This is the accepted manuscript made available via CHORUS. The article has been
published as:
Ultrafast structural dynamics of VO_{2}
Sergiy Lysenko, Nardeep Kumar, Armando Rúa, José Figueroa, Junqiang Lu, and Félix
Fernández
Phys. Rev. B 96, 075128 — Published 14 August 2017
DOI: 10.1103/PhysRevB.96.075128
Ultrafast Structural Dynamics of VO2
Sergiy Lysenko,∗ Nardeep Kumar, Armando Rúa, José Figueroa, Junqiang Lu, and Félix Fernández
Department of Physics, University of Puerto Rico, Mayaguez, Puerto Rico 00681, USA
(Dated: July 27, 2017)
Distinct contribution of acoustic and optical phonons in light-induced lattice transformation was
resolved at different time scales by monitoring the insulator-to-metal phase transition in epitaxial
and non-epitaxial VO2 films. Applying the ultrafast angle-resolved light scattering technique we
demonstrate a significant influence of internal misfit strain in epitaxial films on sub-picosecond phase
transition dynamics. This technique also allows observing a contribution of structural defects in the
evolution of the transient state. The ultrafast structural phase transition dynamics is discussed
in terms of the Ginzburg-Landau formalism. Using a set of experimental data we reconstruct the
thermodynamic potential of photoexcited VO2 and provide a phenomenological model of ultrafast
light-induced structural phase transition.
I.
INTRODUCTION
Vanadium dioxide is a classic example of a ferroelastic material that undergoes first-order insulator-to-metal
transition (IMT) at temperature Tc =341 K1 . The transition is accompanied by structural change2,3 from low5
temperature monoclinic (M1 -structure, space group C2h
4,5
) to high-temperature rutile (R-structure, space group
14 6
D4h
) lattice symmetry. Owing to strong electron correlations, the phase transition dynamics of VO2 is a very
complex process. Crystallinity and internal strain significantly alter the width and tilt of thermal hysteresis
and can shift the temperature of IMT by several degrees
above or below Tc point of stoichiometric unstrained
VO2 crystal.7,8 Numerous theoretical approaches have
been developed to describe VO2 phase stability. While
some models explain the metal-insulator transition in
terms of the Mott-Hubbard mechanism of correlation gap
opening,9–12 other approaches consider band models with
Peierls instability13–15 or describe it by combining both
mechanisms.16–18 Moreover, all these substantially different approaches provide fairly close qualitative and quantitative explanations of VO2 physical properties. Therefore, in order to better understand the actual mechanisms
of IMT in VO2 , along with theoretical approaches new
experimental methods are required to track the process
dynamics.
While the IMT of VO2 can be induced by heat, this
transition can be also initiated or altered by strain,8,19–22
electric current,23–25 by doping with different metal
ions,26,27 by THz radiation28 and by light.29–34 The investigation of ultrafast photoinduced processes in phasechange materials is of special interest since methods of
ultrafast spectroscopy can potentially track electron and
phonon lattice dynamics separately. Existing theoretical
models provide a satisfactory description of ultrafast dynamics of photoexcited VO2 .35–37 However, these models also reveal the great complexity of the problem, and
the general picture of evolution for electronic and lattice subsystems is still poorly understood. The problem
of the nature of photoexcited states remains open. In
most cases, experimental and theoretical studies assume
for simplicity only the photogeneration of dense electron-
hole plasma but neglect the possibility of excitonic and
polaronic states in VO2 .
As shown in numerous works,22,38–45 the photoinduced
IMT in VO2 depends on incident light intensity, wavelength, and also on film crystallinity, structural defects,
and internal strain. Recent studies indicate that the ultrafast structural phase transition (SPT) is likely to be
triggered by a screening of electron correlations on the
sub-picosecond scale.34,36,46,47 The study of ultrafast response in VO2 by Mayer et al.28 shows that the subpicosecond nonthermal IMT of VO2 can also be induced
by a strong THz pulse. Since the THz excitation does not
produce direct photodoping, the proposed model of the
IMT involves carrier tunneling in the presence of strong
THz field. We note that the utilization of strong THz
field for triggering and monitoring the IMT in VO2 is
of special interest since it can provide new information
about bonded electronic states, cooperative effects and
the possible presence of polaronic states,10,48 and their
role in structural and electronic properties of VO2 .
In this paper, we demonstrate a comprehensive approach describing the pathways of photoinduced firstorder structural transition in VO2 versus optical excitation level, material strain, morphology and structural
defects. The discussion starts by presenting a semiclassical computation (Sec. II) of vibration density of
states and adiabatic relaxation of kinetic energy on the
sub-picosecond time scale. We apply a method of molecular dynamics (MD) which neglects electron-electron correlations and, therefore, simulates the actual case of photoinduced screening of the correlations. In order to better
understand the results of MD computation and to find
relations between different degrees of freedom of ultrafast structural dynamics, we performed the experimental study of VO2 with techniques described in Sec. III.
In Sec. IV we report on the experimental results of
time-resolved transmission, reflection and angle-resolved
light scattering within nanosecond, picosecond and subpicosecond time scales. We show a noticeable influence
of photoacoustic strain on nonlinear optical (NLO) dynamics of VO2 on the nanosecond time scale. Then, applying time-resolved light scattering technique, we observe grain-size-dependent phase transition on the sub-
picosecond time scale at different levels of optical excitation. The dynamics of photoinduced SPT shows a
considerable anisotropy for epitaxial VO2 film due to influence of the misfit strain. The evolution of excited state
depends on size of VO2 grains/domains and on concentration of structural defects. In Sec. V we summarize our
experimental data and present a phenomenological model
of ultrafast structural dynamics of VO2 . Our modeling
is based on the numerical integration of the equation of
motion for the effective displacement of lattice ions in
terms of Ginzburg-Landau formalism.
II.
SUB-PICOSECOND MOLECULAR
DYNAMICS
Recent progress in experimental and theoretical studies of phase transition in VO2 evidences that the screening of electron correlations is a key factor which triggers the SPT.12,18,34,36,46 Photoinduced metallization of
VO2 occurs during the light pulse interaction, showing
Mott-type behavior with possible band-gap collapse on
a timescale less than ∼60 fs.46,49 Nevertheless, the total
structural transformation of VO2 lattice from monoclinic
to rutile symmetry is slower with durations ranging from
∼80 fs to several hundred femtoseconds39 or even several picoseconds,50,51,54 depending on the level of optical excitation, wavelength, and morphology of the sample. In this context, the transient lattice distortion can
be modeled by using semi-classical computational methods of molecular dynamics (MD) which neglect electronelectron correlations. We note that the MD method does
not permit computation of exact lattice dynamics during
SPT, but provides a meaningful estimation of quantitative parameters of these dynamics (i.e. root-mean-square
(rms) atomic displacements, vibrational density of states,
kinetics of energy relaxation, etc.).
To compute the MD we used the QuantumWise software package55 . The method of MD employs BornOppenheimer approach with semi-classical ReaxFF (reactive force field) potential which describes interactions
between all atoms.56,57 This approach significantly reduces the computation cost and allows analyzing large
atomic clusters. ReaxFF potential consists of various potential energy functions, including Coulomb and van der
Waals interactions, which provide accurate descriptions
of bond breaking and bond formation. A full description of these functions can be found in Ref. 56. Each
ReaxFF function is based on an appropriate many-body
expansion and is obtained from solving the Schrödinger
equation for fixed positions of the nuclei.58
The ReaxFF method retains nearly the accuracy of
quantum mechanical calculations.56,57 However it neglects electron-electron correlations.
Therefore this
method is very promising to simulate the lattice dynamics for the case when electron-electron correlations
are screened by photoexcitation of a dense electron-hole
plasma.
FIG. 1. (a) The time evolution of kinetic energy and effective
transient temperature T * for pure monoclinic VO2 (M1 ) and
for structure with 2.5% of oxygen vacancies (VO2 (M1 )+2.5%
O vac.) at T =15 K and for rutile VO2 (R) at T =385 K. (b)
Kinetic energy and T ∗ for VO2 (M1 ) at T =313 K. The envelope is shown by dashed lines for VO2 (M1 ) at t >20 fs.
VO2 is a strongly-correlated oxide and electron correlations play an important role in stabilization of the lowT monoclinic VO2 lattice. The semi-classical method of
MD provides rather an estimation than the actual trajectory of the lattice transformation, where the final structure corresponds to an equilibrium uncorrelated structure
of VO2 which is neither monoclinic nor rutile. Therefore,
taking the real ’correlated’ structure of VO2 as the input
for MD calculations, we obtain the evolution of correlated to uncorrelated VO2 structure. The method of MD
does not use an external ’excitation’. To apply the concept of MD, we assume that the ultrafast photoexcitation
produces only screening of electron-electron correlations,
while all other physical characteristics of VO2 remain unchanged.
Upon photoexcitation of VO2 , the lattice transformation within sub-picosecond time scale is considered as
an adiabatic process, where the energy exchange with
the surroundings is neglected. As the input parameters
for MD calculation we use only the initial temperature
and cluster of vanadium and oxygen atoms constructed
according to VO2 lattice parameters. Initial velocities
of atoms were set according to the Maxwell-Boltzmann
distribution, and the calculation was performed for adiabatic relaxation with constant number of particles, volume and total energy of the system.
To model the photoinduced lattice dynamics, the initial lattice parameters of the monoclinic VO2 (M1 ) unit
2
ports the estimation of characteristic time for the resonant elastic scattering of optical phonons on V-V dimers
performed in Ref. 59. The time τ =120 fs is comparable to experimentally observed characteristic time of the
fastest component of the photoinduced structural transition in VO2 , which ranges from ∼80 fs to several hundred
femtoseconds.39 Therefore, it is very likely that the elastic phonon-phonon scattering in a highly non-equilibrium
environment plays a significant role in the SPT dynamics, triggering the lattice transformation from monoclinic
to rutile symmetry on the sub-picosecond time scale.
The computation of VO2 (M1 ) lattice dynamics at
low-temperature T (0 fs)=15 K results in an increase of
the effective temperature up to T *=768 K [Fig. 1(a)]
which is very close to the Debye temperature of VO2
(TD =750 K).60,61 This fact indicates a possibility of excitation of all vibration modes in high-T VO2 62 and,
very likely, it contributes to structural instability of photoexcited low-T phase, even at initial temperature as
low as 15 K. On the observed timescale the average
kinetic energy approaches to Ekin =2.29×103 cal/mole.
The difference between this energy and the energy which
corresponds to the transition temperature Tc *=341 K
(Ekin =1.02×103 cal/mole) is 1.27×103 cal/mole. It
is noted that this value is very close to the latent
heat of the phase transition experimentally obtained by
Ryder60 (1.02×103 cal/mole) and by Chandrashekhar
et al.61 (1.12×103 cal/mole). In terms of thermodynamics, this means that the photoinduced screening of
electron-electron correlations triggers an exothermic reaction which guarantees the thermally-induced phase
transition of VO2 , even if the initial temperature of the
sample is near absolute zero.
cell were taken from experimentally obtained data in
Ref. 5: am =5.7517 Å, bm =4.5378 Å, cm =5.3825 Å, the
angle between am and cm axes is βm =122.646◦ . The
MD computation was performed for a VO2 (M1 ) cluster
of 20,217 atoms at temperatures T =15 K and T =313 K.
