Laser-induced changes of nonlinear electronic transport properties in La_0.75Ba_0.25MnO₃ and (La_0.6Pr_0.4)_0.67Ca_0.33MnO₃

Perovskite manganites display a variety of intriguing physical phenomena such as ferro- and antiferromagnetic ordering, metal-to-insulator and structural phase transition as well as electronic phase separation [1, 2]. In their general formula ${R}_{1-x} {A}_{x} {\rm MnO_3}$, where $R = {\rm La}, {\rm Pr}, \ldots$ denote rare earth metals and $A = {\rm Ca}, {\rm Sr}, {\rm Ba}, \ldots$ alkaline earth metals, the doping level x gives the fraction of ${\rm Mn^{4+}}$ ions and 1  −  x the complementary portion of ${\rm Mn^{3+}}$. This ratio of ${\rm Mn^{3+}}/{\rm Mn^{4+}}$ ions and the cation radii of R and A determine the interplay of two competing interactions: one is double exchange [3], which leads to the delocalization of Mn e_g electrons and stabilizes a ferromagnetic metallic ground state, and the other is electron–phonon coupling originating from Jahn–Teller (JT) polaron [4, 5] formation favoring an insulating paramagnetic or antiferromagnetic state. Due to the nature of a JT polaron being an e_g electron accompanied by a lattice distortion in the form of an elongation of the ${\rm MnO_{6}}$ octahedron in one direction and a contraction in the other, multiples of these deformations can arrange themselves next to each other to further gain energy forming correlated polarons [6]. On the other hand, the Coulomb repulsion between the electrons limits the formation of correlated JT polarons [7–9] leading to long-range charge ordering for higher dopings, $x \sim 0.5$. There is strong evidence [10] that these correlations play an important role for the well known [11] colossal magnetoresistance (CMR) effect in connection with an electronic phase separation scenario [12].

The CMR is a classic example of the extreme susceptibility of manganites to different external stimuli which results in a drastic change of properties via the phase transition. Such stimuli include chemical doping, temperature, pressure [13], strain [14], magnetic [15] and electric [16, 17] fields as well as electromagnetic radiation, including light [18, 19]. For example, Rini et al [20] were able to drive an insulating ${\rm Pr_{0.7}Ca_{0.3}MnO_3}$ (PCMO) single crystal into a transient metallic state by means of a coherent laser excitation of a low energy JT phonon mode. Moreover, incoherent—with respect to JT phonons—electronic excitations [21, 22] with much higher photon energies in the ultraviolet, visible and near infrared spectral region have been shown to strongly affect electric transport and structural properties [23]. Thus, manipulating electron-lattice correlations by electronic transitions can have direct effects on the structure of a material as shown by DMFT calculations [24]. The photoinduced changes of electrical resistance were found to be mostly pronounced for small-bandwidth manganites, i.e. PCMO, at low temperatures, for which the electronically and structurally phase separated state can be destroyed or strongly modified by electromagnetic radiation [25].

Here we report the influence of laser excitation with ultra-short laser pulses in the near-infrared regime ($h \nu = 1.55~{\rm eV}$) on the nonlinear third harmonic voltage ($\newcommand{\Udromega}{{U_{3\omega}}} \Udromega$) of middle-bandwidth strongly correlated $\newcommand{\LBMObr}{\rm {La_{0.75}Ba_{0.25}MnO_3}} \LBMObr$ (LBMO) and $\newcommand{\LPCMObr}{{\rm (La_{0.6}Pr_{0.4})_{0.67}Ca_{0.33}MnO_3}} \LPCMObr$ (LPCMO) manganites. Both these compounds show a metal-to-insulator (MI) transition which is highly influenced by the strength of the electron–phonon coupling in the system. For LBMO the coupling is rather weak, which leads to an MI transition near room temperature, whereas in LPCMO the coupling is strong and thus the transition occurs at lower temperatures, $T \sim {200}~{\rm K}$ [26]. This coupling manifests itself in the formation of correlated JT polarons, which give rise to non-linear behavior in electric transport due to their quadrupolic nature. Namely, the third harmonic coefficient, $\newcommand{\Uomega}{{U_\omega}} \newcommand{\Udromega}{{U_{3\omega}}} \newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega = \Udromega / \Uomega$, was shown to represent the density of correlated JT polarons [27]. The laser excitation changes $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega$ near the transition temperatures, $T=200{\rm{-}}{300}~{\rm K}$, thus indicating a manipulation of electron–phonon coupling.

