509, 1999.

93

[14] Q. Du, R. Superfine, E. Freysz, and Y. R. Shen. Vibrational spectroscopy of water at the

vapor/water interface. Physical Review Letters, 70:2313–2316, 1993.

[15] J. D. Eaves, J. J. Loparo, C. J. Fecko, S. T. Roberts, A. Tokmakoff, and P. L. Geissler. Hydrogen
bonds in liquid water are broken only fleetingly. Proceedings of the National Academy of
Sciences, 102:13019–13022, 2005.

[16] A. Vila Verde, P. G. Bolhuis, and R. Kramer Campen. Statics and dynamics of free and
hydrogen-bonded OH groups at the air/water interface. The Journal of Physical Chemistry B,
116:9467–9481, 2012.

[17] C.-S. Hsieh, R. K. Campen, M. Okuno, E. H. G. Backus, Y. Nagata, and M. Bonn. Mechanism
of vibrational energy dissipation of free OH groups at the air-water interface. Proceedings of
the National Academy of Sciences, 110:18780–18785, 2013.

[18] T. Ishiyama and A. Morita. Molecular dynamics study of gas-liquid aqueous sodium halide
interfaces. I. Flexible and polarizable molecular modeling and interfacial properties. The
Journal of Physical Chemistry C, 111:721–737, 2007.

[19] T. Ishiyama and A. Morita. Molecular dynamics study of gas-liquid aqueous sodium halide
interfaces. II. Analysis of vibrational sum frequency generation spectra. The Journal of
Physical Chemistry C, 111:738–748, 2007.

[20] X. Wei, P. B. Miranda, and Y. R. Shen. Surface vibrational spectroscopic study of surface

melting of ice. Physical Review Letters, 86:1554–1557, 2001.

[21] R. Superfine, J. Y. Huang, and Y. R. Shen. Nonlinear optical studies of the pure liquid/vapor
interface: Vibrational spectra and polar ordering. Physical Review Letters, 66:1066–1069,
1991.

[22] X. Wei. Sum-Frequency Spectroscopic Studies I. Surface Melting of Ice II. Surface Alignment

of Polymers. PhD thesis, University of California at Berkeley, 2000.

[23] D. Eisenberg and W. Kauzmann. The Structure and Properties of Water. OXFORD UNIV PR,

1969.

[24] A. J. Lock, S. Woutersen, and H. J. Bakker. Ultrafast energy equilibration in hydrogen-bonded

liquids. The Journal of Physical Chemistry A, 105:1238–1243, 2001.

[25] Z. Gengeliczki, D. E. Rosenfeld, and M. D. Fayer. Theory of interfacial orientational relaxation

spectroscopic observables. The Journal of Chemical Physics, 132:244703, 2010.

[26] S. Ashihara, N. Huse, A. Espagne, E.T.J. Nibbering, and T. Elsaesser. Vibrational couplings
and ultrafast relaxation of the O-H bending mode in liquid H2O. Chemical Physics Letters,
424:66–70, 2006.

[27] J. A. McGuire and Y. R. Shen. Ultrafast vibrational dynamics at water interfaces. Science,

313:1945–1948, 2006.

94

[28] M. Smits, A. Ghosh, M. Sterrer, M. Müller, and M. Bonn. Ultrafast vibrational energy transfer
between surface and bulk water at the air-water interface. Physical Review Letters, 98:098302,
2007.

[29] M. L. Cowan, B. D. Bruner, N. Huse, J. R. Dwyer, B. Chugh, E. T. J. Nibbering, T. Elsaesser,
and R. J. D. Miller. Ultrafast memory loss and energy redistribution in the hydrogen bond
network of liquid H2O. Nature, 434:199–202, 2005.

[30] S. Woutersen, U. Emmerichs, H.-K. Nienhuys, and H. J. Bakker. Anomalous temperature
dependence of vibrational lifetimes in water and ice. Physical Review Letters, 81:1106–1109,
1998.

[31] Z. Zhang, L. Piatkowski, H. J. Bakker, and M. Bonn. Ultrafast vibrational energy transfer at
the water/air interface revealed by two-dimensional surface vibrational spectroscopy. Nature
Chemistry, 3:888–893, 2011.

[32] I. V. Stiopkin, C. Weeraman, P. A. Pieniazek, F. Y. Shalhout, J. L. Skinner, and A. V. Benderskii.
Hydrogen bonding at the water surface revealed by isotopic dilution spectroscopy. Nature,
474:192–195, 2011.

[33] M. Sovago, R. Kramer Campen, H. J. Bakker, and M. Bonn. Hydrogen bonding strength
of interfacial water determined with surface sum-frequency generation. Chemical Physics
Letters, 470:7–12, 2009.