In order to reduce the random component of atomic motion, the main calculations were performed for T =15 K.
To estimate the influence of oxygen vacancies on MD, additional calculations were performed for a VO2 (M1 ) cluster with 2.5% of vacancy defects randomly distributed
inside the lattice. The 2.5% oxygen deficiency is close to
the stability limits for the VO2 structure. To compare
MD of monoclinic VO2 (M1 ) and rutile VO2 (R), we also
calculated MD for a VO2 (R) cluster of 49,152 atoms at
temperature T =385 K, slightly above the Tc point. Crystal structure parameters for rutile VO2 (R) were taken
from Ref. 6 (ar =4.53 Å, br =4.53 Å, cr =2.869 Å).
The computational study of adiabatic lattice dynamics reveals significant instability of the VO2 lattice with
monoclinic symmetry, while the rutile phase was found to
be relatively stable. Fig. 1(a) shows the time evolution
of kinetic energy Ekin and effective transient temperature T * for VO2 (M1 ) and VO2 (R). The temperature T *
was calculated from kinetic energy through the equation
Ekin = 3/2N kB T *, where N is the number of atoms
and kB is Boltzmann constant. A comparison of kinetic
energy evolution for monoclinic and rutile phases shows
that Ekin for VO2 (M1 ) increases within 900 fs by two orders of magnitude, from 45 to 2.29×103 cal/mole, while
for VO2 (R) this change is noticeably less. For VO2 (R)
Ekin changes from 1.15×103 to 1.05×103 cal/mole: after
a small drop it returns to nearly the same level. The
dynamics of VO2 (M1 ) is more pronounced, and increase
of the temperature from T (0 fs)=15 K to T (0 fs)=313 K
[Fig. 1(b)] does not noticeably affect the relaxation kinetics, but results in an increase of the final average level
of Ekin to 2.7×103 cal/mole (T *=905 K).
A significant difference in the relative change of Ekin
for correlated low-T monoclinic phase and for noncorrelated high-T rutile phase was a priori expected as
the result of photoinduced screening of electron correlations, and eventually reliably confirmed by the ReaxFFbased MD simulation. Thus, a screening of Coulomb
repulsion for non-correlated rutile VO2 does not provide
a considerable change of Ekin , while for the correlated
low-T monoclinic phase the screening results in instantaneous rise of kinetic energy followed by lattice instability.
Also, this indicates that the electron correlations in the
low-T phase are the major factor in stabilizing the VO2
monoclinic symmetry.
The envelope of Ekin (t) for t >20 fs in Fig. 1 can be approximated by the function E0 ±E1 exp(−t/τ ), where E0
and E1 are fitting constants and τ is a characteristic relaxation time. The transient dynamics of kinetic energy
shows near-exponential oscillatory decay with τ =120 fs.
This process is associated with elastic phonon-phonon
scattering and randomization of the phase of coherent
phonon oscillations. Moreover, this result strongly sup-
FIG. 2. The computation of molecular dynamics for vanadium (V), oxygen (O) and for all atoms (V+O) of VO2 (M1 )
at T =15. (a) Velocity autocorrelation function. (b) Velocity
probability distribution within 500 fs (c) rms atomic displacement.
3
Additional analysis of the VO2 (M1 ) cluster with 2.5%
of oxygen vacancies shows that the structural defects
slightly increase the kinetic energy released during the
phase transition [Fig. 1(a)], while the nonequilibrium dynamics remains nearly the same as for pure crystalline
VO2 . Therefore, it could be expected that the vacancies
facilitate the ultrafast SPT. However, the MD calculations do not consider an influence of defects on the formation of specific localized electronic states which can
significantly contribute in SPT dynamics. As a result,
the real photoinduced dynamics in VO2 with high concentration of structural defects can be more complex.
Figures 2(a) and 2(b) show the velocity autocorrelation
function (ACF) and velocity probability distribution for
vanadium and oxygen atoms, respectively. The velocity
ACF shows oscillatory decay within ∼500 fs. This behavior indicates initial coherent atomic motion, but coherency eventually vanishes in ∼500 fs. Within this time
scale the rms displacement of all atoms [i.e. weighted average rms value for V and O atoms, Fig. 2(c)] shows some
oscillatory behavior, approaching the value of 0.32 Å with
respect to initial position. While the rms displacement
is not, strictly speaking, an order parameter ηc (effective
displacement of lattice ions during photoinduced SPT),
it compares well with the value of ηc = 0.26 Å obtained
recently by van Veenendaal for ultrafast SPT in VO2 .37
Therefore, it is reasonable to expect a correlation between these two values within ∼500 fs.
The vibrational density of states (DOS) for the
VO2 (M1 ) cluster was computed by Fourier transform of
the atomic velocities.63 Figure 3 shows the DOS computed within 1 ps time scale for all atoms and for Oand V-atoms separately. Additionally, DOS was calculated for the lattice with 2.5% of oxygen vacancies. The
presence of structural defects results in smearing of sharp
resonance peaks, but the overall spectrum of vibrational
DOS remains the same. MD computation clearly shows
two broad resonances of optical phonons with maxima
at 6.2 THz (25.6 meV) and at 17.7 THz (73 meV). The
VO2 lattice oscillations near 6 THz were observed in several previous studies.32,33,39,43,49,64,65 The second resonance was clearly observed by Kübler et al. in Ref. 32.
Thus, the calculated spectrum of vibrational DOS is fully
supported by experimental time-domain and THz spectroscopy data.
III.
FIG. 3. Vibrational density of states of all atoms (V+O) and
only oxygen (O) and vanadium (V) atoms. Calculations are
performed for pure VO2 (M1 ) and for VO2 (M1 ) with 2.5% of
oxygen vacancies (V+O+vac.).
lic vanadium (99.95% purity) target, using ∼4 J/cm2
laser fluence. VO2 films were grown at 30 mTorr pressure
in the PLD chamber, in oxygen and argon atmosphere,
with O2 and Ar flow rates of 20 and 5 std. cm3 /min,
respectively. The substrate temperature was maintained
at 550◦ C.
All samples were prepared at identical nominal conditions using the same protocol of PLD growth. Deposition of VO2 on different substrates provided a variety of desired structures and morphologies of the material. The VO2 films’ phase, crystallographic orientation, and morphology was determined by x-ray diffraction
(Bruker D8 Discover X-ray diffractometer) and atomicforce-microscopy (AFM, Park Scientific Instruments, Autoprobe CP). Only VO2 films with single monoclinic M1
phase at room temperature were selected for the current
study.
The synthesis of VO2 on single crystal sapphire resulted in epitaxial films and relatively low concentration of structural defects.42,66 These films experience
moderate misfit strain owing to lattice mismatch at the
film/substrate interface.67–69 As determined from the
XRD data for VO2 /Al2 O3 (A-cut), the film’s out-of-plane
orientation is with its bm axis normal to the substrate surface. Additional azimuthal ϕ-scans of the sample were
performed to determine the in-plane orientation of VO2
crystallites. It was found that the film am axis is normal
to the sapphire c-axis [0001]. Therefore, the cm axis is
also in the film plane.
For VO2 deposited on (0001) Al2 O3 (C-cut) substrate,
the XRD reflection at 2θ=39.8◦ indicated that the (010)m
planes are parallel to the substrate surface. Due to the
three-fold symmetry of sapphire around the c-axis, the
EXPERIMENT
Several high quality epitaxial and non-epitaxial VO2
films with substantially different morphology, different
thicknesses, domain structure and in-plane oriented crystallites were synthesized. VO2 films with thicknesses
of 30 nm, 50 nm, 80 nm and 100 nm were grown by
pulsed laser deposition (PLD) technique on amorphous
SiO2 and single crystal Al2 O3 (C-cut and A-cut sapphire)
substrates. An excimer KrF laser with 20-ns pulses of
248 nm wavelength, was used to ablate a rotating metal4
scatterometer setup is illustrated in Fig. 4. In all scattering measurements, the pump (λ=800 nm) and probe
(λ=400 nm) pulses are overlapped on the sample surface
at normal incidence and focused to a spot size of 0.7 mm
and 70 µm respectively. To prevent nonlinear interaction
of probe pulse with the sample, its intensity was reduced
by several orders of magnitude compared to pump pulse.
Probe light was linearly polarized by a Glan-type prism
GP. The polarization of pump pulse was set circular by
quarter-wave plate λ/4. A computer-controlled optomechanical delay line was used to set a time delay t between
pump and probe pulses with a resolution of 10 fs. Samples were placed at the focal point of the custom-built
metallic elliptical mirror with 20-cm diameter aperture.
The mirror was used to collect scattered light within the
whole hemisphere over the sample surface and to project
the image to a 16-bit charge-coupled device (CCD).72
Color filter F was used to filter out the pump wavelength
light from the recorded image. Obtained scattering patterns were recalculated into indicatrices of BidirectionalScatter-Distribution-Function (BSDF) versus polar θ and
azimuthal ϕ angles or spatial frequency f of the surface.
BSDF is the function with close resemblance to surface
power spectral density, and it can be used for the surface
analysis within ”scatter prediction” approach.73 Two amplified silicon detectors PD1 and PD2 were used to monitor transient total integrated scattering and reflection
correspondingly. Detectors are conjugated with a gated
data processor.
deposited VO2 film is twinned, with orientation of the
am axis along three equivalent crystallographic directions
of the substrate as [100]m ||[100]Al2O3 , [100]m ||[010]Al2O3
and [100]m ||[1̄1̄0]Al2O3 .42
VO2 films deposited on SiO2 show only out-of-plane
orientation, and are expected to have higher concentration of oxygen vacancies and other structural defects. The strong reflection observed at 2θ=27.98◦ for
VO2 /SiO2 indicated that the (011)m plane is parallel to
the substrate.
In the present study, the VO2 films deposited on
Al2 O3 (C-cut) and SiO2 substrates were used to observe
the influence of photoacoustic excitations and structural
defects on SPT. The 50-nm-thick VO2 /Al2 O3 (C-cut) film
was used to monitor fluence-dependent evolution of SPT
on picosecond time scale in order to reconstruct the thermodynamic potential of VO2 in its nonequilibrium photoexcited state. The VO2 /Al2 O3 (A-cut) film was used
to study the influence of misfit strain and structural
anisotropy on SPT dynamics.
FIG. 4. Experimental setup for ultrafast angle-resolved light
scattering measurements within the hemisphere. BS: beam
splitter; DM: dichroic mirror; GP: Glan-type polarizer; NF:
neutral density filter; F: color filter; L: lens; λ/4: quarter-wave
plate; SH: sample holder; PD1 and PD2: silicon photodetectors; RM: removable mirror.
IV.
RESULTS AND DISCUSSION
The light-induced SPT in VO2 is a complex process
which depends on excitation wavelength, fluence and
sample morphology. A laser pulse with fluence above
the phase transition threshold (F0 ≃3 mJ/cm2 ) induces
ultrafast structural transition, accompanied by a noticeable change of optical and electronic properties of VO2 .
