U
NIVERSITEIT VAN
A
MSTERDAM
X-ray Spectroscopy and Electron
Diffraction on VO
2
Author:
Steef Smit
Supervisor:
Prof. M.S. Golden
Quantum Matter Group Van der Waals-Zeeman Institute - IOP
iii
Universiteit van Amsterdam
Abstract
FNWI
Van der Waals-Zeeman Institute - IOP
Master of Science
X-ray Spectroscopy and Electron Diffraction on VO2 by Steef Smit
The ability to understand and control the conductivity of a material has been a great goal of physicists and engineers for decades. This would pave the road to energy efficient electronic devices that could cause a revolution in the electronics industry. One of the most promising materials in this respect is vanadium dioxide, a correlated electron material with a metal-insulator transition (MIT) just above room tempera-ture. Not only is it a promising material for future appliances, it is rich in beautiful physics not yet fully understood despite decades of work by many different people. Especially in the ultrafast (<ps) regime, the MIT has not been investigated to the full extent. Due to great developments and increasing availability of free electron lasers capable of measuring on these timescales, investigating this side of transition is be-coming increasingly appealing. In this work, a novel sample geometry of VO2 is investigated using a broad range of experimental techniques. For the investigation of the electronic structure techniques such as static and ultrafast X-ray absorption were used. Orbital ordering was investigated with polarisation dependent X-ray ab-sorption, and the available electronic states were tracked around the transition tem-perature. These experiments were carried out at large scale X-ray facilities around the world, such as the BESSY synchrotron in Berlin, and the LCLS Free Electron Lasers at SLAC. The structural aspects of the MIT were investigated using the ASTA electron diffraction facility at SLAC. The experiments described in this thesis form the basis for future ultrafast experiments and lensless imaging studies, as well as attempts at purely electronic switching of the MIT using THz radiation.
v
Contents
Abstract iii 1 Introduction 1 1.1 Vanadium dioxide. . . 1 2 Experimental 5 2.1 Sample development . . . 5 2.2 X-ray techniques . . . 7 2.2.1 Synchrotron radiation . . . 72.2.2 X-ray Absorption Spectroscopy . . . 10
2.2.3 X-ray Linear Dichroism . . . 12
2.3 Electron based techniques . . . 13
2.3.1 Ultrafast Electron Diffraction . . . 13
2.4 Free Electron Laser . . . 15
2.4.1 XAS at FEL . . . 16
3 Background and Theory 19 3.1 VO2 . . . 19
3.1.1 Structural properties . . . 19
3.1.2 Electronic aspects of the MIT . . . 20
3.1.3 XAS spectra of VO2 . . . 22
3.1.4 Electron Diffraction. . . 23
3.1.5 The (Ultrafast) Insulator to Metal Transition in VO2 . . . 25
4 Results 27 4.1 Soft X-ray Absorption Spectroscopy at the Synchrotron . . . 27
4.2 Ultrafast electron diffraction . . . 32
4.2.1 Simulation . . . 34
4.3 Soft X-ray absorption spectroscopy at LCLS . . . 35
5 Concluding remarks and Outlook 39 5.1 Concluding remarks . . . 39
5.2 Outlook: two paths forward . . . 40
5.2.1 Spatially resolving the MIT . . . 40
A Data treatment and simulations 43
A.1 Normalisation to the beam and the detector.. . . 43
A.1.1 XAS in Transmission . . . 44
A.1.2 Saturation . . . 44
A.1.3 Electron diffraction . . . 46
A.2 Simulating electron diffraction data. . . 47
1
Chapter 1
Introduction
Transition metal oxides, and correlated electron materials in general, have been a major topic of investigation for solid state physicists around the world for many decades. They earned the attention due their display of fascinating properties such as Metal-Insulator Transitions (MIT) and related phenomena like colossal magne-toresistance, high Tc superconductivity, and the formation of 2D electron gasses at heterointerfaces of insulating oxides.
FIGURE 1.1: TM-oxides have a lot of degrees of freedom that help determine their physical properties. Spin, orbital momentum, their charge distribution and lattice degrees of freedom are closely cou-pled in these systems, and their interaction energy is all of the same order of magnitude as their kinetic energy. External parameters such as temperature, strain or electric fields can be used to to tune macro-scopic properties such as crystal structure and electrical resistance.
[1]
Intriguing as these systems are, a lot is still unknown about the fundamental physical processes that cause their odd behaviour. The research in this thesis aims to gain more understanding about one of the touchstone systems in the field of cor-related electrons, vanadium dioxide
1.1
Vanadium dioxide
Vanadium dioxide and its Metal-Insulator Transition are one of the go-to systems for research on correlated electron systems. In conventional band theory, systems with partially filled outer electron bands such as is the case in VO2, should show
FIGURE 1.2: Impression of a Mott-FET. The active layer where the conductivity can be controlled by an external parameter such as temperature or electric field, is imagined to be a material ex-hibiting Mott-MIT, such as VO2. Picture taken from
http://www-ssrl.slac.stanford.edu/stohrgroup/research/research.html
metallic behaviour. At low temperatures however, the resistance of VO2 increases by 4-5 orders of magnitude and the compound starts to behave like an insulator. This effect, due to the opening of an insulating gap is a result of complex electron-electron interactions first explained by Nevill Mott, and was later summarized in his book [2], which could be seen as the start of this field in correlated electron systems, something carried out together with Rudolph Peierls. The main reason that VO2 in particular is so interesting compared to similar systems, lies in the fact that its MIT occurs relatively close to room temperature, around 350 Kelvin, and is even tuneable by straining the crystal [3]. This makes it an ideal candidate for future applications in for example information technology, where one can think of applications in the form of sensors, or low energy switching systems that could be used as the active material in small and efficient Field Effect Transistors, so called Mott-FETs, an example of which is shown in figure1.2.
While this is a great goal to work towards in the distant future, this thesis re-search project is more concerned with uncovering the fundamental physics behind the MIT occurring in VO2. To explore, and ultimately control the complicated elec-tronic, spin and lattice dynamics occurring is absolutely necessary before one can start thinking about technical feasibility of electronic devices as such.
It has been known for a long time that the electronic transition is accompanied by a structural change in the crystal. Measuring both the electronic and structural changes across the MIT requires a wide variety of experimental techniques, and thus this thesis reflects that. The dynamics of the transition, especially the coupling -or
1.1. Vanadium dioxide 3
more exciting still, the possibility of decoupling- between the structural and elec-tronical transition, is something that needs further investigation. Recent studies sug-gest that to a certain degree the structural and electronical transition can be decou-pled and an intermediate, metastable state exists where the material can be switched from insulating to metallic, without immediately inducing the structural change [4]. Both X-ray based experiments such as X-ray Absorption Spectroscopy (XAS) at large synchrotron radiation facilities and Free Electron Lasers, and electron scat-tering techniques like Transmission Electron Microscopy (TEM) and Ultrafast Elec-tron diffraction (UED) have been used to characterize and measure samples forming the subject of this thesis, which were made by collaborators in Twente. The ultimate goal of this part of the complex oxide FOM programme in which this masters project is embedded, is to unravel the transition in its completeness both temporally- and spatially resolved. With the recent development of extremely powerful free electron X-ray laser facilities that can achieve extremely good beam coherence, beam bright-ness AND temporal resolution, these type of experiments are becoming feasible.
5
Chapter 2
Experimental
2.1
Sample development
A significant development in sample growing technology by our collaborators from the University of Twente, was part of the reason for starting this line of research. Up until recently, it has not been possible to get high quality, single crystalline samples of complex oxides that are suitable for transmission geometry experiments. Free-standing films of epitaxially grown VO2 were not yet in the realm of the possible. To be able to do X-ray or electron transmission experiments, the samples have to be grown on a membrane that does not interfere with the probing beams too much. Often amorphous Si3N4 is used for this purpose. The innovation applied by our Twente collaborators is to get the VO2 to grow epitaxially on Si3N4, by the use of so called-nanosheets. A schematic layout of the sample structure can be seen in figure
2.1.
