betaFeedback

WorksPhysics and Astronomy

Initial electron thermalization in metals measured by attosecond transient absorption spectroscopy

Abstract

Understanding initial electron thermalization has relevance to both fundamental scientific knowledge and application to the construction of novel devices. In this study, attosecond transient absorption is used to directly measure initial electron thermalization times of 38 ±\pm 8 fs, 15 ±\pm 3 fs, 4.2 ±\pm 1 fs, and 2.0 ±\pm 0.3 fs for Mg, Pt, Fe, and Co, respectively. Through time dependent density function theory calculations, it is shown that the fast electron thermalization observed in Fe and Co is correlated with a strong local field effect. We find that a simple analytical model can be used to calculate the initial electron thermalization time measured by the transient extreme ultraviolet absorption spectroscopy method performed here. Our results suggest that the most significant contributions to the initial electron thermalization times are the basic metal properties of the density of states volume available for scattering and screened electron interaction. Many-body effects contribute less, but still significantly to the initial electron thermalization time. Ultimately the information gained through this study shows the unique view that attosecond transient absorption spectroscopy contributes to unraveling and monitoring electron dynamics and its connection to many-body effects in metals and beyond.

Introduction

What processes contribute to non-equilibrium electron thermalization has been a long-standing question for many-body physics. The nature of electron thermalization is not only a fundamental question, but it is of interest for device applications. It has been shown that spatial gradients created by non-equilibrium electron distributions result in transient behavior, such as ultrafast magnetic phenomena.[1, 2] Additionally, energy transport of non-equilibrium electrons can be up to two orders of magnitude greater than their equilibrium equivalents.[3, 4] These unique properties driven by electron thermalization are significant for the development of novel devices ranging from magnetic storage devices to hot carrier solar cells.

In the pursuit of understanding electron thermalization, metals are the natural starting place as they have the simplest material properties and have a large range of well-developed theory available. The photoexcitation process in metals promotes electrons from states below the Fermi level to above it, leaving these photoexcited electrons in a highly non-equilibrium, nascent, distribution. This nascent distribution decays on the femtosecond timescale to a hot-Fermi Dirac distribution through electron-electron collisions.[5] Finally electrons return to equilibrium by transferring their energy to phonons, defects, surfaces, or impurities. Some of the main factors dictating the efficiency and therefore the timescale of the initial electron thermalization, i.e. the transition from the nascent to hot Fermi-Dirac distribution, are the number of states available to facilitate electron-electron scattering, the strength of coupling between scattering partners, and the temperature of the non-excited scattering partners.[6, 7]

From an experimental perspective, very few techniques operate on a fast-enough timescale to directly measure the initial electron thermalization, which can occur within a femtosecond of excitation.[7] Two-photon photoemission (2PPE) is a common spectroscopic technique for measuring the lifetime of an electron in an excited state, however the final state of the probe electron is outside the material, which makes the probe blind to material responses related to electron-electron interactions. 2PPE is also limited to a low excitation density in order to avoid space charging of ejected photoelectrons, and the method is subject to a dependence between measured electron lifetimes and the temporal duration of the excitation pulses employed.[8] Although valuable insight has been gained through 2PPE studies, knowledge of non-equilibrium electron thermalization could greatly benefit from another technique, such as attosecond transient extreme ultraviolet light (XUV) absorption spectroscopy (ATAS), which probes the bulk of the material, is robust to high excitation densities, and has few femtosecond or subfemtosecond time resolution.

Previous ATAS studies in metals have measured the fluence dependence of electron thermalization times in Ni, showed the influence of the local field effect on spectral features and charge localization in Ti, and observed broadening of the electron distribution and a shift of the chemical potential in Al.[9, 10, 11] In this work we broaden previous 2PPE and ATAS studies by using ATAS to measure the initial electron thermalization time in four metals (magnesium, platinum, iron, and cobalt) in order to gain insight into the fundamental relationship between the measured initial electron thermalization time, many body effects, and material properties.