Therefore dynamics of the SPT can be tracked by monitoring transient reflectivity R(t), transmittance T r(t) or
light scattering Is (t). Fig. 5 shows the typical evolution
of ultrafast light scattering signal (inset) upon photoinduced phase transition within ∼500 fs and also differential reflectivity (main panel) within several nanoseconds for an epitaxial 50 nm VO2 /Al2 O3 (C-cut) film.
Rigorous observation of the SPT in different samples
by pump-probe optical techniques allows distinguishing
several characteristic time scales for qualitatively different transient dynamics. These are time scales (i) up to
∼500 fs, (ii) between 500 fs and ∼40 ps,74 and (iii) above
∼40 ps. In the next sections, we discuss the SPT and specific optical response for each timescale. We will address
first the dynamics (iii) in Sec. IV.A, then (i) in Sec. IV.B
and IV.C, and finally (ii) in Sec. IV.D.
A Spectra-Physics Ti:Sapphire femtosecond laser system was used as a source to induce and to monitor the
phase transition dynamics of VO2 at room temperature.
The system generates light pulses with λ=800 nm wavelength and 130 fs duration. Owing to the relatively slow
recovery rate of non-epitaxial VO2 /SiO2 after optical
excitation,70,71 to monitor SPT in these films the repetition rate of laser pulses was set to 200 Hz. To study
epitaxial VO2 /Al2 O3 films the laser repetition rate was
set to 1 kHz. In this work, we performed different pumpprobe transient reflection, transmission, and light scattering measurements. Depending on experimental geometry, the wavelengths of pump and probe pulses were
λ=800 nm or frequency-doubled λ=400 nm by a BBO
crystal.
The time- and angle-resolved hemispherical elastic
light scattering (TARHELS) technique was used to study
multiscale dynamics of photoinduced SPT in VO2 . The
5
observed only at F ≥ 3 mJ/cm2 , while at lower fluence the increased noise significantly affected the measurement accuracy. For 100-nm-thick VO2 /SiO2 film
the NLO response was much stronger. By plotting the
NLO signal at 1 ps delay versus pump fluence for this
sample, it was possible to derive accurately the threshold fluence F0 , required to initiate the ultrafast SPT in
VO2 material.49 Thus, Fig. 6(b) shows the normalized instantaneous change of transmittance |∆T r(1ps)/T r(0)|,
where the increasing optical excitation contributes to the
rise of the optical transmission starting from a fluence
of 3 mJ/cm2 . This specific optical response indicates a
qualitative change in NLO dynamics of VO2 . As shown
by O’Callahan et al.49 , the threshold F0 can vary for
different VO2 samples within a short range: from 2 to
6 mJ/cm2 and, according to the literature31–33,50,54,75 ,
the fluence of 3 mJ/cm2 compares well with the threshold
values for the ultrafast phase transition in VO2 . Therefore, in this study, we assume that the observed qualitative change of NLO signal at fluence F0 =3 mJ/cm2 is
related to the threshold of sub-picosecond SPT of VO2 .
At F <3 mJ/cm2 the sub-picosecond NLO signal
|∆T r(1ps)/T r(0)| is nearly constant. Presumably, in this
case the NLO response is related to photogeneration of
dense electron-hole plasma in the film. However on longer
(picosecond) timescale the transient transmission shows
a decrease within first 200 ps [Fig. 6(a)], associated with
the SPT.
Figure 6(a) shows a striking difference in the pathways of phase transition dynamics for VO2 /SiO2 below
and above F =6 mJ/cm2 pump level. At lower excitation, the system recovers back into the initial insulating
phase within several nanoseconds, while at higher excitation the recovery process does not start on the observed
timescale, and the film is continuously switching into its
metallic state. This transition is associated with nucleation and growth of metallic phase inside of photoexcited
monoclinic VO2 . Moreover, above F =6 mJ/cm2 the total SPT dynamics does not depend much on excitation
level, and the full recovery of the system occurs on microsecond time scale, as shown in Fig. 6(c) for the fluence
9 mJ/cm2 .
Figures 6(a) and 6(c) show that the increase in the
excitation level above 6 mJ/cm2 to 9 mJ/cm2 increases
the characteristic recovery time τR by more than two
orders of magnitude: from τR =1.5 ns (F =1.5 mJ/cm2 )
to τR =200 ns (F = 9 mJ/cm2 ). This evidences the major contribution of heat to the SPT on a nanosecond
time scale at pump fluence above 6 mJ/cm2 : the heat
increases the film temperature above Tc and, therefore,
stabilizes the metallic phase. As a result, the recovery
time depends only on heat sink into the substrate within
several microseconds. However, it is very likely that at
laser excitation below 6 mJ/cm2 the average temperature of the film does not reach Tc point and the system
recovers rapidly, within several nanoseconds. We note
that the repetition rate of the laser pulses was reduced to
200 Hz for all measurements of non-epitaxial VO2 /SiO2
FIG. 5.
Evolution of transient differential reflectivity
∆R(t)/R0 upon photoinduced phase transition of epitaxial
50-nm VO2 /Al2 O3 (C-cut) film at different levels of optical
excitation. R0 is the reflectivity of an unperturbed sample,
∆R(t) = R(t)−R0 . The wavelengths of the optical pump and
probe pulses are 400 nm and 800 nm, respectively. The inset
shows transient differential signal ∆Is (t)/I0 of light scattering
integrated within the hemisphere. I0 is the scattering signal
of the unperturbed sample, ∆Is (t) = Is (t) − I0 .
A.
The role of photoacoustic excitations in phase
transition dynamics on nanosecond time scale
Using VO2 films with different thicknesses and morphologies, it is possible to enhance and monitor specific
nonequilibrium processes. The transient reflection and
transmission both provide essentially the same information about the photoinduced phase transition in VO2 .
For 50 nm VO2 /Al2 O3 film, the differential signal of
transient reflection is much stronger as compared to the
transmission. Therefore, to obtain higher signal-to-noise
ratio, the NLO dynamics for this sample was studied in
reflection geometry. On the other hand, for thicker, 100
nm VO2 /SiO2 film, the interference effects introduced
some artefacts into the reflection signal. Therefore, for
this sample reliable information about the phase transition dynamics was obtained from transmission measurements.
Figures 5 and 6 show a significant difference between
two NLO processes for the nanosecond time scale, for
50-nm thick epitaxial VO2 /Al2 O3 and for non-epitaxial
100-nm thick VO2 /SiO2 films. It was found that the
VO2 /Al2 O3 film switches into the metallic phase rapidly
without noticeable posterior dynamics at F =15 mJ/cm2 .
The recovery process back to insulating phase starts at
∼1.3 ns, as the laser fluence drops to 5.5 mJ/cm2 . The
photoexcitation of 100 nm thick VO2 /SiO2 film shows
more complex dynamics [Fig. 6(a)].
A similar behavior of the optical signal for VO2 /SiO2
and VO2 /Al2 O3 films was found within several picoseconds after optical excitation. The NLO response within
1 ps for thinner (50 nm) VO2 /Al2 O3 film was reliably
6
FIG. 6. Nonlinear optical dynamics of 100 nm VO2 /SiO2 film. (a) Transient transmittance of the film upon optical excitation. Numbers specify the laser pump fluence in mJ/cm2 . Gray arrows show two distinctly different directions of relaxation
kinetics for the fluence below and above 6 mJ/cm2 . The dashed curve is a fit using Eq.(1). (b) Magnitude of relative change
|∆T r(1ps)/T r(0)| of transmittance at 1 ps delay, where ∆T r(1ps) = T r(1ps) − T r(0). (c) Transient transmittance showing
recovery of VO2 /SiO2 film into insulating phase at different levels of optical excitation. The oscillatory part of the signal at
F =1.5 mJ/cm2 fitted (dashed curve) by Eq.(1). The signal on the right panel was obtained by using a continuous wave laser
source. (d) AFM topography of 2.6×2.6 µm2 area. The lower panel shows a cross-section of the AFM image corresponding
to the dashed line in the image. (e) Two consecutive measurements of transient transmittance within the same area of the
sample, confirming a pronounced oscillatory behavior. Dashed curves show the fit to equation A0 − A2 | sin(2πνt + ϕ0 )|.
At the fluence F =2.0 mJ/cm2 the frequency of oscillations decreases to ν=7×108 Hz [Figs. 6(a)]. This
low-frequency oscillatory response is difficult to associate
with acoustic waves propagating in normal direction to
the film surface. Additional contribution into oscillatory
dynamics can be produced by acoustic waves propagating
in lateral direction. According to AFM data [Fig. 6(d)],
the grain size in the film ranges from 0.2 µm to ∼1 µm.
Rough estimation shows that the acoustic wave with
ν=7×108 Hz can be excited within a large grain of size
1–2 µm.
In some areas of the film we observed a strong oscillatory signal with ν=8×108 Hz and with 2% modulation
of the sample transmittance, as shown in Fig. 6(e). The
signal is proportional to A0 − A2 | sin(2πνt + ϕ0 )|. This
distinct behavior may be assigned to photoacoustically
driven ferroelastic SPT, where a standing acoustic wave
modulates the strain field in the large VO2 grain.
Obtained experimental data allows describing the
light-induced SPT of VO2 in terms of the free energy difference ∆G between insulating and metallic phases. The
potential barrier ∆G is a function of temperature T0 at
the phase boundary and actual temperature T , molecu-
film. This rate was sufficient to provide complete heat
sink to the substrate and to prevent any possible accumulation of the heat in the film during the repetitive
excitation of the sample.
The NLO dynamics of VO2 /SiO2 shows the complex
behavior. For F =1.5 mJ/cm2 [Fig. 6(c)] it was possible
to resolve oscillations with frequency ν=4×109 Hz and
with impulse response:
T r(t) ≈ A0 +A1 [1−exp(−t/τR )]+A2 sin(2πνt+ϕ0 ), (1)
where A0 , A1 and A2 and time τR are fitting constants
and ϕ0 is the initial phase. Taking into account relatively
low frequency of observed oscillations, the oscillatory signal was assigned to photoacoustic material response.
The speed of the acoustic wave propagating across
the film of thickness d can be estimated using the following equation:76–78 v = 4dν. For d=100 nm and
ν=4×109 Hz, one obtains the speed v=1600 m/s. This
value is less than half the speed of sound in single-crystal
VO2 (v=4000 m/s).79 Such a large difference can be related to an amorphous structural component of the film
as well as to possible change of mechanical properties of
VO2 in its photoexcited nonequilibrium state.
7
lar latent heat q, number of molecules N , pressure p and
specific surface energy σ, and can be expressed as80
∆G = −(N q/T )∆T −N kB T ln(p/p0 )+N ∆µ+σ∆s, (2)
where ∆T = T − T0 , ∆s is the surface area of nucleating
grain, domain or cluster, ∆µ is the chemical potential
related to the difference in the bonding of VO2 molecules
for metallic and insulating phases; p0 is the equilibrium
pressure at the thermodynamical phase boundary.