Nanosheets are extremely thin crystalline films that can be deposited on amor-phous substrates using a Langmuir-Blodget process [5]. These nanosheets, when chosen to have the correct lattice parameters, can act as a crystalline substrate al-lowing epitaxial growth of the sample, while still being thin enough to not interfere with the probing electron or photon beams when performing experiments in trans-mission geometry. After nanosheet deposition, the VO2is grown using Pulsed Laser Deposition with a KrF laser (248 nm wavelength). The VO2samples are grown using a V2O5 target, and a laser fluence of 1.3 J/cm2. The substrate temperature is kept at 793 K, under a partial oxygen pressure of 7.5 mTorr1. Atomic Force Microscopy
1All samples were created in the MESA+ laboratory in Twente, by Abhi Rana and Kevin Hofhuis.
(A) nanosheets on amorphous SiN (B) VO2on nanosheets on SiN FIGURE 2.2: Atomic Force Microscopy (AFM) images of the nanosheets with (right) and without (left) VO2 on top. The single nanosheets can be seen to be on the order of square micrometers big in the lateral dimension. In the rutile phase the VO2 is epitaxial on each nanosheet. AFM images courtesy of Kevin Hofhuis (MESA+
Twente)
images of the VO2thin films grown in this manner, are shown in Fig.2.2.
The choice of which nanosheet system is used also determines the direction of growth of the VO2 samples. For the majority of the samples used in this work the nanosheets used were made of T iOxwhere x ≈ 2. These nanosheets cause the VO2 to always grow with the rutile crystallograpic c-axis in plane of the film. The az-imuthal orientation of the individual nanosheet crystallites is not controled during nanosheet deposition. This means that the resulting VO2 sample has a mosaic like pattern of grains, all with rutile c-axis in-plane but without a preferred orientation of the VO2rutile c-axis in the film plane. When cooling down, the VO2 monoclinic phase nucleates at multiple locations within each rutile single crystalline patch, ulti-mately causing monoclinic domains with sizes varying between a few 10’s to a few 100’s of nm. Fig. 2.3 shows TEM characterisation of VO2 on nanosheet films like those studied in this thesis. According to van Tendeloo et al [6], four different types of monoclinic domains have to form to satisfy the reduction of the symmetry in this rutile to monoclinic transition. Fig. 2.3shows TEM imaging of a VO2 film at room temperature, where the monoclinic phase of VO2 will dominate. The individual monoclinic domains can be seen to be of order 100nm, on average. In Fig. 2.3band
2.3ccrystal orientation maps are shown. These are the result of pixel by pixel elec-tron diffraction experiments, and show the orientation of the monoclinic and rutile phases. The rutile phase clearly grows with full epitaxy on the nanosheet in ques-tion, with the [110] direction out of the film plane. On cooling to room temperature (Fig. 2.3b), the monoclinic phase develops from the rutile one, and domains with four different orientations are shown on the map using four different (false) colours.
2.2. X-ray techniques 7
(A) (B) (C)
FIGURE 2.3: (A): Transmission Electron Microscopy image (room temperature) of the VO2 on T iOx nanosheets. The monoclinic do-main size is a few hundred nm on average. (B): Crystallite orientation map of the monoclinic VO2 phase (T=300K). Each of the four false colours represents a single orientation of the monoclinic domain. (C) Crystallite orientation map of the VO2 on a single nanosheet. The inset shows that the single domain rutile phase is [110] oriented. TEM data courtesy of Dr. Nicolas Gauquelin (EMAT, Universiteit
Antwerp)
2.2
X-ray techniques
For investigating the properties of materials at the atomic scale, such as their crys-tallographic or electronic structure, X-ray based techniques are very widely used. Since the main part of this thesis research consists of X-ray spectroscopy, the cre-ation, manipulation and detection of these X-rays is a very important subject. In this part of the thesis, the instrumentation and measurement techniques used will be summarized.
2.2.1 Synchrotron radiation
A lot of the data gathered in this research was taken at the UE56-PGM1 beamline at the BESSY II synchrotron radiation source located in the Helmholtz center in Berlin, Germany. A synchrotron is a circular electron storage ring. It uses magnetic fields to keep electrons orbiting in a ring-like vacuum chamber when under acceleration by means of radio frequency fields. This section is based upon [7].
The operating principle of any synchrotron radiation source relies on the fact that accelerating charged particles over a curved path leads to the emission of elec-tromagnetic radiation (or: synchrotron radiation). This was first observed in 1946 by Frank Elder et al. [9]. While at first this emssion was considered as a parasitic effect, preventing the acceleration of particles to the speed of light, since then it has been harnassed and used as a powerful source of high brightness and highly tunable radiation, that is used in fields ranging from physics to biology.
FIGURE 2.4: Schematic structure of modern synchrotron radiation source. The synchrotron radiation is created in the undulators or wig-glers that are placed in straight sections in the electron storage ring. The electons are first accelerated in a linear accelerator and boosted up to speed in a booster ring before being injected into the storage
ring. Picture from [8]
The higher the kinetic energy of the electrons in the ring, the more the radia-tion is colimated along the tangential direcradia-tion of their orbit, as can be seen in fig.
2.5. This is very useful for our experiments, as this means a highly focussed, high flux beam can be directed to our sample. The half-opening angle of the beam, ψ, is approximately given by [7]: ψ ≈ mc 2 E ≈ p 1 − β = 1 γ (2.1)
where m is the electron mass, c the speed of light and E its kinetic energy. γ is the Lorentz factor.
FIGURE 2.5: Accelerating a charged particle along a curved path causes synchrotron radiation. This radiation can be colimated in a tangential direction to the path of the particle if the kinetic energy is
very high. Adapted from [7]
Synchrotron radiation from the bending magnets of a storage ring usually con-tains a very broad spectrum of wavelengths, ranging from the infrared to hard X-rays. To get a specific photon-energy for an experiment, often an undulator in com-bination with a monochromator is used. An undulator is a so-called insertion device
2.2. X-ray techniques 9
that is placed inside the storage ring. It consists of a periodic structure of dipole magnets alternating in polarity, as shown schematically in Fig. 2.6a. The electrons passing through this magnetic structure are forced to oscillate, and therefore radiate. Given a magnetic structure with a sufficient number of periods, interference effects cause emission of synchrotron radiation in a series of energetically fairly narrow harmonics. The parameter that can be used to characterize the nature of the electron motion inside an undulator is the parameter K:
K = eBλu 2πmec
(2.2) with e electron charge, B the magnetic field, λu the period of the undulator, me the mass of the electron and c the speed of light. for K << 1, the emitted radiation is intense with narrow spectral bandwidth. The energy of the harmonics can be shifted by altering the magnetic field, B, most often carried out by altering the size of the gap between the banks of permanent magnets.
For the experiments reported here, the energy width of 10eV of an undulator harmonic is still of order 100 times too broad. This means that the synchrotron radi-ation from the insertion device is sent through a monochromator.
A X-ray monochromator consists usually out of two very precisely spatially ori-ented crystals or a ruled metal grating as shown in Fig.2.6b. When hit with the still relatively broad spectrum coming from the undulator, the grating diffracts the beam of X-rays according to Braggs Law:
2d sin(θ) = nλ (2.3)
By choosing the angle of the monochromator with respect to an exit slit, a certain wavelength λ can be selected out of the spectrum. By careful, synchronised motion of both the monochromator angle and the undulator gap value, the photon energy can be swept over the desired ranges.