LBMO/${\rm SrTiO_3}$(1 0 0) and LPCMO/MgO(1 0 0) thin films were prepared by metalorganic aerosol deposition [28]. The films were microstructured to a bar with a width of $w=30~{\mu {\rm m}}$ and a length of $l= 120~\mu{\rm m}$ (LBMO) and $l=100~\mu{\rm m}$ (LPCMO), respectively, using electron beam lithography on a Raith Elphy-Plus and chemical wet etching. In addition, metal contacts (${30}~{\rm nm}$ Au on top of ${10}~{\rm nm}$ Cr) were deposited by thermal evaporation onto the bar to get a hall-bar geometry for electrical four-probe measurements. A nonlinear $3\omega$ measurement technique [27] was used with an ac current sent through the bar while the voltage drop was measured on the fundamental $\newcommand{\Uomega}{{U_\omega}} (\Uomega)$ and third harmonic frequency ($\newcommand{\Udromega}{{U_{3\omega}}} \Udromega$) by means of lock-in amplifiers to calculate the third harmonic coefficient $\newcommand{\Uomega}{{U_\omega}} \newcommand{\Udromega}{{U_{3\omega}}} \newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega = \Udromega / \Uomega$. Therefore the 5 V-sine output of a Stanford SR830 was used to provide a constant excitation current of ${0.5}~\mu{\rm A}$ at ${37}~{\rm Hz}$ over a 10 MΩ series resistor.

The photoexcitation was realized by two different Ti:sapphire lasers with the same photon energy $h \nu=1.55~{\rm eV}$ and wavelength of $\lambda \sim {800}~{\rm nm}$, but with different duration (τ) and repetition rates (f) of the pulses. The 'Femtolasers Femtosource fusion 20' provides pulses with $\tau = 120~{\rm fs}$ and $f = 74.9~{\rm MHz}$ repetition rate and a spot size of $d_{\rm spot} = 40~\mu{\rm m}$, yielding a relatively low pulse fluence, $F_{\rm l}=11~{\mu {\rm J~cm^{-2}}}$. The second laser 'Coherent RegA 9040' with $\tau = 45~{\rm fs}$ pulses, $f = 250~{\rm kHz}$ and a $d_{\rm spot} = 80~\mu{\rm m}$ results in a much higher fluence $F_{\rm h}=8~{\rm mJ~cm^{-2}}$. To measure the laser-induced effects in a wide range of temperatures, $T = 5{\rm{-}}400~K$, a customized inset with fiber-guided laser insertion for a Quantum Design physical properties measurement system (PPMS) was built for the first laser system. For the second, more powerful system, which demands a free beam configuration, a Peltier-cooled sample holder was built, enabling measurements at temperatures $T=250{\rm{-}}{300}~{\rm K}$.