Heat and strain significantly contribute to SPT on the
nanosecond time scale via changing ∆G. Equation (2)
provides a straightforward explanation of this dynamics. While the first term of Eq.(2) is related to the
conventional contribution of heat, the second term is related to the photoinduced pressure and strain in the film.
The photoexcitation decreases the binding energy difference N ∆µ, increases lattice temperature via electronphonon and phonon-phonon scattering, generates acoustic phonons and, as a result, alters ∆G producing the
SPT on the nanosecond time scale.
Since VO2 /SiO2 is a non-epitaxial film and is expected
to have relatively high acoustic impedance and high thermal boundary resistance,70 generated acoustic phonons
should be confined inside the film and do not propagate into the substrate on the monitored nanosecond time
scale. As a result, these phonons provide significant contribution into ferroelastic SPT. In contrast to VO2 /SiO2 ,
for epitaxial VO2 /Al2 O3 films the photoacoustic response
was not observed clearly due to lower acoustic impedance
[see Fig. 5]. Owing to epitaxial nature of VO2 /Al2 O3 ,
phonons leave the film volume rapidly without noticeable
acoustic modulation of the optical properties. Nevertheless, a recent observation of phonon dynamics in an epitaxial VO2 /Al2 O3 film by Abreu et al.81 shows that the
lowering of the sample temperature increases the signalto-noise ratio and provides reliable detection of acoustic
phonons.
B. Grain-size-dependent sub-picosecond phase
transition dynamics in the presence of anisotropic
internal misfit strain
FIG. 7. (a) Scattering indicatrix for unexcited VO2 /Al2 O3 (Acut) film. Arrows show the orientation of am and cm axes of
VO2 (M1 ). (b) 5×5 µm2 AFM topography of VO2 /Al2 O3 (Acut) film. The average lateral size of grains is 175 nm and
rms surface roughness is 4 nm. (c) BSDF(f ) cross-sections
at ϕ=75◦ at t=0 fs and at t=720 fs after photoexcitation;
F =12 mJ/cm2 . (d) Transient change ∆BSDF(t)/BSDF(0) of
the scattering signal upon light-induced phase transition at
F =12 mJ/cm2 , where ∆BSDF(t)=BSDF(t)-BSDF(0).
As shown above (see Fig. 6) the strain field produced
by photoacoustic wave in VO2 plays an essential role in
the light-induced SPT on the nanosecond time scale. According to recent studies of ultrafast light scattering,82,83
the influence of internal strain on the sub-picosecond
SPT dynamics of VO2 is also significant. In order to
obtain new information about how internal strain affects
the femtosecond SPT, we used a 80-nm-thick epitaxial
VO2 /Al2 O3 (A-cut) film with anisotropic misfit stain.
Since the VO2 film was deposited in its rutile phase,
because of the growth temperature employed, the magnitude of misfit strain in the film can be obtained by
considering the lattice mismatch between VO2 (R) and
sapphire Al2 O3 (A-cut) substrate. The sapphire A-plane
represents the cross-section of the unit cell along [0001]
and [1̄100] directions with respective values csaph and
√
3asaph , where csaph = 12.993 Å and asaph = 4.759 Å.84
These distances are large in comparison with the VO2 (R)
lattice parameters, but accommodation occurs by multiples of the ar and cr parameters. Corresponding to the
substrate [0001] direction it is noted that csaph ≈ 3ar ,
8
FIG. 8. (a) Transient change ∆BSDF(t)/BSDF(0) of the scattering signal upon light-induced SPT in VO2 /Al2 O3 (A-cut)
film; F =7 mJ/cm2 . Arrows show the orientation of am and cm axes of VO2 (M1 ). Dashed rhomb identifies oscillatory area
of scattering indicatrix. (b) Time-dependent cross-sections of the data mapped in (a) at ϕ=215◦ and ϕ=0◦ versus spatial
frequency of surface relief. (c) The cross-sections of the data mapped in (b) at ϕ=0◦ shows gradual change of oscillatory
dynamics versus spatial frequency.
quency f remains unchanged during the SPT. Moreover, the relative transient change of the scattering signal
∆BSDF(t)/BSDF(0) [Fig. 7(d)] does not show noticeable
angular dependence or anisotropy. Thus, despite the epitaxial nature of the film, the excitation of the film at laser
fluence F =12 mJ/cm2 produces simultaneous SPT with
nearly the same rate in all VO2 grains/domains with different spatial frequencies. These data show uniform and
isotropic SPT at this level of optical excitation. Nevertheless, it was found that the SPT dynamics changes
dramatically and becomes essentially anisotropic when
the excitation reduces to the levels close to the threshold
F0 .
with a lattice mismatch of approximately −4.4
√ %. Corresponding to the substrate [1̄100] direction, 3asaph ≈ 3cr
with a lattice mismatch of −4.2 %. Both of these values
imply that as the VO2 (R) film grows on the A-cut sapphire substrate, it will be compressed in all directions
along its surface. As the film is cooled to room temperature, strains will change, because of (i) the phase
transformation and (ii) film-substrate thermal expansion
mismatch. However, consideration of the VO2 (M1 ) lattice parameters and their orientation on the sapphire Asurface shows that the relatively large strain values calculated above will be largely maintained.
To observe the influence of misfit strain on femtosecond
SPT as a function of surface spatial frequency, we performed TARHELS measurements. These measurements
provide spatial resolution of phase transition dynamics
in differently oriented groups of domains, grains, and
clusters.73 To avoid experimental uncertainty owing to
polarization-dependent excitation of VO2 , we used a circularly polarized pump pulse.
To enhance the influence of anisotropic misfit strain on
SPT dynamics in VO2 grains/domains with different spatial frequencies, the pump fluence was set near the threshold F0 , at F =7 mJ/cm2 . As shown previously,32,33,43,65
the photoexcitation of VO2 with relatively low fluence results in activation of coherent phonon mode at ∼6 THz.
This allows us to observe clearly the influence of film morphology on amplitude and evolution of these oscillations.
The BSDF indicatrix of hemispherical light scattering by VO2 /Al2 O3 (A-cut) film is shown in Fig. 7(a).
The anisotropy of the scattering pattern is caused by
the texture of the epitaxial film [Fig. 7(b)] which is
highly influenced by the single-crystal substrate. The
photoinduced SPT at F =12 mJ/cm2 results in uniform
decrease of light scattering intensity for all spatial frequencies. The cross-sections of the scattering indicatrix at delay time t=0 fs and t=720 fs [Fig. 7(c)] show
that the profile of BSDF distribution versus spatial fre-
Transient
change
of
the
scattering
signal
∆BSDF(t)/BSDF(0)
obtained
for
the
same
VO2 /Al2 O3 (A-cut) film shows strong oscillatory
behavior [Fig. 8(a)] associated with photoexcitation
of active Raman modes in monoclinic VO2 due to
stretching and tilting of V-V dimers.32,33,43,64,65 In the
central part of scattering indicatrix, the amplitude of
oscillations is relatively high, but it decreases at larger
polar angles and the transient signal decays rapidly. Due
9
FIG. 9. (a) The Fourier transform of the oscillatory component in the scattering signal for different spatial frequencies.
(b) The thermodynamic potential of photoexcited VO2 . Φ
is altered by internal strain in the film. Strain contributes
in additional positive δφ1 or negative δφ2 component of the
potential Φ. Dashed arrows indicate the initial energy of the
VO2 system right after photoexcitation. Here the same level
of photoexcitation switches the grains/domains with different
thermodynamic potential into different excited states.
to epitaxial orientation of the film, oscillatory response
of grains/domains contributes to symmetric diffraction
pattern outlined by dashed rhomb in the center of
scattering indicatrix.
The cross-section of the ∆BSDF(t)/BSDF(0) indicatrix at two azimuthal directions ϕ=215◦ and ϕ=0◦
[Fig. 8(b)] reveals anisotropy in the evolution of coherent lattice oscillations which accompany the SPT. The
dashed line in the figures separates two mainly different
areas for the SPT process. Taking into account the essentially nonlinear dynamics of photoexcited VO2 , the
dashed line defines a set of time points when the character of structural dynamics changes qualitatively. For
VO2 structures with higher spatial frequencies the SPT
occurs faster, while for structures with f . 1.95 µm−1
at ϕ=215◦ and f . 1.2 µm−1 at ϕ=0◦ a monotonic
phase transformation does not start on the monitored
timescale. This behavior evidences the size-dependent
and anisotropic SPT dynamics, which rate depends on
size and orientation of VO2 grains/domains in the film.
The oscillation of ∆BSDF(t)/BSDF(0) for larger crystallites is anharmonic and contains several oscillatory
modes [see f =1.0 µm−1 in Fig. 8(c)]. However for smaller
grains [see f ≥ 1.5 µm−1 in Fig. 8(c)] the oscillations become harmonic. The Fourier transform of the oscillatory
component for the scattering signal at f =2.3 µm−1 shows
a single frequency centered at ν0 =6.1 THz [Fig. 9(a)],
while for lower spatial frequency f =1.0 µm−1 the Fourier
spectrum is more complex.
The strong dependence of ∆BSDF(t)/BSDF(0) oscillations versus spatial frequency is attributed to the influence of the misfit strain on thermodynamic potential Φ
of photoexcited VO2 . The detailed description of this potential will be given in Sec. IV.D. Different strain in VO2
grains/domains of different size and orientation alters Φ
FIG. 10. (a) Transient change ∆BSDF(t)/BSDF(0) of the
scattering signal at ϕ=250◦ versus spatial frequency of surface relief; F =6 mJ/cm2 . (b) Upper panel: the crosssections of the data mapped in (a) along the time axis
marked by dashed lines. Lower panel: the oscillatory part
of ∆BSDF(t)/BSDF(0) obtained by high-pass filtering of the
signal at f =2.46 µm−1 (circles). Solid line is the fit to equation A0 sin(2πνt + ϕ0 ) with ν=11.2 THz.
as shown in Fig. 9(b). As a result, the same level of photoexcitation switches these grains/domains into different
excited states. In consequence, nonequilibrium dynamics
of VO2 becomes grain-size-dependent.
It is important to note that at certain scattering angles we were able to resolve oscillations with nearly doubled frequency ∼2ν0 . Figure 10(a) shows TARHELS
data obtained at F =6 mJ/cm2 , at azimuthal direction
ϕ=250◦ , for the same VO2 /Al2 O3 (A-cut) film within
f =1.8–2.5 µm−1 . The observed oscillations are localized at certain spatial frequencies and represent nonequilibrium dynamics of certain groups of VO2 grains. The
cross-sections of Fig. 10(a) at different spatial frequencies
[upper panel of Fig. 10(b)] show noticeable oscillatory
behavior of the transient signal at f =2.12, 2.24, 2.264,
and 2.46 µm−1 . The lower panel of Fig. 10(b) shows
the oscillating part of the signal at f =2.46 µm−1 , which
10
FIG. 11. Ultrafast light scattering by VO2 /SiO2 at F =15 mJ/cm2 . (a) log(BSDF) indicatrix of the unperturbed sample. (b)
Relative change ∆BSDF(t)/BSDF(0) of the scattering signal upon light-induced SPT. Dashed circle separates two regions with
different nonlinearity. (c) BSDF and power spectral density (PSD) of the surface obtained by the cross-sections of scattering
indicatrix at ϕ=0◦ . PSD was calculated using Elson’s theory.87 (d) Same as (c) for f =2.3–2.5 µm−1 .
shown in Figs. 8 and 10, and is not related to coherent
excitation of optical phonons. As will be shown below,
this signal originates from growing optical inhomogeneity
in the film.