(A) An undulator consists of rows of magnets with alternating polarity as indicated by the two colours. The resulting magnetic wiggles the charged electrons, which start to radiate. Figure taken from
wikipedia.com
(B) The monochromator disperses the light coming from the undulator, and uses an exit-slit to select a spe-cific wavelength. Figure taken from
[10]
2.2.2 X-ray Absorption Spectroscopy
The high tunability in energy of the X-ray photons available at a synchrotron light source gives an opportunity to look at specific elements inside complex materials. In a crystal lattice, the outermost electrons of the atoms, the valence electrons, have low binding energies and are involved in chemical bonding or delocalized, forming bands. They hold little information about the specific atoms in the material. The deeper lying ’core’ electrons, however, are strongly localized at their host atoms and have very specific binding energies. These core-electrons can be excited to higher, unoccupied states using photons of a specific energy as depicted in Fig. 2.7. This local process in which an electron is promoted to an unoccupied state is called X-ray Absorption.
The chance of a X-ray photon being absorbed in this way is described by a cross section σ, and is given in terms of Fermi’s Golden Rule. The cross section of an electron-photon interaction is proportional to the probability of the transition per unit time from a specific energy eigenstate |ψni to another eigenstate (empty orbital) hψm|: σ ∼ 2π ¯ h |hψm| H |ψni| 2 δ (Ef− Ei− hν) (2.4) Where Ef and Ei the energy of the final and initial state, hν the photon energy, and H the hamiltonian describing the interaction which couples the two eigenstates of the system. The delta-function is used to ensure the conservation of energy. The interaction hamiltonian H can be written as a Taylor expansion, where the first and most important term describes the electric dipole transition:
Hdipole= e
mcp · E (2.5)
where p is the electron momentum operator and E the photon electric field. The smaller terms that follow, such as the quadrupole operator, are often ne-glected. The dipole operator only yields a positive intensity for certain transitions, which are summed up in the dipole selection rules. An extensive treatment of all transition possibilities and probabilitys can be found in Ref. [11]. For this thesis, the most important results from this derivation is that electrons from an s orbital that get excited by a photon, will go into an unoccupied p orbital, and p electrons will be ex-ited into either unoccupied s or d states. The inner product between p and E means that the direction of the electric field of the photon with respect to the electron orbital has a great influence on the absorption probability. This can be used in a technique called Linear Dichroism, on which will be expanded in the next subsection.
Core levels of atoms are described by the quantum numbers n (principal quan-tum number), l (orbital momenquan-tum) , s (spin momenquan-tum) and j (total angular mo-mentum j = l ± s. In XAS, the transitions are labeled by the atomic shells (K,L,M etc), which are indicated in Fig2.7.
2.2. X-ray techniques 11
FIGURE 2.7: A few examplary aborption edges in XAS. The main edges studied for VO2are the vanadium L-edge (2p → 3d) transition and the oxygen K-edge (1s → 2p). Figure taken from wikipedia.com
So the empty, un-occupied orbitals are what is being measured with XAS. When a core level electron is excited by a photon to a higher lying un-occupied state, it leaves behind an empty core hole. As the core hole is a high energy state, there are multiple pathways by which the system can relax. The core hole can be filled by an electron from a higher lying shell, resulting in a release of energy. This energy can be emitted in the form of another photon, but it can also be used to eject a second electron all the way out from the atom to the continuum. This is called the Auger-effect. The two options are depicted in Fig.2.8.
FIGURE2.8: Two pathways for a decaying corehole. Either the excess energy from the falling electron is released in the form of photon, or it kicks a second electron out of its shell into the vacuum leaving behind
a charged ion. Figure taken from www.lpdlabservices.co.uk
This Auger process removes an electron from the material, leaving it charged. The same is the case for other, secondary electrons created as the system distributes the core hole energy. This charged state is compensated for by connecting the sample to the ground, and the resulting current (called the drain- current) can be measured with a sensitive ammeter. Measuring this current as a function of incoming photon
energy gives a XAS spectrum measured in what is called the Total Electron Yield (TEY) mode (Fig. 2.9a). In general this is a good measure of the absorption process. When working with a sufficiently thin sample, some of the X-rays may not be ab-sorbed at all, and pass through. These residual photons can be detected with a diode or a CCD camera. If this signal is compared with the intensity of the incoming X-ray beam, this gives the most direct measure for the absorption inside the sample, see Fig.2.9b. This is the basis of doing XAS in transmission, and this is one of the main techniques used in this thesis. Measuring XAS in transmission is commonplace for the hard ray regime, as samples do not need any special preparation. In the soft X-ray regime, the benefits of the excellent sensitivity to transition metal d and oxygen p orbitals, as well as to spin and orbital degrees of freedom are offset by the fact that transmission is only possible through very thin (<100nm) membranes, due to the high absorption cross-section for soft X-rays. The difference between the physics in-formation contained in the XAS spectra obtained in the TEY and transmission modes lies in the fact that transmission is a measure for the absorption throughout the en-tire sample, whereas the TEY signal is mostly sensitive to the surface region, probing maximally the outer 10 nm of material. This is because the mean free path of the Auger and secondary electrons that leave the sample, yielding the drain current, is in the order of a few nanometers.
(A) (B)
FIGURE2.9: (A) XAS in Total Electron Yield mode. The sample cur-rent to the ground, originating from electrons being excited to the continuum is a measure of the X-ray absorption.(B) XAS in Transmis-sion mode. Comparing the intensity of the X-ray beam before and after passing through the sample is the most direct measure of
ab-sorption possible.
2.2.3 X-ray Linear Dichroism
Since not all orbitals are symmetric in all spatial dimensions, the orientation of the empty orbitals is also an important factor in whether or not an incoming polarized photon is absorbed, as discussed in the context of the absorption in section 2.2.2. If the inner product between the electric field vector and the unoccupied orbital is maximal, the interaction cross section can be greatly enhanced. This feature can be very usefull in determining the charge order inside a material, by comparing the absorption of two known, different polarizations of light. This is what is done in the
2.3. Electron based techniques 13
technique called X-ray Linear Dichroism (XLD), as illustrated in Fig. 2.10. More on the specifics of XLD on VO2 is given in section3
FIGURE2.10: Schematic representation of the principle behind X-ray Linear Dichroism. Imagine an unoccupied p-orbital on the sample surface as shown. In this case the vertically polarized light, where the electric field is aligned with the available orbital, will have a large
dipole transition probability (p · E is large).
————————————————-2.3
Electron based techniques
The second major tool of investigation in this research next to X-rays is to make use of electrons. Just like X-rays, electrons behave like waves with the main difference being that they posses a charge and mass. Because of this mass they interact much more strongly with materials, and have a very short wavelength compared to X-rays. This makes them very usefull in scattering and diffraction experiments. These kind of experiments are most often used to determine crystallographic and electronic structure.
2.3.1 Ultrafast Electron Diffraction
One of the most interesting aspects of the Insulator to metal transition (IMT) is its dy-namics, and especially the coupling between the structural and electronical changes, taking place during the IMT. This is a relatively unexplored area, but with recent advances in ultrafast laser technology and measuring techniques such as Ultrafast Electron Diffraction and free electron lasers, this is a side of the transition that is in-creasingly being explored. Finding the ’speed limit’ of the different components of
the transition would be a very important step to developing a microscopic under-standing of a Mott MIT, and would also provide key information towards eventual device development, where one would like to switch the material millions to maybe billions of time per second. With that idea in mind, an attempt was made to measure the insulator-metal transition at the Ultrafast Electron Diffraction set-up at SLAC National Accelerator Laboratory, California USA.