This is achieved by first quantifying the static XUV absorption spectra, the changes in XUV absorption upon optical excitation, and the electron thermalization times with ATAS measurements. Next spectral line shape decomposition, provided through the ab-initio modeling, is used to unravel the relevant excited state population changes and many-body interactions that occur after photoexcitation in order to correlate these effects with the experimentally measured initial thermalization times.[12] Finally, the measured initial electron thermalization times are compared to thermalization times calculated by a simple analytical model to provide a means of estimating the thermalization time without the need for measurements and to gain insight into the factors that contribute to the thermalization time.

Results and Discussion

The metal films studied here were grown by thermal evaporation onto silicon nitride membranes (Appendix § 3). The transient XUV absorption apparatus is described in detail in Appendix § 4. The static XUV absorption spectrum was used to determine film thickness (Figure 1).

Image not extracted

Figure 1. Equilibrium XUV absorption at the Mg L2,3L_{2,3}–edges (A), Pt N7N_{7}–edge (B), Fe M2,3M_{2,3}-edges (C) and Co M2,3M_{2,3}-edges (D). The dashed lines indicate the energies of the expected static absorption edges from previous x-ray absorption measurements (in some cases with a large spin-orbit splitting).[13, 14, 15, 16]

In these measurements, the XUV absorption of the silicon nitride membrane was subtracted from the sample in order to measure the XUV absorption of the metal alone. Using tabulated XUV absorption data for Mg, Fe, and Co, and a calibrated sample for Pt, it was determined that the Mg, Pt, Fe, and Co films were 130±1130 \pm1 nm, 35±135 \pm1 nm, 18±118 \pm1 nm, and 9±29 \pm2 nm thick, respectively (Appendix § 4). The XUV light used in absorption measurements was produced through the process of high harmonic generation (HHG) in argon gas driven by 4 fs, 1.3 eV – 2.6 eV pulses at a 1 kHz repetition rate (Appendix § 5). The resulting attosecond pulse train of XUV light continuously spanned from 20 eV - 80 eV. In the transient experiments an optical pump with the same bandwidth as the HHG driving field and a fluence of 20 mJ cm−2^{-2}, 23 mJ cm−2^{-2}, 29 mJ cm−2^{-2}, and 21 mJ cm−2^{-2} was used to excite charge carriers in Mg, Pt, Fe, and Co, respectively. The density of excited charge carriers was 2.7×10202.7 \times10^{20} cm−3^{-3}, 1.1×10211.1 \times10^{21} cm−3^{-3}, 1.9×10221.9 \times10^{22} cm−3^{-3}, and 1.7×10221.7 \times10^{22} cm−3^{-3} for Mg, Pt, Fe, and Co, respectively (Appendix § 11). The resulting change in absorption was then calculated for increasing time delays between the optical pump and XUV absorption probe, where negative time delays correspond to the pump arriving after the probe.

The resulting change in absorption, dA=−log⁡Ipump onIpump offdA = -\log\frac{I_{\mathrm{pump\ on}}}{I_{\mathrm{pump\ off}}}, was then calculated for increasing time delays between the optical pump and XUV absorption probe. Here, negative time delays correspond to the pump arriving after the probe and positive time delays correspond to the optical pump arriving before the probe. Further detail on the experimental methods can be found in Appendix § 5.

The transient absorption spectra (Appendix E) were integrated over the time ranges of 60 – 70 fs, 35 – 45 fs, 15 – 25 fs, and 2 – 12 fs for Mg, Pt, Fe, and Co, respectively, to obtain the transient absorption line shapes (Figure 2, panels A – D). All four metals display a large positive change in absorption at energies below the energy of their respective static absorption edges (Figure 2 dashed lines). This positive change in absorption has been observed in transient XUV absorption spectra of other metals and has been previously identified as a state filling or state blocking change in the spectra. [9, 10, 11, 12, 17, 18] At energies directly above the Mg L3L_3-edge (49.65 eV) and the Pt N7N_7-edge (71.2 eV) a negative change in absorption is observed (Figure 2 A, B), while the change in absorption remains positive at energies above the Fe M2,3M_{2,3}-edge (52.7 eV) and Co M3M_3-edge (58.9 eV) (Figure 1 C, D). The origin of the spectral features above the static absorption edge energy contain contributions from state filling and many-body effects. Both the state filling / opening contributions and many body effects are discussed in more detail later in section B.