Taking into account the evolution of the dielectric constant during the IMT from εi ≃7.4+i5.5 to
εm ≃4.7+i5.46 for probe wavelength λ=400 nm,85 the
scattering cross-section of VO2 should decrease by ∼35%.
This estimation is consistent with ∼20% drop of the
scattering signal in Fig. 7(d). We note that the optical constants significantly depend on film morphology
and concentration of structural defects, and the scattering signal can decrease by a low percentage only. Nevertheless, at excitation levels well above the threshold
F0 , light scattering cannot increase for pure VO2 and
does not show any signature of ”transition opalescence”
on the sub-picosecond time scale, as shown in Refs. 82
and 86. However, Fig. 11(b) and the cross-sections
[Figs. 11(c), 11(d)] of the scattering indicatrix [Fig. 11(a)]
at different time delays show the rise of the scattering signal above f =2.35 µm−1 (θ =70◦ ). This behavior can be
interpreted as a consequence of increased local optical inhomogeneity in the film, and cannot be assigned to the
uniform change of dielectric constant as VO2 switches
from an insulator to metal.
The transient optical inhomogeneity can originate from
several major factors: from structural defects of vanadium dioxide and inhomogeneous nucleation of VO2 sites
or from film twinning and geometrical reconstruction of
the surface. As shown above [Figs. 7(c), 7(d)] and in
can be approximated by the function A0 sin(2πνt + ϕ0 )
with ν=11.2 THz. The obtained frequency of 11.2 THz
is nearly doubled frequency of the active Raman mode
shown in Fig. 9(a). We note that similar oscillations with
the frequency ∼2ν0 were previously resolved by multiterahertz spectroscopy33 (12 THz) and by transient reflectivity technique43 (10 THz).
C.
Sub-picosecond phase transition in the presence
of structural defects
The influence of structural inhomogeneities on transient optical properties was observed for 30 nm thick
non-epitaxial polycrystalline VO2 /SiO2 film. In contrast
to epitaxial VO2 /Al2 O3 (A-cut) film, for VO2 /SiO2 the
oscillations of scattering signal were not resolved. It is
very likely that the oscillatory signal is suppressed owing to fluctuation of the initial phase of lattice oscillation in different grains and domains, because of significant randomness in orientation and distribution of VO2
grains on the surface, as well as the relatively high concentration of structural defects in the non-epitaxial film.
At a laser excitation well above the transition threshold
(F =15 mJ/cm2 ) the scattering signal shows the monotonic rise at θ >70◦ (f >2.35 µm−1 ) [Figs. 11(a), 11(b)].
We note that a similar rise of the signal was observed also
for epitaxial VO2 /Al2 O3 (A-cut) film [Fig. 7(d)], however with much lower relative intensity. This behavior
is different from the dynamics at lower optical excitation
11
posterior evolution on picosecond time scale. While these
two components were observed previously in numerous
studies,31,43,50,90 most of the attention has centered so far
in the femtosecond response of VO2 . Both components of
NLO signal are related to light-induced phase transition
in VO2 grains, domains and/or clusters. We also found
that the slower (picosecond) component completely vanishes when the laser fluence exceeds ∼30 mJ/cm2 . This
result agrees with data obtained in Ref. 43.
Refs. 82 and 86, the ultrashort light pulse induces uniform SPT in all VO2 grains of the thin film and does not
modify the surface morphology on the picosecond time
scale. Therefore, it is proposed that the most probable
origin for the transient increase of scattering signal at
the relatively high optical excitation of F =15 mJ/cm2
is the presence of structural defects (presumably oxigen vacancies) in smaller grains with spatial frequencies
f >2.35 µm−1 .
The non-epitaxial growth of VO2 on SiO2 substrate
results in a quite disordered structure with numerous
oxygen vacancies and other point-like defects. The concentration of structural defects is higher in the smallest grains of the film,88 and it is very likely that the
grains with f >2.35 µm−1 also contain inclusions of VOx
(1 ≤ x ≤ 2.5) oxides,89 other than VO2 . Thus, VO2
would undergo the phase transition upon light illumination, while VOx would remain in the same phase or show
different nonlinear optical response as compared to VO2 .
As a result, the system disorder rapidly increases as the
VO2 grains with high concentration of structural defects
are switching into the metallic phase.
The MD calculations show that the presence of oxygen
vacancies does not noticeably alter the kinetics of SPT
[Fig. 1(a)]. This is consistent with experimental data
obtained for non-epitaxial VO2 /SiO2 film with higher
structural disorder [Fig. 11], as compared to epitaxial
VO2 /Al2 O3 [Fig. 7]. Thus, the complete phase transition in VO2 /SiO2 occurs within the same timescale of
∼500 fs, as in the case of VO2 /Al2 O3 . The phase transition rate within the full monitored range of spatial frequencies of the VO2 /SiO2 film was found to be nearly
same. However these films, in contrast to epitaxial films,
contain highly disordered structures with f > 2.35µm−1 ,
where the transient signal rises up and has significantly
increased noise [Fig. 11(d)]. It is very likely that this
noticeable noise component is a signature of alteration
of thermodynamic potential Φ and SPT trajectories by
structural defects in the smallest VO2 grains with embedded VOx sites.
D.
Picosecond lattice relaxation and
thermodynamic potential
FIG. 12. (a) Transient change of VO2 /Al2 O3 (C-cut) reflectivity within 40 ps at different levels of optical excitation.
Dashed curves are the exponential fit. Inset: 5×5 µm2 AFM
image of VO2 /Al2 O3 (C-cut) sample. The average lateral size
of grains is 250 nm and rms roughness is 14 nm. (b) The relaxation rate for picosecond component of photoinduced phase
transition versus laser pump fluence.
In this section, we discuss the nonequilibrium dynamics on a ∼40 ps time scale and perform quantitative reconstruction of the VO2 thermodynamic potential versus photoexcitation level. For this study we used an
epitaxial 50 nm VO2 /Al2 O3 (C-cut) film which absorbs
75% of incident radiation and is sufficiently thin for uniform photoexcitation. Figure 12(a) shows typical transient reflectivities of the film at three different excitation
levels: 4 mJ/cm2 , 6 mJ/cm2 and 27 mJ/cm2 . The inset shows an AFM image of the sample surface, which is
more uniform than the VO2 /SiO2 sample [Fig. 6(d)]. The
observed NLO response has two distinctive components:
instantaneous change of reflectivity within ∼500 fs and
Taking into account observed NLO dynamics at different levels of optical excitation, we assume that at relatively low laser fluence, much below 30 mJ/cm2 , not all
VO2 sites (i.e. domains, grains or clusters) are switched
into the complete metallic rutile phase within ∼500 fs.
However their number increases with the pump level.
As a result, the relative change of reflection [Fig. 12(a)]
12
(4) and (5) can be interpreted as a constant which is
proportional to the degrees of freedom and number of
ions involved in phase transition process.
or transmission [Figs. 6(a), 6(b)] also increases with the
pump fluence. The rest of the VO2 sites, which are not
switched by light within ∼500 fs are still in monoclinic
but nonequilibrium excited state. These sites can undergo the SPT on longer time scales, where the firstorder transition is mainly triggered by electron-phonon
and phonon-phonon scattering processes. These processes contribute into nucleation and growth of new rutile
phase.59
The slower (picosecond) nonequilibrium dynamics of
VO2 within ∼40 ps depends on laser excitation level.
However, we did not find a noticeable dependence of the
relaxation rate versus film thickness for different samples,
as well as we did not detect oscillations of optical signal
which could be assigned to acoustic phonon contribution
on this timescale. In Ref. 74, Brady et al. also found
that the phase transition dynamics within 40.5±2 ps time
scale is independent on sample morphology.
The structural transformation within several picoseconds involves electron-phonon and anharmonic phononphonon scattering processes which allow overcoming the
potential barrier ∆G(F ) between insulating (monoclinic)
and metallic (rutile) phases. Increasing pump fluence decreases the difference N ∆µ in binding energy for these
phases and, as a result, reduces the barrier. The relaxation rate for the picosecond component of SPT is determined by equation80
τ −1 = τ0−1 exp(−∆G(F )/kB T ),
(3)
where τ is the characteristic relaxation time and τ0 is the
shortest detectable relaxation time for this component of
the SPT.
In this study, we performed a series of pump-probe
measurements of VO2 dynamics at different levels of optical excitation and then derived the time τ by an exponential fit of experimental data. We have obtained the
relaxation times which compare well with the characteristic times obtained by Wall et al.43 for similar processes
in VO2 /SiO2 film. The rate τ −1 versus pump fluence is
shown in Fig. 12(b). These results can be well approximated by
e
τ −1 = AF N ,
FIG. 13. (a) Potential barrier ∆G(F ) between monoclinic
(M ) and rutile (R) phases derived from experimental data using Eq.(5). The inset shows the reconstructed thermodynamic
potential Φ as a function of effective ion displacement η for
the unperturbed (dashed line curve) and photoexcited (solid
line curves) VO2 . (b) The photoexcitation of VO2 at moderate fluence. The vertical dashed arrow 1 shows the initial
under-barrier energy state of the VO2 system right after photoexcitation. In this case, the interaction of the system with
an optical phonon of frequency ωph results in SPT (shown
by horizontal dashed arrow). The vertical dashed arrow 2
indicates the initial over-barrier photoexcited state of VO2 .
Here the additional energy is gained by the ion subsystem via
resonant excitation of Raman modes.
(4)
with fitting constants A=1.0×10−2 cm2 /(mJ·s) and
e =1.65. The slowest relaxation process, with the rate
N
τ −1 =2.9×1010 s−1 was still observed at pump fluence
F =1.4 mJ/cm2 , which is slightly below the threshold
F0 for the femtosecond component of I-M PT discussed
above. As the pump level increases to Fmax =30 mJ/cm2 ,
the slower (picosecond) component of NLO signal vanishes, providing an upper limit for the rate of SPT
τ0−1 =2.7×1012 s−1 . Equations (3) and (4) yield the dependence of the potential barrier on pump fluence
e kB T ln(F/Fmax ).
∆G(F ) = −N
Recently, the phenomenological Ginzburg-Landau
formalism91 was successfully applied to describe the
second-order photoinduced phase transitions in several
phase-change materials.92–95 In our study we extend this
concept to the first-order SPT in VO2 and consider a
thermodynamic potential Φ which depends on effective
ion displacement η as
Φ=
(5)
α(F ) 2 β 4 γ 6
η + η + η ,
2
4
6
(6)
where α(F ), β, and γ are experimentally derived constants, and α(F ) >0, β <0, γ >0. In order to describe the ultrafast structural dynamics of the VO2 lat-
Figure 13(a) shows the experimentally derived ∆G versus
e in
the level of optical excitation. The fitting constant N
13
system into an excited state above the potential barrier,
as indicated by the vertical dashed arrow 2 in Fig. 13(b).