The Ultrafast Electron Diffraction experiments in this work have been performed during a secondment at the ASTA beamline at SLAC. ASTA provides Ultrafast Elec-tron Diffraction with a optical pump-MeV elecElec-tron probe set-up.
The lay-out of the beamline can be seen in Fig.2.11.
FIGURE 2.11: Lay-out of the UED set-up of the ASTA experiment at SLAC. Not shown is the 50 MW Klystron generating the radiofre-quency waves used to accelerate the electrons to MeV kinetic
ener-gies. From [12]
The electrons are created in a photocathode radio frequency (rf) gun with help of a kHz, 5-mJ Ti:Sapphire laser. A ultrastable rf generator, a 50 MW klystron, accel-erates and compresses electron bunches into packets with bunch lengths of ± 100 fs and up to 3.6 MeV kinetic energy. [12]
As with every pump-probe experiment, the timing of the probe pulses with re-spect to the pump is crucial for a succesful experiment. The optical pump pulse is created with the same Ti:Sapphire laser as the electron gun uses by means of a beam splitter. This is both efficient and convenient, as this guarantees that the electron bunches and the pump pulses are created with the exact same frequency. The pump pulse is send through a physical delay stage, which consists of two mirrors that can be moved with respect to each other to increase the path length and thus the arrival time of the pulse. The electron bunches are first calibrated with the pump pulse on a reference gold sample. When the pump and probe pulses arrive at the same time, the time delay between them can be very precisely adjusted by changing the length of the delay stage.
2.4. Free Electron Laser 15
FIGURE2.12: Schematic picture of a pump-probe electron diffraction experiment. Picture from slac.stanford.edu
A typical experiment runs at 100 Hz, which means the sample is pumped, probed and relaxes again 100 times a second. Measuring a diffraction pattern for single time-delay can take 10’s of minutes as data from thousands of pulses needs to be summed up so as to obtain sufficient signal to noise. After this, the pump-probe delay is altered and a new pattern is measured. In this manner, step by step a time-dependent diffraction ’movie’ can be measured.
2.4
Free Electron Laser
The Free Electron Laser (FEL) called the Linac Coherent Light Source (LCLS) at Stan-ford is a unique apparatus capable of delivering femtosecond long, extremely bright, coherent X-ray pulses. It makes use of a linear accelerator to accelerate bunches of electrons over more then 3 kilometers. The electron bunches are guided through the magnetic structures of an undulator to generate electromagnetic radiation, similar to those in storage ring facilities discussed previously in the section above. Initially the radiation emitted is incoherent, but the electric field of the radiation then inter-acts with the accelerated electrons, causing them to form microbunches modulated precisely in phase with the light. The consecutively emitted radiation then adds coherently to the beam. The LCLS is capable of producing X-rays a billion times brighter then synchrotron X-ray facilities, exceeding 1013 photons per pulse. The
FIGURE2.13: Arial view of SLAC, including the LCLS facility. The Linear accelerator is over 3 km long
pulsed nature of the light makes it possible to do time-resolved pump-probe exper-iments.
2.4.1 XAS at FEL
One of the major challenges in X-ray FEL science has been measuring high resolu-tion X-ray absorpresolu-tion in a reasonable time. While the energy and pulse duraresolu-tion of the photons can be precisely controlled, the intensity of the incoming X-ray pulses is hard to keep track of. The way this is done in synchrotron sources, by measuring a monitor current upstream of the beam (see sectionA), does not work in this regime of such enormous numbers of photons/pulse. The measured monitor current be-comes very non-linear for such brilliance, and it bebe-comes non-trivial to normalise the measured absorption to the intensity of the incoming beam this way. The inten-sity can vary significantly between pulses, making this task even more complicated. A clever idea around this was proposed by Dr. Bill Schlotter, an instrument scien-tist at SLAC. The method has been tried out for the first time, using the VO2samples discussed in this thesis as a test case. The main idea is to split the soft X-ray FEL beam into two identical parts by means of diffraction off of a zone plate (Fig.2.14) . Thus one part of the pulse passes through the sample, and the other part of the pho-ton pulse flies past the sample and is measured downstream as a reference. When dividing the transmitted sample signal by the reference beam intensity, the pulse-to-pulse fluctuations are normalized. Just as with the electron diffraction experiment, an 800nm optical pump was used for the pump probe experiment.
2.4. Free Electron Laser 17
FIGURE2.14: Idea behind the ’split and measure’ technique utilised at the SXR beamline at SLAC. The initial photon beam is split into two identical parts, which are seperately measured on the same CCD detector. One is used as a reference, while the other is passed through the sample. This enables normalising the absorption spectrum to the reference beam, by avoiding the non-linearity problems of measuring a monitor current. This increases the signal-to-noise level to photon
shot noise. Image courtesy of Dr. Bill Schlotter.
The system is run and read-out at 120 Hz, utilizing a special pnCCD camera [13]. With this, enough statistics for a photon noise limited spectrum can be gathered in a matter of minutes.
19
Chapter 3
Background and Theory
3.1
VO
2Vanadium dioxide’s metal to insulator transition contains a lot of interesting physics besides just the five orders of magnitude drop in resistance when crossing the transi-tion temperature. Vanadium dioxide is a correlated electron insulator. According to regular band theory, a material with a partially filled outer band should be metallic at all temperatures. Based on this non-interacting particle picture, VO2 should fall in this class. The low temperature insulating state is, therefore, not accounted for within single particle theories, and arises due to different mechanisms.
3.1.1 Structural properties
For a long time it has been known that there is both a structural and an electronic reconfiguration taking place during the MIT. In the high temperature metallic phase, VO2has a rutile crystal structure as show in figure3.1awith a P 42/mnmspace group symmetry. The crystallographic a- and b axes are similar is this structure, with a unit length of 4.5546 Å, and a c-axis with length 2.8514 Å[14]. When crossing the transi-tion temperature, an anti-ferroelectric distortransi-tion makes the vanadium atoms along the c-axis dimerize and makes these dimers twist slightly, changing the crystal struc-ture to monoclinic. This doubles the unit-cell length and causes a zig-zag pattern of vanadium atoms along the [110] direction. The oxygen atoms do not move signifi-cantly during the transition.
(A) (B)
FIGURE3.1: (A): both crystal structures of VO2. Taken from [15]. (B) top view of the crystal, with the oxygen atoms omitted. The dimer-ization and tilting along the rutile c-axis are ultimately the cause of
the gap opening [16].
3.1.2 Electronic aspects of the MIT
Goodenough [17] was the first to present schematic band diagrams for VO2 based on a crystal-field model that intuitively showed the electronic structures of both the metallic and insulating phases.
Crystal field theory describes the breaking of degeneracies in electron orbital states, due to an electric field produced by a surrounding charge distribution. The vana-dium atoms in the VO2 crystal are surrounded by an octahedron of oxygen atoms, that produce a static field lifting the five-fold degeneracy of the d-orbitals at the vanadium sites. The vanadium d-orbitals are split into three-fold degenerate low en-ergy t2gorbitals and two-fold degenerate egorbitals as shown in figure3.2. The lobes of the egorbitals (dx2−y2 and dz2) point along the coordinate axes, generating strong
’σ’ type bonds with the oxygen ligands, while the lobes of the t2g orbitals (dxy,dyz and dxz) point in between the coordinate axes and generate the slightly weaker, ’π’ type bonds.