Image not extracted

Figure 2. Panels A – D show the spectral line shape of the change in ΔA\Delta A for Mg, Pt, Fe, and Co, respectively. The dashed lines correspond to the energy of the static absorption edges.

Temporal evolution of the state filling feature

To show the evolution of the state opening feature in the time domain, the transient absorption spectra were integrated directly below the static absorption edge; 49.2 eV – 49.7 eV, 70.5 eV – 71.3 eV, 51.8 eV - 52.6 eV, and 56.5 eV – 65.2 eV for Mg, Pt, Fe, and Co, respectively (Figure 3, points). The inset in Figure 3 compares Co (gray) and Fe (green) from -15 to 15 fs in order to better visualize the difference in the temporal response for these two metals. The experimental data were fit with an exponentially modified Gaussian function (Figure 3, solid lines) to determine rise times of 38±838 \pm8 fs, 15±315 \pm3 fs, 4.2±14.2 \pm1 fs, and 2.0±0.32.0 \pm0.3 fs for Mg, Pt, Fe, and Co, respectively. More details on the integration and fitting can be found in Appendix § 13.

Image not extracted

Figure 3. Integrating the change in absorption of the positive portion (newly created absorption) of the state-blocking feature (points) shows the dramatic difference in response to optical pumping for Mg (red), Pt (blue), Fe (green), and Co (gray). This is then fit with an exponentially modified Gaussian function to quantify the rise time of the signal (solid lines). The slight difference in rise time between Fe and Co is highlighted in the inset.

Discerning contributions to the transient absorption spectral lineshape with TDDFT

Determining the factors that contribute to the transient absorption lineshape aid in understanding the underlying processes in the metal that leads to the dramatic difference observed in the timescales of the state opening feature. Figure 4 shows time dependent density functional theory (TDDFT) computation results for three spectral aspects, the state filling / opening component (orange solid line), local field effect (LFE) component (purple solid line), and the change in absorption coefficient (blue solid line) for Mg (Figure 4A) and Co (Figure 4B). The full calculated change in absorption coefficient (blue) is to be compared to the experimental transients for Mg (red dashed line) and Co (gray dashed line). More detail on the TDDFT calculations can be found in Appendix § 14.

Image not extracted

Figure 4. TDDFT calculations of the full change in absorption coefficient (blue), state filling contribution (orange), and local field effect contribution (purple) show that the measured change in absorption lineshape of the change in absorption signal for Mg (dashed red line) comes primarily from state filling / opening (A), while both state filling / opening and the local field effect contribute to the measured change in absorption lineshape of Co (dashed gray line) (B).

The positive feature of the state filling / opening contribution refers to an observation of a non-equilibrium distribution of holes and electrons in the XUV absorption probing. When the optical pump promotes electrons from below to above the Fermi level new transitions are opened to the XUV absorption probe below the static XUV absorption edge and previously available transitions above the static XUV absorption edge are blocked. This is viewed as a characteristic increase followed by a decrease in the change of absorption lineshape versus photon energy, centered around the static XUV absorption edge.

The LFE is a many-body effect, where a charge carrier is influenced by the electric field created by other nearby charges.[19, 20, 21] Here, charge carriers excited by the optical pump couple with other excited carriers, resulting in a collective excitation mode. The LFE shifts the dielectric function to higher energy, thereby moving the XUV absorption edge to higher energy.[22, 23, 24, 25] This is seen as a negative followed by a positive change in absorption versus photon energy centered around the static XUV absorption edge.

The experimentally measured transient absorption lineshape of Mg is well-described by the calculated state-filling / opening feature (Figure 4A), while the experimentally measured lineshape of Co includes contributions of both state filling / opening and the LFE (Figure 4B). The LFE in Co reduces the intensity of the positive change in absorption attributed to state opening and the negative portion of the state filling contribution is completely masked by the LFE contribution. The change in absorption lineshape of Mg and Co is reasonably well matched by the TDDFT calculations, however there are two notable differences between the full TDDFT calculations and measurements. First, the calculated spectra are broader than the measurements. This possibly stems from the broadening parameter in the TDDFT calculations overestimating the experimental linewidth. The TDDFT calculations may also overestimate the LFE. For Mg this suggests that the small calculated LFE component may be negligible. Co has a much stronger LFE component, so the overestimation in the TDDFT calculations leads to the 60.3 eV peak carrying more spectral weight than the 58.5 eV peak. Despite these two differences, the trend that ATAS measurements of Mg can be viewed primarily as a direct observation of state filling / opening whereas Co includes both state filling and LFE contributions still holds.