In this case, ion motion overcomes the barrier ∆G(F ),
and the structural phase transition occurs rapidly within
∼500 fs.
If the optical pulse excites the VO2 system below the
potential barrier ∆G(F ) in the Φ diagram, the SPT cannot occur due to tunneling through the barrier because of
the relatively high mass of ion subsystem. However, additional energy of lattice vibrations, sufficient to overcome
barrier ∆G(F ), can be gained due to electron-phonon relaxation and due to anharmonic coupling of two or more
optical phonons. The rate of anharmonic optical phonon
scattering can be calculated using the equation obtained
by Klemens97
tice caused only by electronic excitations and by scattering of optical phonons within several picoseconds, the
contribution of transient photoacoustic stress was not included in the expression (6). Using the experimentally
derived potential Φ we avoid the necessity to know exact
details about the nature of photoexcited states and electronic mechanism of nearly instantaneous modification of
Φ by light with respect to the ground state.
Using the experimentally derived ∆G(F ) [Fig. 13(a)],
we performed the reconstruction of the thermodynamic
potential Φ (see Supplemental Material96 ). In the unperturbed state, it has two global minima at ±ηc which
correspond to two different domains. A VO2 microcrystal resides only in one of these domain states. The inset
in Fig. 13(a) shows the thermodynamic potential (6) as a
function of effective ion displacement η for three different
excitation levels. According to the Landau theory, the
phase transition threshold corresponds to the case when
minima of potential wells for different phases coincide.
In this study we consider the photoinduced screening
of electron-electron correlations as a near-instantaneous
process which occurs on a timescale comparable to or
less than the duration of the femtosecond light pulse.
Moreover, in our model we assume that only this process modifies the shape of the thermodynamic potential
Φ. Electron-phonon and phonon-phonon scattering processes can contribute only to the change of the energy of
the ion subsystem, without altering the potential Φ. In
this scenario, right after the illumination of the material
by a femtosecond pulse, the electronic state of VO2 is
changed by photoexcited free carriers and is characterized by a new thermodynamic potential Φ. However the
positions of atoms remain unchanged due to the relatively slow response of the lattice to femtosecond photoexcitation. The corresponding initial states of VO2
right after photoexcitation are marked on the Φ diagram
in Figs. 13(a) and 13(b) by dashed vertical arrows. Figure 13 shows that the minimum of the potential Φ of
photoexcited VO2 no longer coincides with the minimum
of the unperturbed ground state and, therefore, VO2 is
switched into a nonequilibrium state. The formation
of nonequilibrium state triggers the SPT. The quantitative modelling of subsequent structural dynamics will
be given in Sec. V.
Figure 13(b) shows that at relatively low optical excitation, slightly above the threshold F0 , the VO2 system
switches into an excited metastable state below the potential barrier ∆G(F ). We note that the presence of
the metastable state was directly observed previously
in experiments on ultrafast electron diffraction.51,54 In
this case, the SPT occurs during several picoseconds via
anharmonic scattering of optical phonons. However, a
sub-picosecond above-barrier pathway for the potential
in Fig. 13(b) is also possible if we take into account
the fact that the broadband femtosecond pulse instantaneously produces resonant Raman oscillations of the
VO2 lattice.32,33,39,49,64 These oscillations can provide
additional kinetic energy to ions, sufficient to switch the
−1
τas
=ω
J 2 ~ω a3 ω 3
γ
24π M v 2 v 3
(7)
where ω is the angular frequency of the phonon, M is
atomic mass, a is atomic size, v is the speed of sound, γ
is the Grüneisen parameter and ~ is Planck constant.
Parameter J ranges from 1 to 6 and corresponds to the
number of different phonon scattering processes. Taking into account only single longitudinal-to-longitudinal
phonon scattering, J=1. The Grüneisen parameter was
previously obtained only for rutile high-T metallic phase
of VO2 as γa = γb ≃ 2.0 and γc ≃ 4.5 for ar , br , and cr
crystallographic directions, correspondingly.98 In order
to calculate the phonon scattering rate, we used the averaged value γ=2.8, ω=38×1012 rad/s and v=4000 m/s.79
Since the Klemens theory considers monoatomic solid
such as Si, for the case of VO2 we used the averaged value of atomic mass M =4.6×10−26 kg and
atomic size a=1.9 Å. From (7) one obtains the rate
−1
τas
=1.3×1011 s−1 . This value belongs to the range of
experimentally obtained relaxation rates for the phase
transition process [Fig. 12(b)] and, therefore, provides
strong support for the proposed model of light-induced
structural phase transition in VO2 where anharmonic decay of optical phonons contributes to the picosecond SPT
dynamics.
V. PHENOMENOLOGICAL MODEL OF
ULTRAFAST STRUCTURAL DYNAMICS
The ultrafast solid-to-solid structural phase transition
can be well described in terms of the phenomenological Ginzburg-Landau theory.91–95 The oscillatory dynamics of VO2 during the light-induced SPT is essentially nonlinear: it depends on pump fluence, film crystallinity, internal strain and size of VO2 grains and clusters. The metastability of phase-change material can
be described by the thermodynamic potential Φ (Eq.6).
Previously, it was shown experimentally that the thermodynamic potential of VO2 includes at least two lattice distortions.50–53 However, in order to avoid possible ambiguities in the present study, we consider only a
14
single generalized lattice distortion associated with lattice transformation from monoclinic to rutile phase. Figure 14 shows the reconstructed potential Φ for three different levels of optical excitation. These levels correspond
to the threshold laser fluence F0 =3 mJ/cm2 [Fig. 14(a)],
moderate excitation at F =10 mJ/cm2 [Fig. 14(b)], and
to excitation at F =30 mJ/cm2 when the potential well
of monoclinic phase in Φ diagram vanishes [Fig. 14(c)].
The main pathways of lattice relaxation from monoclinic to rutile symmetry can be found by solving the motion equation for effective ion displacement η. This displacement is accompanied by photoinduced phonon oscillations with resonance frequency
ν0 =6.1 THz (ω=38×1012 rad/s), as shown in Fig. 9(a)
and also reported in Refs. 32, 33, 39, 43, 49, 64, and 65.
In terms of the microscopic theory of dynamic processes
in structural phase transitions,99 the equation of motion
can be written as
m̃
∂2η
∂η
∂Φ
+L
=−
,
∂t2
∂t
∂η
As was experimentally observed in this work, the optical signal related to the fastest component of the phase
transition (i.e. rapid change of the optical signal within
∼500 fs) at threshold excitation level F0 =3 mJ/cm2 is
relatively small. However, as excitation increases, the
amplitude of this signal asymptotically increases and approaches some constant level [see Figs. 5 and 12(a)].
Above F ∼10–15 mJ/cm2 the signal nearly saturates,
but still follows by a minor component of slower (picosecond) evolution. The picosecond component of NLO signal vanishes above F =30 mJ/cm2 [see Fig. 12(b)]. The
modeling of ultrafast structural dynamics in Fig. 14 fully
supports the experimentally observed NLO dynamics of
VO2 .
According to the model, at threshold fluence
F0 =3 mJ/cm2 the optical pulse excites the system below the potential barrier ∆G which separates two phases
[Fig. 14(a)]. Such excitation returns the system back
to its monoclinic phase: the system does not overcome
the ∆G barrier and relaxes to the bottom of the potential well which corresponds to the excited (and also
slightly distorted ) metastable monoclinic phase. The
corresponding trajectory in the η η̇ diagram [Fig. 14(d)]
is spiral dashed curve 1. The frequency of dissipative
oscillations during this relaxation is nearly double the
frequency ν0 and is ∼12 THz [Fig. 14(g)]. It is important to note that very similar oscillatory dynamics
with 11.2 THz frequency was experimentally observed in
this study (Fig. 10). Also, 12 THz and 10 THz oscillations were observed, respectively, by Pashkin et al.33 and
by Wall et al.43 in the low-fluence regime. These facts
strongly support the correctness of the model proposed.
Different trajectories of SPT in Fig. 14 are defined
by different initial conditions during photoexcitation of
the matarial. An additional contribution to the ultrafast SPT can be produced by resonant Raman process.
Coherent Raman excitation of the lattice is a nearly instantaneous process which occurs within the timescale of
light interaction with the sample. This process can provide additional kinetic energy to the system, switching it
to the phase trajectories 2-5 in Figs. 14(d), 14(g). Trajectory 2 is a separatrix which corresponds to the case when
the system passes a saddle point and can be switched either to rutile or monoclinic phase. Trajectory 3 is the
transition to metallic rutile phase.
It is interesting to note the possibility of switching VO2
into another domain state of the monoclinic phase. Separatrix 4 and trajectory 5 in Figs. 14(d) and 14(g) show
the relaxation of VO2 into the second potential well of the
monoclinic phase, related to another domain state. Here
we only mention this possibility which could be potentially observed.100 However, this requires verification by
additional rigorous experimental studies of photoinduced
dynamics.
According to the model of SPT shown in Fig. 13,
the photoinduced screening of electron correlations is a
nearly instantaneous process which alters only the thermodynamic potential. If the coherent Raman process
(8)
where m̃ is effective mass of ion subsystem and L is the
kinetic coefficient. According to experimental data on
photoinduced coherent phonon oscillations obtained by
Wall et al. for the sub-picosecond time scale,43,65 this
coefficient increases with laser fluence. Substituting (6)
in (8), we find
∂2η
2g ∂η
+ 2
+ α̃η + β̃η 3 + γ̃η 5 = 0,
ω 2 ∂t2
ω ∂t
(9)
where g = L/2m̃ corresponds to the damping of the vibrational modes, α̃ = α/m̃ω 2 , β̃ = β/m̃ω 2 , γ̃ = γ/m̃ω 2 .
The phase trajectories on the η η̇ phase plane and transient evolution of η in Figs. 14(d)-14(i) show possible
pathways of photoinduced ultrafast lattice transformation within 1.5 ps obtained by numerical integration of
Eq.(9). To perform the calculations, the kinetic coefficients were estimated from previously obtained experimental data of photoinduced phase transition, yielding g=3.3×1012 s−1 (F0 =3 mJ/cm2 ), g=5.0×1012 s−1
(F =10 mJ/cm2 ) and g=5.2×1012 s−1 (F =30 mJ/cm2 ).
We note that these coefficients are sufficiently close to
the kinetic coefficients which can be derived from damping ratio constants obtained in Ref. 43. The effective mass m̃ was estimated at the threshold fluence
F0 as m̃=1.7×10−25 kg using experimentally obtained
constants for the potential Φ(F0 ) (see Supplemental
Material96 ).
As shown above, the characteristic time of SPT and,
as a result, the kinetic coefficient depends on grain size.