3.1. Sample characterization

Firstly, we examined the general characteristics of the films to ensure their quality. X-ray diffraction (XRD) in Bragg–Brentano geometry ($\theta-2\theta$, Cu $K_\alpha$ radiation) confirms the perovskite crystal structure with a pseudo-cubic out-of-plane lattice constant $c({\rm LBMO})= 3.91~{\mathring{\rm A}}$ and $c({\rm LPCMO})=3.87~{\mathring{\rm A}}$ and film thicknesses $d({\rm LBMO})=30~{\rm nm}$ and $d({\rm LPCMO})=60~{\rm nm}$. Scanning tunneling microscopy (not shown here) reveals a smooth surface for both films with a mean-square roughness value ${\rm RMS} < {0.27}~{\rm nm}$ for LBMO and ${\rm RMS} < {1.6}~{\rm nm}$ for LPCMO. In figure 1 we present dc resistivity versus temperature curves used to map the MI transition, which is clearly visible at $\newcommand{\TmiB}{{T_{\rm MI}^{\rm (B)}}} \TmiB=280~{\rm K}$ for LBMO and at $\newcommand{\TmiPC}{{T_{\rm MI}^{\rm (PC)}}} \TmiPC=201~{\rm K}$ for LPCMO. Interestingly, the latter sample also shows a second maximum in resistivity at $\newcommand{\Tnd}{{T_{\rm 2nd}}} \Tnd=182~{\rm K}$, which was not present before the microstructuring of the LPCMO film. We interpret this feature to be a signature for a large scale phase separation effect [29]. For the second phase we assume a slightly different Pr-doping, which cannot be identified in the XRD pattern since the lattice constant only differs by less than ${0.01}~{\mathring{\rm A}}$. If we furthermore assume that this phase forms elongated spatial domains, electrical currents can circumvent these domains in the full film. In the structured sample this short-circuit is eventually inhibited. By comparing the 2nd maximum to its full phase expected value, which could be roughly calculated by extrapolating from the slope towards the main maximum, one gets an estimation of about  ∼10% fraction of the second phase. However, this second phase shows a stronger nonlinear behavior than the main transition as we will see later.

Zoom In Zoom Out Reset image size

3.2. Electronic transport

The actual study of laser-induced effects in manganite films was performed using the following protocol developed by us: (1) The desired temperature was set and we waited until it was stable; (2) measuring $\newcommand{\Uomega}{{U_\omega}} \Uomega$ and $\newcommand{\Udromega}{{U_{3\omega}}} \Udromega$ was started without laser excitation; (3) after ${50}~{\rm s}$ the laser with a pulse length of $\tau=120~{\rm fs}$ ($\tau=45~{\rm fs}$) and a repetition rate of $f=74.9~{\rm MHz}$ ($f=250~{\rm kHz}$) for the low (high) fluence was switched on; (4) ${150}~{\rm s}$ later the excitation was switched off, meaning that during this period the sample was actually illuminated for only ${1.3}~{\rm ms}$ or ${1.7}~\mu{\rm s}$, respectively, due to the pulsed nature of the laser beam; (5) after further ${200}~{\rm s}$ the measurement was stopped. The levels of the saturated signal for 'laser off' and 'laser on' states were extracted and assigned to the set temperature, resulting in the corresponding $\newcommand{\Uomega}{{U_\omega}} \Uomega(T)$ and $\newcommand{\Udromega}{{U_{3\omega}}} \Udromega(T)$ dependencies. We have shifted the 'excited' $\newcommand{\Uomega}{{U_\omega}} \Uomega(T)$ up to higher temperatures so that it matches the 'unexcited' curve. This was necessary to account for the unavoidable heating of the sample caused by a repetitive laser light absorption during the continuous electric measurement. Although the T shift was determined empirically, a simple finite-element simulation (not shown here) was also carried out using COMSOL Multiphysics [30] to confirm it. Our model describes the steady state temperature distribution in an unstructured manganate film on top of an STO substrate, subject to local heating at the surface. The relevant parameters for heat transport (density, specific heat, thermal conductivity) were taken from [31]. Experimental laser heating was implemented as a spatially distributed heating source, reflecting the properties of the laser system in Greifswald. This simulation confirmed the empirically determined T shift and this shift was then also applied to $\newcommand{\Udromega}{{U_{3\omega}}} \Udromega(T)$. This was done for all temperatures from ${300}~{\rm K}$ down to $\sim {250}~{\rm K}$ for LBMO and down to ${170}~{\rm K}$ for LPCMO using the PPMS.