Due to computational time and complexity, the line shape contributions were not calculated for Pt or Fe; the results of Mg and Co can in principle be extended to these metals. The experimental line shape of Pt (Figure 2B) is similar to Mg, showing that state filling / opening is the major contributing factor for both Pt and Mg. In similar fashion, the lineshape of Fe (Figure 2C) follows the line shape observed in Co, suggesting that both state filling / opening and LFE contribute to the line shape of these metals. Therefore, it can be concluded that state filling / opening contributes to the transient absorption line shape of all four metals and the LFE additionally contributes to the line shape of Fe and Co.

The spectral analysis confirms the earlier assertion that changes in the electron occupation (the state opening feature) were temporally quantified in Figure 3. Therefore, the timescales reported here correspond to the initial electron thermalization. It also indicates a striking correlation between a faster initial electron thermalization time and a strong LFE. This suggests that intrinsic properties of the metals are responsible for the variety of initial electron thermalization times that are observed.

Calculating the initial electron thermalization time

Previous theoretical work developed for 2PPE measurements on metals correlated the initial electron thermalization time with intrinsic properties of the metals.[6, 26, 27] From Fermi liquid theory, the thermalization rate of a single excited electron is[23, 24, 26, 27]

1τ=A(πkBTe)2+E21+e−EkBTe(1)\frac{1}{\tau}=A\frac{\left(\pi k_{\mathrm{B}}T_{\mathrm{e}}\right)^2+E^2}{1+e^{-\frac{E}{k_{\mathrm{B}}T_{\mathrm{e}}}}} \tag*{(1)}

where τ\tau is the thermalization time of the excited electron, AA is the characteristic electron scattering constant, kBk_{\mathrm{B}} is the Boltzmann constant, TeT_{\mathrm{e}} is the electron temperature of the electron bath (the non-excited electron scattering partners), and EE is the energy of the excited electron relative to the Fermi level (i.e. EF=0E_{\mathrm{F}} = 0 eV). Several transient spectroscopy studies of metals have shown that a larger electron temperature (higher excited charge carrier densities) lead to faster electron thermalization for a particular metal.[9, 28] The electron temperatures for the experimental conditions used here are 1300 K, 2300 K, 3700 K, and 3400 K for Mg, Pt, Fe, and Co, respectively (Appendix H). However, because kBTe≪Ek_{\mathrm{B}}T_{\mathrm{e}} \ll E for all of the metal films eq. (1) reduces to

1τ≈AE2.\frac{1}{\tau}\approx A E^2.

Therefore, the difference in electron temperatures between the metals films does not significantly affect the measured thermalization times in this study.

Zarate et al. used a perturbative scattering view of electron thermalization, where the momentum of the initial and final electron states is neglected (the random-kk approximation), to formulate AA for the free electron gas (FEG) as[6]

A=2πℏρ3∣Meff∣2(2)A=\frac{2\pi}{\hbar}\rho^3\left|M_{\mathrm{eff}}\right|^2 \tag*{(2)}

Here, ρ\rho is the density of states (DOS) at the excitation energy, and MeffM_{\mathrm{eff}} is an effective screened electron interaction between the excited electron and scattering partner.

The DOS used here was computed using the open-source package Quantum ESPRESSO (Appendix G).[29] The effective screened electron interaction reduces the actual screened electron interaction by assuming that this matrix element does not vary with energy, spin, or momentum.[6, 7, 30, 31, 32, 33, 34] It has been previously shown that neglecting the energy and momentum dependence is valid below 3 eV.[6, 26, 27, 30] The absorption measurements performed here are not sensitive to electron spin, which justifies averaging the spin components of MeffM_{\mathrm{eff}} for Fe and Co.