FIGURE3.2: Splitting of degenerate d-orbitals due to an octahedral crystal-field into three low-lying t2gand two higher egstates
The octahedra in VO2are slightly distorted: the difference in apical and equato-rial V-O distance cause the t2g orbital group to further split into a highly directional d|| and a residual π∗ states [18]. In the rutile phase both these states form the den-sity of states at the Fermi level, as predicted also by regular band structure calcula-tions. During the MIT the vanadium atoms dimerize along the rutile c-axis, and this causes the d||band to split into a filled bonding, and an empty antibonding state in
3.1. VO2 21
a so called Peierls transition. The π∗ state also shifts up in energy due to the tilting of these dimers in a zig-zag pattern. Now, in the insulating state, the Fermi level lies between the split d||bands, in a 0.6 eV wide energy gap which has been exper-imentally shown by a wide variety of techniques such as VUV reflectance [19] and photoemission [20].
FIGURE3.3: Schematic density of states representation of VO2with all the hybridized O2p/V3d states. According to the Goodenough-picture, the dimerization and zig-zagging of the vanadium atoms opens a 0.6 ev gap in the low-temperature monoclinic phase.
Adapted from [14]
Later this model was adapted by Zylbersztejn and Mott, who proposed that strong electron-electron correlations in the d|| band also played a major role in the coupled structural change and metal-insulator transition. They claimed these effects are initially screened by the π∗band in the high temperature phase, but become ap-parent when that band is shifted upwards and away from d||due to the displacement of the vanadium atoms in the low temperature phase [14] [21]. Biermann et al. [22] predicted that such correlations within the vanadium dimers should yield an addi-tional band, called the dk-singlet.
In the insulating state both the highly directional dk bands, due to the Peierls distortion, and the vanadium-vanadium singlets are unoccupied and ordered along the crystallographic c-axis. As mentioned in section2, this gives an opportunity to investigate them using polarisation dependent XAS experiments. The X-ray spot size at the UE56 endstation at BESSY is 10’s of micrometers large. This is so much larger then the size of the nanosheets, that there will allways be an ensemble of different grains sampled in each measurement. At normal incidence of the X-ray beam at the sample, there is therefore no expected difference in the number of grains providing Ekc for either linear horizontal or linear vertical polarization. There will be a statistically identical proportion of the grains with their c-axis oriented along the electric field vector for both polarisations, so no linear dichroic contrast is to be expected. At grazing incidence to the sample however, horizontally polarized light always has its electric field vector perpendicular to the sample plane. And thus there are no grains with the directional d||bands aligned along the electric field. For vertical polarisation, the E-vector remains in-plane, meaning that a proportion of the
grains will have Ekc. This means that -for grazing incidence- linear dichroic contrast is to be expected.
FIGURE 3.4: The highly directional d|| orbital depicted is preferen-tially oriented along the crystallographic c-axis and are completely unoccupied in the insulating phase of VO2. The π∗ states are more isotropically distributed, and are not expected to contribute to
polar-isation dependent contrast. [16]
3.1.3 XAS spectra of VO2
Vanadium is a transition metal with a partially-filled outer 3d shell. As a single atom, its electronic structure is [Ar]3d34s2. When bonding in a crystal with oxygen it donates its 4 outermost electrons to the octahedral crystal bonds, forming a V4+O2− 2 compound. Its effective electronic structure then becomes [Ar]3d1which in the field of transition metal oxides is usually called a 3d1 complex. It is this single d-electron that, due to a combination of Peierls, Mott-Hubbard and orbital ordering physics, forms an insulating state at low temperature, despite representing a half-filled band situation.
When exciting an electron from the 2p to the 3d states, a transition between the states 2p63d1 and 2p6−13d1+1
, there are two non-degenerate possibilities allowed by the dipole selection rules. Excitation from the 2p3/2level is called the L3peak, and from the 2p1/2it is called L2. These are the two main features in the vanadium L-edge spectrum. The core hole in the 2p band couples to holes in the 3d shell, generating some fine structure on the edge due to multiplet effects. These fine structures include the small shoulder at the onset of the peak. The onset of the L-edge of vanadium can be found at around 515 eV binding energy.
The oxygen K-edge, comprising oxygen 1s to 2p transitions, is very close in en-ergy. It starts around 530 eV and sits on the high energy tail of the V L-edge. Nor-mally transitions from O 1s to V 3d states would be dipole-forbidden, but it is known that also these states are probed at these energies [3]. The fact that at the O-K edge the V 3d orbitals are probed is due to strong covalent hybridization between the oxygen 2p and vanadium 3d levels.
3.1. VO2 23
FIGURE3.5: Example spectrum of rutile VO2with both the V L-edge and the O K-edge. The splitting between the L2 and L3 edges is caused by the V 2p spin-orbit coupling [25]. The sharp double peak in the O-K edge are crystal-field split states, hybridized with the V 3d states and correspond to the π∗ and σ∗ states seen in the density of states picture in Fig. 3.3. The highly directional d||state is expected to lie in between those two features, and should be discernible in the monoclinic phase data measured as part of this thesis research at the
UE56-PGM1 at BESSY II.
In the high temperature, rutile phase there are 3 distinct features in the O:K edge spectrum. A sharp double peak between 530 and 534 eV, and a broader structure about 10 ev higher. The sharp double peak is linked to the V 3d state. According to the model of Abate [23], the V 3dx2−y2 orbital overlaps with the O 2p which accounts
for the higher energy, sharp peak called σ∗. The V 3dxz and V 3dyz do not overlap with the O 2p and are called π∗ (see Fig. 3.4). The broad feature at high energy is linked to the covalent overlap of the O 2p states with the V 4s and 4p bands [23] [24]. An example XAS spectrum of VO2is shown in figure3.5.
3.1.4 Electron Diffraction
By diffracting electrons from the material, one can learn about the crystal structure. In the UED experiments reported here, this takes place in transmission geometry. The interference pattern of the transmitted electron beam is measured by means of a fluorescent screen in the far field. The diffraction pattern from a plane wave from a crystal is a representation of the reciprocal lattice of the crystal. The electrons diffract from the periodicity of lattice planes (as illustrated in Fig.3.6), and each set of planes causes positive and negative interference which can be detected as intensity fluctu-ations on the detector.
Electron diffraction is an elastic process, meaning that the energy and wavenum-ber of the incoming waves are conserved. Just the direction is changed by a momen-tum transfer vector denoted s = kf − ki, where kf and ki are the final and initial wave numbers, respectively. The relation to the lattice spacing is : s = 2d1 . From the
FIGURE3.6: (A): cartoon of the top view of a crystal with c-axis in plane. For drawing-simplicity here the a-axis is also drawn in plane, such that a (010) surface is pictured. In the real VO2 films grown on nanosheets the film plane is (110). Some example sets of crystal-planes that can act as a diffraction grating are indicated with their
Miller indices.
measured momentum transfer one can thus determine the lattice plane spacing at the origin of the diffracted beam.
Thus, although the diffraction peak position is relatively straightforward to un-derstand, calculating the expected intensities of the peaks is more difficult. Each single atom in the crystal contributes to the scattering intensity of multiple Miller planes. Besides depending on the incoming wave, the diffracted intensity of a spe-cific set of planes depends on both the amount of atoms in that particular plane, their position in said plane, and their atom specific properties. All this atom-specific scattering information (such as atomic weight, electron shell occupation, etc) is cap-tured in the atomic scattering factor, f , of these atoms. The structure factor equation is used to compute the expected scattering intensity Ihkl2 = F (s)by summing over all the atoms in an (hkl) plane, using their scattering factors:
F (S) =X n
fne2πi(hx+ky+lz) (3.1)
Where the complex exponent is used to represent the fact that the incoming wave has both an amplitude and a phase. The h,k,l values are from the corresponding planes giving rise to the diffraction peaks with the given Miller indices, and the x,y,z values are the Wyckoff positions from the atoms in the unit cell. The structure factors fnfor electrons scattering of atoms are usually approximated as a sum of n Gaussians: f(e)(s) = n X i=1 aiebis 2 (3.2)
3.1. VO2 25
where the fitting parameters ai,bi can be found for most available elements in [26].