Figure 2 shows the data on LBMO with the two different fluences used. Both $\newcommand{\Uomega}{{U_\omega}} \Uomega$ and $\newcommand{\Udromega}{{U_{3\omega}}} \Udromega$ display a pronounced maximum at the MI transition and one can clearly see that $\newcommand{\Uomega}{{U_\omega}} \Uomega$, after temperature compensation, is not influenced by the laser excitation for both intensities. In contrast, $\newcommand{\Udromega}{{U_{3\omega}}} \Udromega$ is slightly increased in the case of the small fluence and is evidently decreased under irradiation with a much higher fluence. Thus, the third harmonic coefficient $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega = 3.5\times 10^{-3}$ at $\newcommand{\TmiB}{{T_{\rm MI}^{\rm (B)}}} \TmiB$ is also changed correspondingly by  ∼+1% and  ∼$-{2.5}{\%}$, for small and large fluence respectively.

Zoom In Zoom Out Reset image size

The LPCMO film could only be measured with low fluence since the required low temperatures were available within the PPMS setup only. The results from these measurements are shown in figure 3. Here the maximum in $\newcommand{\Udromega}{{U_{3\omega}}} \Udromega$ is located at $\newcommand{\Tnd}{{T_{\rm 2nd}}} \Tnd$ which corresponds to the second maximum in $\newcommand{\Uomega}{{U_\omega}} \Uomega$. Additionally there is also a small maximum in $\newcommand{\Udromega}{{U_{3\omega}}} \Udromega$ at $\newcommand{\TmiPC}{{T_{\rm MI}^{\rm (PC)}}} \TmiPC$, but it is by a factor of 50 smaller than the main one. Again, the laser excitation does not change $\newcommand{\Uomega}{{U_\omega}} \Uomega$, even the heating was found to be very small ($\sim {0.1}~{\rm K}$), but in $\newcommand{\Udromega}{{U_{3\omega}}} \Udromega$ a clear effect at $\newcommand{\Tnd}{{T_{\rm 2nd}}} \Tnd$ is visible in $\newcommand{\Udromega}{{U_{3\omega}}} \Udromega$. Expressed in terms of $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega$, which is about $2\times 10^{-2}$ at $\newcommand{\TmiPC}{{T_{\rm MI}^{\rm (PC)}}} \TmiPC$, this implies a laser-induced decrease of the polaron density by about ${2.5}{\%}$.

Zoom In Zoom Out Reset image size

The small increase of $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega$ in the case of LBMO with small fluence could be originating from an incoherent excitation of JT breathing modes similar to the coherent stimulation of these modes observed by Rini et al [21]. Even though the energy of this mode $E_{\rm JT} = 71~{\rm meV}$ is much smaller than the excitation energy $h \nu=1.55~{\rm eV}$, some part of the photon energy could have been transferred from the initially excited electron system via dissipative processes into the JT optical phonon modes because of strong electron–phonon coupling (see figure 4). Being close to the phase transition enables a large susceptibility to the enhancement of the short-range polaronic order mediated by this activation of JT phonon modes. This would explain the rise in $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega$ caused by the laser.

Zoom In Zoom Out Reset image size

On the other hand, with high laser fluences one could melt the correlated polarons in LBMO and thus decrease $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega$. This can be understood by looking at the electronic structure, in which due to the crystal field and JT splitting the fivefold degeneracy of the 3d electrons is lifted and two separated e_g levels are formed. The lower level is occupied by one electron in the ${\rm Mn^{3+}}$—where JT distortion is favored—and can be excited with ${1.55}~{\rm eV}$ photons to the upper level (see figure 4) [21, 22]. In this case, the distortion no longer gains energy and so the JT polarons and the electron-lattice correlations are destroyed.

Another mechanism could be that the free upper e_g level is populated from the low-lying O-2p states. In fact, the photon energy, $h \nu = 1.55~{\rm eV}$, is well above the typical JT e_g splitting of $\Delta_{\rm JT} \sim {0.6}~{\rm eV}$. However, even though this process is possible, it should not be the predominant one. As DFT calculations for the band structure of the related system PCMO show [22, 32], this excitation requires photon energies of $h \nu > {2.5}~{\rm eV}$ to be significant. Anyway, such an excitation would induce a change in the Mn valence, which in turn destroys the energetic balance between Coulomb repulsion and JT distortion. Thus, one can expect a destruction of correlated polarons as a consequence.