[34] K. B. Eisenthal. Measurement of intermolecular energy transfer using picosecond light pulses.

Chemical Physics Letters, 6:155–157, 1970.

[35] S. A. Corcelli and J. L. Skinner. Infrared and Raman line shapes of dilute HOD in liquid H2O

and D2O from 10 to 90◦C. The Journal of Physical Chemistry A, 109:6154–6165, 2005.

[36] J. R. Schmidt, S. A. Corcelli, and J. L. Skinner. Pronounced non-Condon effects in the ultrafast

infrared spectroscopy of water. The Journal of Chemical Physics, 123:044513, 2005.

[37] J. J. Loparo, S. T. Roberts, R. A. Nicodemus, and A. Tokmakoff. Variation of the transition
dipole moment across the OH stretching band of water. Chemical Physics, 341:218–229,
2007.

[38] R. A. Kaindl, M. Wurm, K. Reimann, P. Hamm, A. M. Weiner, and M. Woerner. Generation,
shaping, and characterization of intense femtosecond pulses tunable from 3 to 20 µm. Journal
of the Optical Society of America B, 17:2086, 2000.

[39] A. Ghosh, M. Smits, J. Bredenbeck, N. Dijkhuizen, and M. Bonn. Femtosecond time-
resolved and two-dimensional vibrational sum frequency spectroscopic instrumentation to
study structural dynamics at interfaces. Review of Scientific Instruments, 79:093907, 2008.

[40] S. Babar and J. H. Weaver. Optical constants of Cu, Ag, and Au revisited. Applied Optics,

54:477, 2015.

[41] J. Sagiv. Organized monolayers by adsorption. 1. Formation and structure of oleophobic mixed
monolayers on solid surfaces. Journal of the American Chemical Society, 102:92–98, 1980.

95

[42] W. Rasband. Imagej. https://imagej.nih.gov/ij/, 1997-2017.

[43] M. Brugnara. Contact angle plugin (for imagej software). https://imagej.nih.gov/ij/

plugins/contact-angle.html, 2006.

[44] W. Hua, D. Verreault, E. M. Adams, Z. Huang, and H. C. Allen. Impact of salt purity on
interfacial water organization revealed by conventional and heterodyne-detected vibrational
sum frequency generation spectroscopy. The Journal of Physical Chemistry C, 117:19577–
19585, 2013.

[45] S. Ye, S. Nihonyanagi, and K. Uosaki. Sum frequency generation (SFG) study of the pH-
dependent water structure on a fused quartz surface modified by an octadecyltrichlorosilane
(OTS) monolayer. Physical Chemistry Chemical Physics, 3:3463–3469, 2001.

[46] A. Eftekhari-Bafrooei, S. Nihonyanagi, and E. Borguet. Spectroscopy and dynamics of the
multiple free OH species at an aqueous/hydrophobic interface. The Journal of Physical
Chemistry C, 116:21734–21741, 2012.

[47] D. E. Goldsack and R. Franchetto. The viscosity of concentrated electrolyte solutions. I.
Concentration dependence at fixed temperature. Canadian Journal of Chemistry, 55:1062–
1072, 1977.

[48] D. K. Chattoraj and K. S. Birdi. Adsorption and the Gibbs Surface Excess. Springer US, 1984.

[49] A. W. Omta, M. F. Kropman, S. Woutersen, and H. J. Bakker. Negligible effect of ions on the

hydrogen-bond structure in liquid water. Science, 301:347–349, 2003.

[50] J. W. Gibbs. The Collected Works of J. W. Gibbs. 1928.

[51] L. Onsager and N. N. T. Samaras. The surface tension of Debye-Hückel electrolytes. The

Journal of Chemical Physics, 2:528–536, 1934.

[52] T. Yamaguchi, O. Lindqvist, T. Claeson, and J. B. Boyce. EXAFS and X-ray diffraction
studies of the hydration structure of stereochemically active Sn(II) ions in aqueous solution.
Chemical Physics Letters, 93:528–532, 1982.

[53] Y. Kameda, H. Saitoh, and O. Uemura. The hydration structure of NO−

3 in concentrated
aqueous sodium nitrate solutions. Bulletin of the Chemical Society of Japan, 66:1919–1923,
1993.

[54] T. Yamaguchi, M. Niihara, T. Takamuku, H. Wakita, and H. Kanno. Scandium(III) hydration
in aqueous solution from X-ray diffraction and X-ray absorption fine structure measurements.
Chemical Physics Letters, 274:485–490, 1997.

[55] J. C. Hindman, A. Svirmickas, and M. Wood. Relaxation processes in water. A study of the
proton spin-lattice relaxation time. The Journal of Chemical Physics, 59:1517–1522, 1973.