For a monoclinic crystal, the distance ’d’ between specific, Miller-indexed planes (hkl) can be calculated via the following formula, where a,b,c are the unit cell lengths in their respective directions, and α, β, γ the corresponding angles:
dhkl= h a 2 +k sin βb 2+ lc2−2hl cos βac (sin β)2 (3.3)
For the rutile crystal structure this formula becomes:
dhkl = h2+ k2 a2 + l c 2 (3.4) As for the lattice parameters, the values stated by Eyert [14] are used. The rutile crystal structure is a simple tetragonal lattice with space group P 42/mnm. All angles in the unit cell are 90 degrees, and the lattice constant are: aR = bR= 4.5546Å and cR= 2.8514Å. The vanadium atoms are located at the Wyckoff positions (0,0,0) and (12,12,21). The oxygen atoms occupy ± (u,u,0), ±(12 + u,12 − u,1
2), with u = 0.3001 Å .
In the monoclinic M1 phase, the crystal is characterized with spacegroup P 21/c. The lattice constants and monoclinic angle are: aM 1 = 5.743Å, bM 1= 4.517Å, cM 1 = 5.375Å and βM 1 = 122.646 degrees. The Wyckoff positions of the vanadium and oxygen atoms in the unit cell are given in the figure:
FIGURE3.7: Wyckoff positions of the atoms in the monoclinic unit-cell. Table taken from [14]
All the calculated reflections for the Miller-indexed planes (100) - (334) in both phases are given in the appendixA.
3.1.5 The (Ultrafast) Insulator to Metal Transition in VO2
The IMT is thought to be determined by a complex interaction between charge, lat-tice, spin and orbital degrees of freedom. Recent experiments have shown that these processes might happen on very different timescales. For example [27] has shown using optical spectroscopy experiments that when photoinducing the IMT using pump pulses of energy 1.5 eV, the transition occurs on a timescale of roughly 100 fs. Ultrafast electron diffraction experiments on the other hand show that the com-plete lattice transformation is a lot slower, only saturating some 10’s of ps after the
optical excitation pulse [15]. This is in agreement with recent rocking curve pump probe X-ray diffraction experiments reported by Gray et al. [4].
Measuring the timescale and contributions of all the different components active in the IMT can only be done using such pump probe experiments. Investigating whether the electronic and structural changes can take place in a temporally de-coupled fashion is one of the main goals of the complex oxide collaboration within which this masters research project took place.
27
Chapter 4
Results
4.1
Soft X-ray Absorption Spectroscopy at the Synchrotron
(A) (B)
FIGURE4.1: Two polarisation dependent spectra at both (A) low and (B) high temperatures recorded from VO2thin films in TEY mode. These spectra are taken at grazing incidence (75
degrees off normal).
In figure4.1 two typical polarisation dependent soft X-ray absorption scans at the V- L2,3 and O-K edges and their difference is shown for both the high tempera-ture (metallic) and the low temperatempera-ture (insulating) phase of VO2 recorded in TEY mode. The changes in the XLD between the two temperatures is especially large at the onset of the V L-edge, between 514-516 eV and over the whole O K-edge (527-534 eV). These two regions have been highlighted in Fig. 4.1using boxes. The XLD at both temperatures indicates the existence of in-plane ordered orbitals at both tem-peratures, but is more pronounced in the monoclinic phase as predicted in section
3, where the highly directional orbitals are oriented along the dimerized vanadium atoms, as a result of the Peierls-like distortion. The shape of the XLD is very similar to bulk crystal measurements as reported by [16] and [28].
(A) (B)
FIGURE4.2: Two polarisation and temperature dependent scans in transmission geometry, taken at an incidence angle of 55 degrees from normal. XAS spectra for Ekc (blue) and E⊥c (red) are shown for 280K (A) and 380K (B). The shape of the spectra at first sight is very different to the ones in TEY, the reason behind this is discussed in appendixA. The dichroic contrast at the vanadium onset and on the O K-edge however is similar to the TEY in shape
and position
In figure4.2 analogous spectra are plotted that were measured in transmission geometry. At the V L-edge onset and O K-edge, the dichroic contrast is very similar to the data measured in TEY mode, for both temperatures. At the peaks of the V L-edge the contrast is not visible in transmission due to saturation of the absorption signal. This effect is covered in appendixA. These XAS spectra are the first of their kind, taken in transmission mode on thin film VO2 samples. These experiments prove that our high quality VO2 samples are suitable for soft X-ray experiments in transmission, giving the opportunity to attempt further, sophisticated experiments for investigating the IMT as discussed in the introduction.
4.1. Soft X-ray Absorption Spectroscopy at the Synchrotron 29
FIGURE 4.3: Zoom in on the oxygen K-edge using XAS in transmission. From top to bottom: changing from maxi-mally grazing to normal incidence. The dichroic contrast in the low temperature data (280 K, blue traces) disappears com-pletely (black arrows). Solid and dash-dotted lines are horizontal and vertical polarisation respectively. Red and blue represent metallic (380K) and insulating phase (280K). At normal incidence (θ = 90◦), there is no XLD visible for both
metallic and insulating phase.
Zoomed in at the O-K edge, there are some significant changes in both the spec-tral shape and XLD between high T (metallic, rutile) and low T (insulating, mon-oclinic) VO2. Firstly, very prominent is the disappearance of the dichroic contrast on the high energy side of the main peak (between 532 - 533.5 eV) when changing the incidence angle of the beam as can be seen from figure4.3. This is indicative of unoccupied orbitals ordered in the plane of the film as discussed in chapter3. The large dichroic contrast around 532.5 eV corresponds to the "Peierls" dkband, and ap-pears very similarly in our transmission data to that reported by Gray et al for TEY measurements [16] . The smaller, dk-singlet band they report to see on the leading edge of the O:K absorption is not visible in the dichroic signal for our samples. The reason for this lies in the fact that we use samples with the c-axis in-plane, which, due to the ensemble averaging across multiple individual nanosheets, makes mea-suring a ’pure’ dichroic contrast impossible and reduces our overall sensitivity for this particular aspect of the monoclinic phase of VO2.
Secondly, in our data the leading edge of the oxygen K edge spectrum shifts to lower energy for the metallic state. This is in accordance with the DOS picture sketched in figure3.3, and agrees with what has been reported in [16] and [29]. In the metallic, rutile phase, the π∗and σ∗states have shifted approximately 0.2 eV to lower energy for higher temperature, compared to the monoclinic, insulating case.
The third, temperature dependent major feature is an extra ’shoulder’ appearing at low temperature in between the π∗and σ∗states. This spectral weight corresponds to the dk band that was mentioned in chapter3, and is screened by the π∗ states in the high temperature phase.
(A) (B)
FIGURE4.4: Temperature dependent transmission spectra at the oxy-gen K-edge. For increasing temperature, the π∗ and σ∗ states are shifted approximately 0.2 eV to lower energy. At low temperature, the spectral weight of the dkstate is shifted up in energy, visible as an
extra shoulder on the π∗peak.
The temperature dependence of the O K-edge transmission spectrum all the way through the MIT is shown in figure4.4. The orbital-related density of states picture that relates the spectra to the theoretical predictions are plotted to the side. Each XAS trace is taken in small increments of 5K around the transition temperature. In these data, the shift of the π∗ and σ∗ states with increasing temperature can be followed with great accuracy, just as the appearance of the d∗kfrom underneath the π∗ can be seen in the low temperature, monoclinic phase.