For the measurements of LPCMO, our model also explains the findings, except that there the lower fluence is already sufficient to break the polaronic state. This can be understood by taking into account the stronger electron–phonon coupling [26] and thus larger polaron density, wich in turn increases the probability of polaron melting for LPCMO in comparison to LBMO.

There are two important points to be accounted for in all measurements presented here. First, the spot sizes of the lasers do not illuminate the whole bar, leaving some unexcited parts, which therefore keeps its unaltered response in $\newcommand{\Udromega}{{U_{3\omega}}} \Udromega$. So the changes in $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega$ only originate from the area below the laser spot. Because the contribution from the excited and unexcited parts to $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega$ add up in a serial manner and are proportional to their areas, we will now scale the measured effect with the ratio of these areas. This enables us to estimate the impact on the illuminated part only. Thus, for the LBMO bar ($l=120~\mu{\rm m}$) we get a rescaled change in $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega$ of $+{3}{\%}$ for the low fluence ($d_{\rm spot} = 40~\mu{\rm m}$) and $-{3.75}{\%}$ for the high fluence ($d_{\rm spot} = 80~\mu{\rm m}$). For the LPCMO bar ($l=100~\mu{\rm m}$) the rescaling yields $-{6}{\%}$ for the low fluence.

The second point is that, due to the ultra-short pulses and the moderate repetition rate of the lasers, the sample was illuminated for a time fraction of about 10⁻⁵ for low fluence and 10⁻⁸ for high fluence during the measurement time. Note that the electric measurements of $\newcommand{\Uomega}{{U_\omega}} \Uomega$ and $\newcommand{\Udromega}{{U_{3\omega}}} \Udromega$ are continuous and the observation of a measurable time-averaged photoinduced signal depends solely on the lifetime $\tau{\rm e}$ of the excitations and the pulse period $\Delta t = 1/f$. On short timescales directly after photoexcitation, we expect a much larger change to $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega$ than the observed values. Let us first have a look at the high fluence case for LBMO, where we had a repetition rate $f = 250~{\rm kHz}$, which corresponds to $\Delta t = 4~\mu{\rm s}$. From previous studies [33] it is known that $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega$ can be reduced by static modification from 10⁻¹ to 10⁻⁴. Taking this relative change as an initial suppression of $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega$ in our experiment as well, we deduce a lifetime of $\tau_{\rm e} \sim {150}~{\rm ns}$ of the photoexcited state. A lifetime of the same order of magnitude was also found by Fiebig et al [34] in their investigations on the reflectivity dynamics of PCMO.

If we also assume this lifetime for the experiments on LBMO and LPCMO conducted with the small fluence laser system, we can now deduce the initial change in $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega$ directly after laser excitation for those experiments. Taking into account that the pulse period, $\Delta t = 13.4~{\rm ns}$, is by a factor of 11 smaller than the assumed lifetime, we calculate an initial change of $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega$ of $+{3.1}{\%}$ (LBMO) and ${-6.3}{\%}$ (LPCMO) for the low fluence measurements. This is in agreement with our experimental results. Note that compared to the high fluence measurements discussed above, one can understand the much smaller initial change in $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega$ due to the three orders of magnitude smaller pulse fluence.

We have shown the influence of laser light on correlated JT polarons near the phase transition for thin films of LBMO and LPCMO with relatively low and strong electron–phonon coupling, respectively. The stabilization or melting of correlated polarons, depending on the manganite system and the excitation fluence, was demonstrated. Rescaled to the known possible change in $\newcommand{\Kdromega}{{K_{3\omega}}} \Kdromega$, we could calculate a lifetime of ${150}~{\rm ns}$ for the photoexcited polaronic state.

We would like to thank the DFG for financial support through SFB 1073 TP B01.