[56] D. W. G. Smith and J. G. Powles. Proton spin-lattice relaxation in liquid water and liquid

ammonia. Molecular Physics, 10:451–463, 1966.

96

[57] M. F. Kropman and H. J. Bakker. Dynamics of water molecules in aqueous solvation shells.

Science, 291:2118–2120, 2001.

[58] M. F. Kropman and H. J. Bakker. Femtosecond mid-infrared spectroscopy of aqueous solvation

shells. The Journal of Chemical Physics, 115:8942–8948, 2001.

[59] A. Einstein. Investigations on the theory of the Brownian movement. Fürth, Methuen, London,

1926.

[60] R. Buchner, T. Chen, and G. Hefter. Complexity in “simple” electrolyte solutions: Ion pairing

in MgSO4(aq). The Journal of Physical Chemistry B, 108:2365–2375, 2004.

[61] W. Wachter, Š. Fernandez, R. Buchner, and G. Hefter. Ion association and hydration in aqueous
solutions of LiCl and Li2SO4 by dielectric spectroscopy. The Journal of Physical Chemistry
B, 111:9010–9017, 2007.

[62] R. Buchner, S. G. Capewell, G. Hefter, and P. M. May. Ion-pair and solvent relaxation processes
in aqueous Na2SO4 solutions. The Journal of Physical Chemistry B, 103:1185–1192, 1999.
[63] I. Benjamin. Theoretical study of ion solvation at the water liquid-vapor interface. The Journal

of Chemical Physics, 95:3698–3709, 1991.

[64] M. A. Wilson and A. Pohorille. Interaction of monovalent ions with the water liquid-vapor
interface: A molecular dynamics study. The Journal of Chemical Physics, 95:6005–6013,
1991.

[65] G. Markovich, R. Giniger, M. Levin, and O. Cheshnovsky. Photoelectron spectroscopy of
negative ions solvated in clusters. Zeitschrift für Physik D Atoms, Molecules and Clusters,
20:69–72, 1991.

[66] G. Markovich, S. Pollack, R. Giniger, and O. Cheshnovsky. Photoelectron spectroscopy of
Cl−, Br−, and I− solvated in water clusters. The Journal of Chemical Physics, 101:9344–9353,
1994.

[67] L. Perera and M. L. Berkowitz. Many-body effects in molecular dynamics simulations of
Na+(H2O)n and Cl−(H2O)n clusters. The Journal of Chemical Physics, 95:1954–1963, 1991.
[68] L. Perera and M. L. Berkowitz. Structure and dynamics of Cl−(H2O)20 clusters: The effect of
the polarizability and the charge of the ion. The Journal of Chemical Physics, 96:8288–8294,
1992.

[69] L. Perera and M. L. Berkowitz. Stabilization energies of Cl−, Br−, and I− ions in water

clusters. The Journal of Chemical Physics, 99:4222–4224, 1993.

[70] L. Perera and M. L. Berkowitz. Structures of Cl−(H2O)n and F−(H2O)n (n=2,3,...,15) clusters.
Molecular dynamics computer simulations. The Journal of Chemical Physics, 100:3085–3093,
1994.

97

[71] L. S. Sremaniak, L. Perera, and M. L. Berkowitz. Enthalpies of formation and stabilization
energies of Br−(H2O)n (n=1,2, ..., 15) clusters. Comparisons between molecular dynamics
computer simulations and experiment. Chemical Physics Letters, 218:377–382, 1994.

[72] I.-C. Yeh, L. Perera, and M. L. Berkowitz. Photodetachment spectra of Cl−(H2O)n clusters.

Predictions and comparisons. Chemical Physics Letters, 264:31–38, 1997.

[73] L. X. Dang and B. C. Garrett. Photoelectron spectra of the hydrated iodine anion from

molecular dynamics simulations. The Journal of Chemical Physics, 99:2972–2977, 1993.

[74] L. X. Dang and D. E. Smith. Molecular dynamics simulations of aqueous ionic clusters using

polarizable water. The Journal of Chemical Physics, 99:6950–6956, 1993.

[75] L. X. Dang. Characterization of water octamer, nanomer, decamer, and iodide-water interac-
tions using molecular dynamics techniques. The Journal of Chemical Physics, 110:1526–1532,
1999.

[76] S. J. Stuart and B. J. Berne. Effects of polarizability on the hydration of the chloride ion. The

Journal of Physical Chemistry, 100:11934–11943, 1996.

[77] S. J. Stuart and B. J. Berne. Surface curvature effects in the aqueous ionic solvation of the

chloride ion. The Journal of Physical Chemistry A, 103:10300–10307, 1999.

[78] G. H. Peslherbe, B. M. Ladanyi, and J. T. Hynes. Structure of NaI ion pairs in water clusters.

Chemical Physics, 258:201–224, 2000.