A presence of structural defects, local deformations and
inhomogeneous strains between domain boundaries or
neighboring grains alters the profile of thermodynamic
potential Φ [Fig. 9(b)]. These factors, and also the resonant excitation of coherent optical phonons, can contribute into different pathways of the SPT dynamics upon
optical excitation, as shown in Figs. 14(d)-14(i).
15
FIG. 14. Thermodynamic potential Φ, phase trajectories on η η̇ phase plane and evolution of η(t) at different levels of optical
excitation: at threshold laser fluence F0 =3 mJ/cm2 (a, d, g), at F =10 mJ/cm2 (b, e, h) and at F =30 mJ/cm2 (c, f, i). Vertical
dashed arrows on Φ diagrams show the initial energy of VO2 system right after photoexcitation at η(0)=ηc .
is excluded from the consideration, the photoinduced
screening does not change directly the position of ions,
and does not provide additional kinetic energy and momentum to the ion subsystem. In this case the most probable trajectory will correspond to the trajectory with initial parameters η(0)=ηc and η̇(0)=0. Figure 14(e) shows
that this is a separatix (trajectory 2) for moderate excitation F =10 mJ/cm2 . That is, above F =10 mJ/cm2 the
majority of phase trajectories lead to complete structural
transition. This dynamics corresponds to sub-picosecond
SPT of mostly all volume of the material. This is supported by our experimental observations of fastest (subpicosecond) component of SPT dynamics. It was found
that the transient reflectance within ∼1 ps time scale
nearly saturates as pump fluence approaches F ∼10–
15 mJ/cm2 [see Figs. 5 and 12(a)]. Rigorous measurements of transient reflectivity versus excitation fluence
performed by Wall et al. in Ref. 65 shows very similar saturation of the transient reflectivity signal as the
fluence approaches F ∼20 mJ/cm2 . We also note that
there is still some possibility of relaxation back to monoclinic phase (trajectory 1) [Figs. 14(e) and 14(h)]. In this
case, the SPT occurs via optical phonon scattering during several picoseconds, as discussed above. This slower
relaxation is observed up to F =30 mJ/cm2 [Fig. 12(b)].
The proposed model describes the ultrafast structural
dynamics within several picoseconds after photoexcitation and agrees with numerous experimental observations
of the SPT in VO2 . In order to extend this model on
nanosecond time scale, the model has to include additional modulation of thermodynamic potential by acoustic strain as well as a growth of phonon entropy.101
VI.
CONCLUSION
It was demonstrated that the photoexcitation of VO2
enables various pathways and possibilities for transformation from its monoclinic to rutile symmetry. Resulting
dynamics strongly depends on excitation level, film morphology, internal strain and strain induced by the optical
pulse.
Semi-classical computation of molecular dynamics for
VO2 reveals significant instability of the monoclinic
phase in the absence of electron-electron correlations.
The computed dynamics of the VO2 lattice shows its
close resemblance to the experimentally observed transient NLO response of VO2 on the sub-picosecond time
scale. The thermodynamic parameters obtained by MD
method show that the screening of electron correlations
results in an exothermic reaction with saturation of the
phonon spectrum at initial temperatures of VO2 as low
as T =15 K. Also, calculations show the relatively small
influence of structural point defects on kinetics of SPT.
This was supported by experimental study of photoinduced dynamics of non-epitaxial films with relatively
high concentration of defects. Nevertheless, a presence
of structural defects produces nonuniform metallic phase
At excitation level of F =30 mJ/cm2 the system dynamics undergoes a qualitative change: the potential well of monoclinic phase and barrier ∆G vanishes
[Fig. 14(c)]. As a result, all possible phase trajectories
change the symmetry of VO2 from monoclinic to rutile
[Figs. 14(f), 14(i)]. No slow (picosecond) relaxation component was observed in NLO signal above F =30 mJ/cm2 ,
since there is no pathway to metastable monoclinic phase.
16
quantitative analysis and numerical modeling of photoinduced dynamics. We show that the modeling of ultrafast
processes in VO2 can be performed in terms of a phenomenological Ginzburg-Landau model. This model provides a reliable explanation of experimentally observed
structural dynamics, where the phase trajectories depend on excitation level. Thus, higher optical excitation
above ∼30 mJ/cm2 corresponds to complete structural
transition of all VO2 grains/domains into metallic rutile phase within ∼500 fs. However, at lower excitations,
grains/domains can be turned to long-lived metastable
monoclinic phase. The presented approach to model the
photoinduced structural dynamics offers unique potential
for the study of different phase-change materials.
nucleation on the sub-picosecond time scale.
On a few-picosecond time scale the structural phase
transition can be considered as a non-thermal process.
However, on the nanosecond time scale the thermal contribution to the transition becomes essential. It was also
shown that photoacoustic stress can potentially induce a
ferroelastic phase transition. Thus, the pronounced oscillatory signal which could be associated with alternating
switching of VO2 phase by a photoacoustic wave was observed at frequency 0.8 GHz [Fig. 6(e)].
It was shown that the internal misfit strain in epitaxial
film noticeably alters the rate of phase transition within
∼500 fs. For epitaxial films with anisotropic strain the
phase transition rate depends on the size and in-plane orientation of VO2 grains/domains. This evidences that the
strain alters the potential energy landscape of photoexcited VO2 and, as a result, changes the phase trajectory
of ultrafast structural dynamics.
In this work, we proposed a technique for a quantitative reconstruction of the thermodynamic potential
of photoexcited phase-change material. While it is
rather an estimation of the energy landscape versus
pump fluence, the obtained energy can be used for semi-
∗
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
ACKNOWLEDGEMENTS
We wish to thank Dr. Andrey Akimov for valuable discussions and comments. The authors gratefully acknowledge support from the U. S. Army Research Laboratory
and the U. S. Army Research Office under contract number W911NF-15-1-0448.
18
sergiy.lysenko@upr.edu
F. J. Morin, Phys. Rev. Lett. 3, 34 (1959).
K. Okimura, J. Sakai, S. Ramanathan, J. Appl. Phys.
107, 063503 (2010).
J. D. Budai, A. Tselev, J. Z. Tischler, E. Strelcov, A. Kolmakov, W. J. Liu, A. Gupta, J. Narayan, Acta Mater. 61,
2751 (2013).
G. Andersson, Acta Chem. Scand. 10, 623 (1956).
J. M. Longo and P. Kierkegaard, Acta Chem. Scand. 24,
420 (1970).
S. Westman, Acta Chem. Scand. 15, 217 (1961).
M. Liu, A. J. Sternbach, M. Wagner, T. V. Slusar,
T. Kong, S. L. Bud’ko, S. Kittiwatanakul, M. Qazilbash,
A. McLeod, Z. Fei, et al., Phys. Rev. B 91, 245155 (2015).
J. H. Park, J. M. Coy, T. S. Kasirga, C. Huang, Z. Fei,
S. Hunter, and D. H. Cobden, Nature 500, 431 (2013).
J. Pouget, H. Launois, T. Rice, P. Dernier, A. Gossard,
G. Villeneuve, and P. Hagenmuller, Phys. Rev. B 10, 1801
(1974).
A. Zylbersztejn and N. F. Mott, Phys. Rev. B 11, 4383–
4395 (1975).
H. Nakatsugawa and E. Iguchi, Phys. Rev. B 55, 2157
(1997).
M. S. Laad, L. Craco, and E. Müller-Hartmann, Phys.
Rev. B 73, 195120 (2006).
V. Eyert, Ann. Phys. (Leipzig) 11, 650 (2002).
V. Eyert, Phys. Rev. Lett. 107, 016401 (2011).
J. B. Goodenough, J. Solid State Chem. 3, 490 (1971).
D. Paquet and P. Leroux-Hugon, Phys. Rev. B 22, 5284
(1980).
T. Koethe, Z. Hu, M. Haverkort, C. Schüßler-Langeheine,
F. Venturini, N. Brookes, O. Tjernberg, W. Reichelt,
H. Hsieh, H.-J. Lin, et al., Phys. Rev. Lett. 97, 116402
(2006).
19
20
21
22
23
24
25
26
27
28
29
30
17
M. Gatti, F. Bruneval, V. Olevano, and L. Reining, Phys.
Rev. Lett. 99, 266402 (2007).
J. Cao, E. Ertekin, V. Srinivasan, W. Fan, S. Huang,
H. Zheng, J. W. L. Yim, D. R. Khanal, D. F. Ogletree,
J. C. Grossman, et al., Nat. Nanotechnol. 4, 732 (2009),
0907.4769.
A. Tselev, V. Meunier, E. Strelcov, W. A. Shelton, I. A.
Lukyanchuk, K. Jones, R. Proksch, A. Kolmakov, and
S. V. Kalinin, ACS Nano 4, 4412 (2010).
J. P. Pouget, H. Launois, J. d’Haenens, P. Merenda, and
T. M. Rice, Phys. Rev. Lett. 35, 873 (1975).
M. Liu, M. Wagner, J. Zhang, A. McLeod, S. Kittiwatanakul, Z. Fei, E. Abreu, M. Goldflam, A. J. Sternbach, S. Dai, et al., Appl. Phys. Lett. 104, 121905 (2014).
A. Pergament, P. Boriskov, A. Velichko, and N. Kuldin,
J. Phys. Chem. Solids 71, 874 (2010).
H.-T. Kim, B.-J. Kim, Y. W. Lee, B.-G. Chae, and S. J.
Yun, Phys. B 403, 1434 (2008).
A. Crunteanu, J. Givernaud, J. Leroy, D. Mardivirin,
C. Champeaux, J.-C. Orlianges, A. Catherinot, and
P. Blondy, Sci. Tech. Adv. Mater. 11, 065002 (2010).
Y. Jiazhen, Z. Yue, H. Wanxia, and T. Mingjin, Thin
Solid Films 516, 8554 (2008).
W. Burkhardt, T. Christmann, S. Franke, W. Kriegseis, D. Meister, B. Meyer, W. Niessner, D. Schalch, and
A. Scharmann, Thin Solid Films 402, 226 (2002).
B. Mayer, C. Schmidt, A. Grupp, J. Bühler, J. Oelmann,
R. E. Marvel, R. Haglund Jr, T. Oka, D. Brida, A. Leitenstorfer, et al., Phys. Rev. B 91, 235113 (2015).
K. Appavoo, D. Y. Lei, Y. Sonnefraud, B. Wang, S. T.
Pantelides, S. A. Maier, and R. F. Haglund, Nano Lett.
12, 780 (2012).
W. Roach and I. Balberg, Solid State Communications 9,
551 (1971).
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
T. L. Cocker, L. V. Titova, S. Fourmaux, G. Holloway, H.C. Bandulet, D. Brassard, J.-C. Kieffer, M. A. El Khakani,
and F. A. Hegmann, Phys. Rev. B 85, 155120 (2012).
C. Kübler, H. Ehrke, R. Huber, R. Lopez, A. Halabica,
R. F. Haglund, and A. Leitenstorfer, Phys. Rev. Lett. 99,
116401 (2007).
A. Pashkin, C. Kübler, H. Ehrke, R. Lopez, A. Halabica,
R. F. Haglund, R. Huber, and A. Leitenstorfer, Phys. Rev.
B 83, 195120 (2011).