The shift of the O K leading edge is displayed even more clearly in figure4.5a. Here the photon energy is set to 529.1 eV, in the middle of the leading edge slope of the O K-edge, where the red arrow is placed in Fig. 4.4. The absorption is measured in TEY mode while the temperature is swept across the transition and back. The hys-teresis curve that is obtained in this way is a measure of the volume fraction of the insulating (low intensity at 529.1 eV) and metallic (high intensity at 529.1 eV) phases under the probe beam during the transition. In figure4.5the resistivity curve of the same sample is given for comparison. The resistivity is not so much a measure of the amount of metallic vs. insulating material, but more an indication of when per-colative pathways are formed to conduct electricity from one contact to another. By comparing the two curves, we see that the hysteresis loop from XAS is at maximum ± 15 K wide, where for the resistivity this is about 12K. The XAS curve seems to
4.1. Soft X-ray Absorption Spectroscopy at the Synchrotron 31
(A) (B)
FIGURE4.5: (A): XAS intensity taken at 529.1 eV in TEY mode (on the leading slope of the O K-edge) as function of temperature. The hysteresis curve is normalized and flipped as to make comparison
with the resistivity curve (B) easier.
decrease and increase more gradually at the onsets of the transition, where the ma-jor change at the transition is a bit more abrupt in the resisitivity data. This suggest that when the volume fraction of the one phase is growing, and within the sensitiv-ity limit of the XAS, for example when heating from the insulating to the metallic phase, in the transport there is not yet a complete metallic pathway formed to con-duct the electricity, although a significant volume fraction is already rutile. Going the other way, cooling from metallic to insulating, not yet all metallic pathways be-tween two contacts may be interrupted when a significant volume fraction of the sample is already monoclinic. To illustrate this, two symbols have been overlaid on the two curves, each represents where the XAS leading edge shift suggests almost phase pure situation.
4.2
Ultrafast electron diffraction
(A)
(B) (C)
FIGURE 4.6: (A): Electron diffraction pattern in monoclinic phase (325K), recorded using the ASTA UED experiment at SLAC. (B) Zoom in at the diffraction pattern in the monoclinic phase.(C) Zoom in at diffraction pattern in the rutile phase (pumped with 1.2 mJ/cm2laser). Two rings in the monoclinic phase are indicated with the yellow arrows in (A), and with dashed line in (B). These rings disappear when pumped (C), indicating a transition of the
crystal to a higher symmetry where these reflections are not allowed.
Two electron diffraction patterns, one for each of the two phases, are shown in Fig.
4.6. Due to the mosaic structure of the samples, the diffraction patterns do not con-sist of single, high intensity spots, but these are azimuthally smeared out to form rings. The grains that are averaged over however are highly crystalline, and this causes some structure to be present in the diffraction rings. Starting from room tem-perature, for which the film will be mainly in the monoclinic phase, there are rings that completely seem to disappear when pumping the sample. Due to the lower symmetry of the crystal in the monoclinic phase, there are multiple planes in the crystal that yield a diffraction feature that do not exist in the high symmetry situ-ation (rutile). The azimuthally integrated patterns of the above diffraction images
4.2. Ultrafast electron diffraction 33
are shown in figure4.7. Some of the features grow more intense when heated, for example those at s=0.16 Å−1and s = 0.21 Å−1. Other features weaken, for example at s = 0.35 Å−1and s = 0.38Å−1.
FIGURE4.7: Azimuthal integration of diffraction patterns with and without pumping. A disappearing ring from figure4.6is indicated with the yellow circle. The patterns are being compared with earlier diffraction experiments performed by Morrison et al [15] on polycrys-talline samples (lower panel). Most of the peaks found by Morisson et al. are also seen for our epitaxial samples, including the disappearing features around 0.45 and 0.53 Å−1. The (30-2),(0-21), (12-2) features indicated in the Morisson paper are not visible in our data. Notable differences are in the (200) and (220) peaks on the one hand, which appear to be much less intense in our epitaxial data, as well as the
(0-11) peak that appears much more intense.
Diffraction patterns have been taken at multiple temperatures across the transi-tion, are normalised and a powerlaw background is subtracted as described in ap-pendixA. The change in the features’ intensity and momentum transfer as function of temperature is shown in Fig.4.8. In this figure, the diffraction pattern at the low-est temperature (325K) is subtracted from the patterns at every other temperature, to visualize how the intensity changes as a function of temperature.
(A) Heating (B) Cooling FIGURE4.8: Azimuthally integrated diffraction patterns from which the lowest T (monoclinic) pattern has been subtracted. Clear hys-teric behaviour of the intensity as function of temperature can be seen in the changing of the patterns. Inset: cartoon of the hysteretic be-haviour of the intensity of a diffraction peak as function of
tempera-ture, at the s value indicated by the blue and orange arrows.
4.2.1 Simulation
One of the contributions of the MSc thesis research to the larger-scale collabora-tion on the dynamics of the VO2 MIT was an attempt to simulate the UED profiles. The simulation of the diffraction patterns are plotted together with the experimental data in figure4.9. The simulation consists of all the calculated reflections from the set spanned by the Miller indices (100) to (334), calculated as covered in the theory section, in chapter3, and in section A.2 of the appendix. Most of the features in the experimental data can be more or less reproduced in the simulation. For the M1 phase, there is one intense feature at 0.1 Å−1, corresponding to the (100) reflection, in the simulation that is not seen in the data. One possibility is that this reflection is not visible because of the epitaxial ordering of our sample. Reflections for which the momentum transfer would be parallel to the electron beam are expected to be lost due to this decrease in degrees of freedom. One feature in the M1 phase on the other hand is a lot stronger in the data then in the calculation, i.e. that at 0.35 Å−1 . This could be because of multiplicity of peaks.This means that, within the experimental resolution and finite width of the diffraction peaks, a number of reflections add up in the experiment, whereas the simulation only foresees a single diffraction solution for a particular s-value. Such multiplicities arise because of crystal symmetries, and are difficult to calculate for complex crystal structures, and their robust theoretical treatment goes beyond the scope of this thesis.
4.3. Soft X-ray absorption spectroscopy at LCLS 35
(A) M1 phase simulation and data (B) rutile phase simulation and data FIGURE 4.9: Simulations of electron diffraction profiles of both the M1 and rutile phase of VO2 together with the appropriate experi-mental data.The largest disagreements between the simulation and experiment for the monoclinic phase are highlighted in green and or-ange. The thin vertical lines show that the simulation is succesful in describing the s-values of the majority of the experimental diffraction
peaks.
4.3
Soft X-ray absorption spectroscopy at LCLS
(A) (B)
FIGURE4.10: (A) two XAS spectra at 300K of 50nm VO2taken at the LCLS. The orange spectrum is taken using the previous state of the art method of measuring XAS at a FEL. The blue spectrum is taken using the new split and measure technique. The signal to noise ratio in the new set-up is an order of magnitude better than before. (B) XAS spectrum at 300K on 50 nm VO2taken at a the BESSY II storage ring
in transmission for comparison.
An example of a XAS spectrum taken at the LCLS can be seen in figure4.10a. Com-parison of these V-L2,3and O-K edge data from the free electron laser, with the data
from BESSY-II shown in Fig.4.10bproves that with the split and measure technique, one is capable of measuring high quality, photon noise limited XAS spectra com-parable with the best available synchrotron methods. The data shown in Fig. 4.10
took less then 10 minutes to measure. Simultaneous XAS measurements with an old technique are also plotted here. Previously the state-of-the-art method of measur-ing XAS at a FEL, used a multichannel plate to measure the fluorescence from a SiN membrane upstream in the beam. More on this technique can be found in Ref. [30]. The improvement in the signal to noise level made by the new set-up is about an order of magnitude.
Time resolved XAS data is shown in figure4.11.