[79] S. Ghosal, A. Shbeeb, and J. C. Hemminger. Surface segregation of bromine in bromide doped
NaCl: Implications for the seasonal variations in arctic ozone. Geophysical Research Letters,
27:1879–1882, 2000.

[80] P. Jungwirth and D. J. Tobias. Ions at the air/water interface. The Journal of Physical Chemistry

B, 106:6361–6373, 2002.

[81] J. H. Hu, Q. Shi, P. Davidovits, D. R. Worsnop, M. S. Zahniser, and C. E. Kolb. Reactive uptake
of Cl2(g) and Br2(g) by aqueous surfaces as a function of Br− and I− ion concentration: The
effect of chemical reaction at the interface. The Journal of Physical Chemistry, 99:8768–8776,
1995.

[82] E. M. Knipping, M. J. Lakin, K. L. Foster, P. Jungwirth, D. J. Tobias, R. B. Gerber, D. Dabdub,
and B. J. Finlayson-Pitts. Experiments and simulations of ion-enhanced interfacial chemistry
on aqueous NaCl aerosols. Science, 288:301–306, 2000.

[83] P. Jungwirth and D. J. Tobias. Surface effects on aqueous ionic solvation: A molecular
dynamics simulation study of NaCl at the air/water interface from infinite dilution to saturation.
The Journal of Physical Chemistry B, 104:7702–7706, 2000.

[84] E. A. Raymond and G. L. Richmond. Probing the molecular structure and bonding of the
surface of aqueous salt solutions. The Journal of Physical Chemistry B, 108:5051–5059, 2004.

98

[85] P. Terpstra, D. Combes, and A. Zwick. Effect of salts on dynamics of water: A raman

spectroscopy study. The Journal of Chemical Physics, 92:65–70, 1990.

[86] F. Rull and J. A. de Saja. Effect of electrolyte concentration on the raman spectra of water in

aqueous solutions. Journal of Raman Spectroscopy, 17:167–172, 1986.

[87] N. Abe and M. Ito. Effects of hydrogen bonding on the Raman intensities of methanol, ethanol

and water. Journal of Raman Spectroscopy, 7:161–167, 1978.

[88] P. Ayotte, G. H. Weddle, J. Kim, and M. A. Johnson. Vibrational spectroscopy of the ionic
hydrogen bond: Fermi resonances and ion-molecule stretching frequencies in the binary
X−·H2O (X = Cl, Br, I) complexes via argon predissociation spectroscopy. Journal of the
American Chemical Society, 120:12361–12362, 1998.

[89] P. Ayotte, G. H. Weddle, and M. A. Johnson. An infrared study of the competition between
hydrogen-bond networking and ionic solvation: Halide-dependent distortions of the water
trimer in the X−·(H2O)3, (X=Cl, Br, I) systems. The Journal of Chemical Physics, 110:7129–
7132, 1999.

[90] D. Liu, G. Ma, L. M. Levering, and H. C. Allen. Vibrational spectroscopy of aqueous sodium
halide solutions and air-liquid interfaces: Observation of increased interfacial depth. The
Journal of Physical Chemistry B, 108:2252–2260, 2004.

[91] G. Archontis, E. Leontidis, and G. Andreou. Attraction of iodide ions by the free water surface,
revealed by simulations with a polarizable force field based on Drude oscillators. The Journal
of Physical Chemistry B, 109:17957–17966, 2005.

[92] E. C. Brown, M. Mucha, P. Jungwirth, and D. J. Tobias. Structure and vibrational spectroscopy
of salt water/air interfaces: Predictions from classical molecular dynamics simulations. The
Journal of Physical Chemistry B, 109:7934–7940, 2005.

[93] L. Vrbka, M. Mucha, B. Minofar, P. Jungwirth, E. C. Brown, and D. J. Tobias. Propensity of
soft ions for the air/water interface. Current Opinion in Colloid & Interface Science, 9:67–73,
2004.

[94] S. Ghosal, J. C. Hemminger, H. Bluhm, B. S. Mun, E. L. D. Hebenstreit, G. Ketteler, D. F.
Ogletree, F. G. Requejo, and M. Salmeron. Electron spectroscopy of aqueous solution
interfaces reveals surface enhancement of halides. Science, 307:563–566, 2005.

[95] C. Tian, S. J. Byrnes, H.-L. Han, and Y. Ron Shen. Surface propensities of atmospherically rel-
evant ions in salt solutions revealed by phase-sensitive sum frequency vibrational spectroscopy.
The Journal of Physical Chemistry Letters, 2:1946–1949, 2011.

[96] L. Piatkowski, Z. Zhang, E. H. G. Backus, H. J. Bakker, and M. Bonn. Extreme surface

propensity of halide ions in water. Nature Communications, 5, 2014.

99