D. Wegkamp, J. Stähler, Progr. Surf. Sci. 90,464 (2015).
Z. He and A. J. Millis, Phys. Rev. B 93, 115126 (2016).
V. Vikhnin, S. Lysenko, A. Rua, F. Fernandez, and H. Liu,
Solid State Commun. 137, 615 (2006).
M. van Veenendaal, Phys. Rev. B 87, 235118 (2013).
D.-S. Yang, P. Baum, and A. H. Zewail, Struct. Dyn. 3,
034304 (2016).
A. Cavalleri, T. Dekorsy, H. H. W. Chong, J. C. Kieffer,
and R. W. Schoenlein, Phys. Rev. B 70, 161102 (2004).
M. Hada, D. Zhang, A. Casandruc, R. J. Dwayne Miller,
Y. Hontani, J. Matsuo, R. E. Marvel, and R. F. Haglund,
Phys. Rev. B 86, 134101 (2012).
H. Wen, L. Guo, E. Barnes, J. H. Lee, D. A. Walko, R. D.
Schaller, J. A. Moyer, R. Misra, Y. Li, E. M. Dufresne,
et al., Phys. Rev. B 88, 165424 (2013).
S. Lysenko, V. Vikhnin, F. Fernandez, A. Rua, and H. Liu,
Phys. Rev. B 75, 075109 (2007).
S. Wall, L. Foglia, D. Wegkamp, K. Appavoo, J. Nag,
R. F. Haglund, J. Stähler, and M. Wolf, Phys. Rev. B 87,
115126 (2013).
C. Chen, Y. Zhao, X. Pan, V. Kuryatkov, A. Bernussi,
M. Holtz, and Z. Fan, J. Appl. Phys. 110, 023707 (2011),
M. A. Huber, M. Plankl, M. Eisele, R. E. Marvel, F. Sandner, T. Korn, C. Schuller, R. Haglund Jr, R. Huber, and
T. L. Cocker, Nano Lett. 16, 1421 (2016).
D. Wegkamp, M. Herzog, L. Xian, M. Gatti, P. Cudazzo,
C. L. McGahan, R. E. Marvel, R. F. Haglund Jr, A. Rubio, M. Wolf, et al., Phys. Rev. Lett. 113, 216401 (2014).
C. Ott, M. F. Jager, C. J. Kaplan, R. E. Marvel,
R. F. Haglund, D. M. Neumark, S. R. Leone, International
Conference on Ultrafast Phenomena, UTu4A–48 (2016).
N. F. Mott, Rev. Mod. Phys. 40, 677 (1968).
B. T. O’Callahan,Callahan, A. C. Jones, J. H. Park, D. H.
Cobden, J. M. Atkin, and M. B. Raschke, Nat. Commun.
6, art# 6849 (2015).
P. Baum, D.-S. Yang, and A. H. Zewail, Science 318, 788
(2007).
Z. Tao, F. Zhou, T.-R. T. Han, D. Torres, T. Wang,
N. Sepulveda, K. Chang, M. Young, R. R. Lunt, and C.-Y.
Ruan, Sci. Rep. 6, 38514 (2016).
A. X. Gray, M. C. Hoffmann, J. Jeong, N. P. Aetukuri,
D. Zhu, H. Y. Hwang, N. C. Brandt, H. Wen, A. J. Sternbach, S. Bonetti, et. al., arXiv:1601.07490.
N. Kumar, A. Rua, F. E. Fernandez, S. Lysenko, Phys.
Rev. B 95, 235157 (2017).
V. R. Morrison, R. P. Chatelain, K. L. Tiwari, A. Hendaoui, A. Bruhács, M. Chaker, B. J. Siwick, Science, 346,
445 (2014).
Atomistix ToolKit, QuantumWise A/S, (2014).
A. C. T. van Duin, S. Dasgupta, F. Lorant and W. A.
Goddard III, J. Phys. Chem. A 105, 9396 (2001).
K. Chenoweth, A. C. Van Duin, and W. A. Goddard III,
J. Phys. Chem. A 112, 1040 (2008).
M. Griebel, S. Knapek, G. Zumbusch, Numerical Simulation in Molecular Dynamics: Numerics, Algorithms, Par-
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
18
allelization, Applications (Springer-Verlag, Berlin, Heidelberg, 2007).
S. Lysenko, V. Vikhnin, A. Rúa, F. Fernández, and H. Liu,
Phys. Rev. B 82, 205425 (2010).
H. J. G. J. Ryder, F. S. L. Hsu and J. E. Kunzler, Phys.
Rev. 185, 1022 (1969).
G. V. Chandrashekhar, H. Barros, and J. Honig, Mater.
Res. Bull. 8, 369 (1973).
J. M. Ziman, Electrons and phonons: the theory of transport phenomena in solids. (Oxford university press, New
York, 1960).
S.-T. Lin, M. Blanco and W. A. Goddard III, J. Chem.
Phys. 119, 11792 (2003).
H.-T. Kim, Y. W. Lee, B.-J. Kim, B.-G. Chae, S. J. Yun,
K.-Y. Kang, K.-J. Han, K.-J. Yee, and Y.-S. Lim, Phys.
Rev. Lett. 97, 266401 (2006).
S. Wall, D. Wegkamp, L. Foglia, K. Appavoo, J. Nag,
R. Haglund Jr, J. Stähler, and M. Wolf, Nat. Commun.
3, 721 (2012).
M. Borek, F. Qian, V. Nagabushnam, and R. K. Singh,
Appl. Phys. Lett. 63, 3288 (1993).
F. C. Case, J. Vac. Sci. Technol. A 2, 1509 (1984).
J. F. De Natale, P. J. Hood, A. B. Harker, J. Appl. Phys.
66, 5844 (1989).
Z. P. Wu, A. Miyashita, I. Nashiyama, H. Naramoto, Philos. Mag. Lett. 79, 813 (1999).
S. Lysenko, A. Rúa, V. Vikhnin, F. Fernández, and H. Liu,
Phys. Rev. B 76, 035104 (2007).
S. Lysenko, A. Rua, F. Fernandez, and H. Liu, J. Appl.
Phys. 105, 043502 (2009).
V. Sterligov, P. Cheyssac, French patent no. 0115232
(2001).
J. C. Stover, Optical Scattering: Measurements and Analysis (SPIE Optical Engineering Press, Bellingham, Washington, 1995).
N. F. Brady, K. Appavoo, M. Seo, J. Nag, R. P. Prasankumar, R. F. Jr. Haglund, D. J. Hilton, J. Phys. Condens.
Matter 28, 125603 (2016).
D. J. Hilton, R. P. Prasankumar, S. Fourmaux, A. Cavalleri, D. Brassard, M. A. El Khakani, J. C. Kieffer,
A. J. Taylor, and R. D. Averitt, Phys. Rev. Lett. 99,
226401 (2007).
A. C. Diebold, Handbook of Silicon Semiconductor Metrology (CRC Press, New York, Basel, 2001).
C. Thomsen, J. Strait, Z. Vardeny, H. J. Maris, J. Tauc,
J. J. Hauser, Phys. Rev. Lett. 53, 989 (1984).
C. Thomsen, H. T. Grahn, H. J. Maris, J. Tauc, Phys.
Rev. B 34, 4129 (1986).
D. Maurer, A. Leue, R. Heichele, and V. Müller, Phys.
Rev. B 60, 13249 (1999).
K. H. Bennemann, J. Phys. Condens. Matter 23, 073202
(2011).
E. Abreu, S. N. G. Corder, S. J. Yun, S. Wang, J. G.
Ramirez, K. West, J. Zhang, S. Kittiwatanakul, I. K.
Schuller, J. Lu, and S. A. Wolf, H.-T. Kim, M. Liu, R. D.
Averitt, arXiv:1701.05531.
S. Lysenko, F. Fernández, A. Rúa, N. Sepúlveda, and
J. Aparicio, Appl. Opt. 54, 2141 (2015).
S. Lysenko, A. Rua, and F. Fernández, International Conference on Ultrafast Phenomena, UTu4A–48 (2016).
E. R. Dobrovinskaya, L. A. Lytvynov, V. Pishchik, Sapphire: material, manufacturing, applications (Springer
Science & Business Media, New York, 2009).
85
86
87
88
89
90
91
92
93
94
H. W. Verleur, A. S. Barker, and C. N. Berglund, Phys.
Rev. 172, 788 (1968).
S. Lysenko, F. Fernandez, A. Rua, and H. Liu, J. Appl.
Phys. 114, 153514 (2013).
J. M. Elson, Waves Random Media 7, 303 (1997).
J. Y. Suh, R. Lopez, L. C. Feldman and R. F. Haglund,
J. Appl. Phys. 96, 1209 (2004).
H. Katzke, P. Toledano, and W. Depmeier, Phys. Rev. B
68, 024109 (2003).
K. Appavoo, B. Wang, N. F. Brady, M. Seo, J. Nag, R. P.
Prasankumar, D. J. Hilton, S. T. Pantelides, and R. F.
Haglund, Nano Lett. 14, 1127 (2014).
Landau and Lifshitz, Statistical Physics, Part 1 (Elsevier,
Butterworth Heinemann, 1980).
C. Giannetti, M. Capone, D. Fausti, M. Fabrizio,
F. Parmigiani, and D. Mihailovic, Adv. Phys. 65, 58
(2016).
R. Yusupov, T. Mertelj, V. V. Kabanov, S. Brazovskii,
P. Kusar, J.-H. Chu, I. R. Fisher, and D. Mihailovic, Nature Phys. 6, 681 (2010).
T. Huber, S. O. Mariager, A. Ferrer, H. Schäfer, J. A.
Johnson, S. Grübel, A. Lübcke, L. Huber, T. Kubacka,
95
96
97
98
99
100
101
19
C. Dornes, et al., Phys. Rev. Lett. 113, 026401 (2014).
P. Beaud, A. Caviezel, S. Mariager, L. Rettig, G. Ingold,
C. Dornes, S. Huang, J. Johnson, M. Radovic, T. Huber,
et al., Nat. Mater. 13, 923 (2014).
See
Supplemental
Material
at
http://link.aps.org/supplemental/
10.1103/PhysRevB.xx.xxxxxx for reconstruction of the thermodynamic
potential of photoexcited VO2 .
P. G. Klemens, Phys. Rev. 148, 845 (1966).
C. Hearn, Solid State Commun. 13, 1139 (1973).
B. A. Strukov and A. P. Levanyuk, Ferroelectric Phenomena in Crystals: Physical Foundations (Berlin, Heidelberg, 1998).
S. A. Dönges, O. Khatib, B. T. O’Callahan, J. M. Atkin,
J. H. Park, D. Cobden, M. B. Raschke, Nano Lett. 16,
3029 (2016).
J. D. Budai, J. Hong, M. E. Manley, E. D. Specht,
C. W. Li, J. Z. Tischler, D. L. Abernathy, A. H. Said,
B. M. Leu, L. A. Boatner, R. J. McQueeney, O. Delaire,
Nature 515, 535 (2014).