(A) (B)
FIGURE4.11: Time-resolved XAS at the vanadium L-edge and oxygen K-edge. (A) shows a 2-dimensional map of absorbed intensity vs. time-delay between the probe and pump laser pulses. Each pixel corresponds to a single flash of the X-ray laser, at a specific energy, with a specific time-delay with respect to the pump pulse. At positive time-delay values, the pump laser arrives before the probe X-ray pulse, and thus a photoexcited state is measured by the probe pulse. Subtle differences are visible in absorbed intensity in this regime compared to negative time-delay (the unpumped, static state). The coloured stripes running in the vertical direction are a 2D, false colour representation of the XAS spectrum (energy on the x-axis) vs. time-delay (on the y-axis). Thus a vertical cut through such a dataset shows the time-delay evolution of that particular energy in the XAS data. Panel (B) shows the 2D energy vs. time-delay map with intensity in false colour. In this case the static data, recorded with negative time-delays have been subtracted from each pixel row of the image. This highlights the changes that occur post-pump. These spectra were recorded with a pump
intensity of roughly 15 mJ/cm2.
The interesting features at the oxygen K-edge are shown in more detail in figure
4.12. Time-delay traces recorded at the leading edge and at the high energy shoulder show that the temporal behavior at the transition consists of at least 2 components. The time-delay traces are fitted with a standard exponential formula with two time components: τBfor an ultrafast response, and τC for the slower change that follows.
4.3. Soft X-ray absorption spectroscopy at LCLS 37
(A) (B)
FIGURE4.12: (A) Two spectra taken at the FEL using the split and measure technique are shown in the M1 phase at 320 K. The orange trace represents static data, taken without a pumping laser, and the blue trace is the intensity at integrated time-delays from 0-5 ps after
a pump pulse of 15 mJ/cm2.
As can be seen in Figs: 4.13a and 4.13b, which show the time dependence of the two energies highlighted in panel (B) of Fig. 4.12, there is a steep change in ab-sorption for time delays of order 100fs after pumping. After this initial, very rapid change, the absorption continues to evolve more gradually, in a process persisting multiple picoseconds. This time constant is similar to the fastest timescales reported in pump-probe reflectivity studies done by Cavalleri et al, [27]. Cavelleri et al. re-ported a ’bottleneck’ timescale, in the sub 100 fs regime, where the electronic excita-tion is largely discoupled from the lattice. In this time-regime (<100fs after pump-ing) the material is hole doped by photoexcitation, but there is no metallic phase yet, and this does not happen until the structural change has caught up. This is in contrast with recent time-resolved X-ray Pump Probe (XPP) studies by Gray et al.[4], claiming a decoupling and a ’golden window’ of about 4ps where there is a purely electronic metallic state, but no structural change yet. The ultrafast response of <100fs found in this work in combination with the structural changes that happen on timescales of 300+ fs, such as those reported by Morrison et al. [15] in their ul-trafast electron diffraction studies, could fit this viewpoint. This could indicate that there is a window (albeit maybe less then 4ps) where the switching of the IMT is purely electronic. Further structural investigation, such as UED, need to be done on our nanosheet templated samples to make more conclusive statements.
The formula used for fitting the time-delay traces is given below. f (t, τa, τb, τc) = A
e−t/τa− Be−t/τb
− C1 − e−t/τc (4.1)
A, B and C are offset parameters used to gain the correct arbitrary units of inten-sity, t the time delay, and τb and τcare the timeconstants after time zero, giving an indication of the speed at which the different changes happen. Assuming nothing changes before time zero, τais set to zero.
(A) (B)
FIGURE4.13: (A) and (B) are delay dependent absorption on two interesting features in the spectrum, the leading edge and the ’d||shoulder’. Two time constants can be distinguished for both energies. A fast, sub 100fs initial change directly following the pump pulse, and a consecutive slower change in the absorption. The time constants mentioned in the figures
39
Chapter 5
Concluding remarks and Outlook
5.1
Concluding remarks
In this thesis static and dynamic aspects of the crystal structure and electronic struc-ture of a novel type of VO2sample, in the form of ultrathin soft X-ray and electron semi-transparent membranes have been investigated. Extremely thin crystalline nanosheets have proven to be a succesfull approach in the growth of high qual-ity, thin films of VO2 in an epitaxial manner suitable for transmission experiments. Using soft X-rays from storage ring facilities, these samples have been benchmarked against the literature available for single crystals, and the absorption spectra have been determined for the first time completely in transmission geometry. Combined X-ray Absorption Spectroscopy and X-ray Linear Dichroism has been used to show the high quality of our samples and their suitability for future more advanced ex-periments aiming towards the investigation of the dynamics of the Mott MIT in the bulk of VO2.
Electron diffraction has been used to investigate the structural properties of VO2. The diffraction patterns show the crystallinity of our epitaxial films, and the samples show clear and reproducible hysteretic behaviour during the Insulator-Metal Tran-sition. Very recently, time resolved pump-probe electron diffraction data became available from our collaborators on these thin films, in a second set of experiments using the same approach as was adopted here and reported in chapter4of this the-sis. At the point of submission of this thesis, the analysis of these new data is still on-going, but the first impressions show promising signs of dynamics on timescales comparable to the behavior reported for non-epitaxial, polycrystalline samples [15]. Finally, time-resolved soft X-ray absorption spectroscopy has been performed in transmission on the IMT in VO2 at the LCLS free electron laser facility at SLAC, Stanford. The data reported in chapter 4 are the first ever, high resolution, pho-ton noise limited XAS data recorded at a free electron laser. Beond this milespho-tone result, time-resolved, pump-probe data shows ultrafast dynamics for spectroscopic features that signal the transformation to the metallic phase that happen at signifi-cantly faster timescales than the changes from the structural transition gained from literature. This could suggests the existence of a time window in which the material,
after pumping from the insulating state, is metallic while the crystal structure is still that of the low temperature phase.
5.2
Outlook: two paths forward
All the results from the previously mentioned experiments are very encouraging for continuing this line of research on VO2. For the next steps there are two branches possible. Now that this work has proven the feasibility of the ultrafast investiga-tion of the Mott MIT in VO2 in transmission, one very significant next step would be to try and spatially resolve all the components of the transition using hologra-phy techniques. Nothing is known about the combination of the nm spatial and fs temporal behavior at the MIT of a complex, correlated material. To what extent and how phase separation, charge and orbital order and pattern formation interplay is wholly uncharted territory. The second avenue builds on the ultrafast XAS data reported in chapter4 of this thesis. The ultrafast fingerprint of the metallic phase that appeared at the sub-100fs timescale - well before the structural transformation away from the low symmetry, monoclinic phase can occur- raises the prospect of attempting to induce the insulator-metal transition using the strong electric fields from a THz pumping laser. The split and measure XAS approach can be used to track the efficacy of the THz electric field switching of VO2- acting as a test case for the development of a feasible switching method for possible future applications.
5.2.1 Spatially resolving the MIT
The next step towards the ultimate goal of unravelling the transition both spatially and temporally simultaneously, relies on getting real, nm-scale spatially resolved images of the material during the transition. Holography experiments designed to do exactly this are underway. The concept is show in Fig. 5.1. The first static mea-surements have already been performed at the SEXTANTS beamline at the SOLEIL synchrotron in Paris, France. The analysis is ongoing, but preliminairy analyses show a resolution of 30nm was achievable. If these attempts are successful, ex-tending this experiment to the time-domain would involve porting the holography samples to the FEL and utilising the same sort of pump-probe as carried out for the XAS data presented here. If imaging using a single X-ray pulse - something already done for metallic magnets [31]- can be shown to work for VO2,then a nm length-scale movie of the Mott MIT with fs frame rate will be within reach.