For Pt, Fe, and Co we use previously determined MeffM_{\mathrm{eff}} values.[31, 34] To the best of our knowledge, MeffM_{\mathrm{eff}} has not been measured or calculated for Mg, however because the DOS for Mg closely aligns with the FEG DOS (Figure 5), we use the FEG calculation of MM.[6, 36]

Image not extracted

Figure 5. The full and partial density of states for Mg (A), Pt (B), Fe (C), and Co (D) was calculated using DFT and shows the variety of structures of the four metals studied here. The red dashed line shows the free electron gas density of states and the pink dashed line marks the Fermi level position.

∣MFEG∣2=3π1281ρ3ℏωpEF2(3)\lvert M_{\mathrm{FEG}}\rvert^2 = \frac{\sqrt{3}\pi}{128}\frac{1}{\rho^3}\frac{\hbar\omega_p}{E_F^2} \tag*{(3)}

where ωp\omega_p is the plasma frequency, and EFE_F is the Fermi level energy.

Table 1 shows the calculated thermalization time for each metal, using equations (1)–(3). More detail on the variables used in these calculations can be found Appendix § 1. The calculated thermalization times are on the same order of magnitude and follow the experimentally measured trend of τMg>τPt>τFe>τCo\tau_{Mg} > \tau_{Pt} > \tau_{Fe} > \tau_{Co}. While the calculated thermalization time for Mg and Pt fall within the error of the experimentally determined thermalization times, the calculated value for Fe is twice as large as the measured value and

that calculated for Co is twice as small as the measured value.

In order to explain the difference between the measured and calculated thermalization times for Fe and Co, the three variables relevant to the calculation of the electron thermalization time, the electron temperature, DOS, and the effective screened electron interaction, are considered. We have already shown that the electron temperature does not significantly affect the results presented here. Previous studies have suggested that the total volume of the DOS participating in scattering principally affects the electron thermalization time, although some of these studies assume the same MeffM_{\mathrm{eff}} values for different metals in the process of showing the importance of the DOS.[30, 33] The constant volume approximation taken here underestimates the total volume of states participating in scattering, which would overestimate the thermalization time. This could explain the deviation observed in Fe, but not that of Co.

For Co, it is more likely that the assumptions made in determining MeffM_{\mathrm{eff}} are responsible for the disparity in the calculated and measured lifetimes. The factors neglected in the effective screened electron interaction are electron spin, energy, momentum, electron localization, and

many-body effects, all of which have been attributed to discrepancies between calculated and measured lifetimes and could contribute to either over or underestimating MeffM_{\mathrm{eff}} depending on the experimental conditions and specific properties of the metal.[30, 35] As the constant DOS assumption does not explain the deviation observed in Co, we believe that MeffM_{\mathrm{eff}} is most likely responsible for the calculated electron thermalization time being underestimated for Co.

Although we cannot exclude the MeffM_{\mathrm{eff}} of Fe contributing to the observed overestimation of the calculated initial electron thermalization lifetime, we favor the simpler explanation that the constant DOS assumption is responsible for the variation observed in Fe. It is beyond the scope of this work to determine which assumptions and excluded factors (i.e. many-body effects, electron spin, electron localization, etc.) are responsible for the differences between the measured and calculated lifetimes. However, this does reflect the importance and difficulty of choosing an appropriate MeffM_{\mathrm{eff}}, which has already been noted by Zhukov and Chulkov, and it demonstrates the need for further theoretical investigation into these effects, particularly on ferromagnetic metals. [31]

Earlier it was observed that Fe and Co both have LFE contributions in their transient XUV absorption lineshape and a faster initial electron thermalization times. This implies that either the LFE influences the initial electron thermalization time or that the basic properties of the metals lead to both a strong LFE and fast initial electron thermalization time. Through the application of the model described above, we get further insight into the correlation between the electron thermalization time and LFE strength. Table 1 shows that the DOS and MeffM_{\mathrm{eff}} values are larger as the thermalization time gets faster, with Pt being a notable exception. This suggests that the specific metal properties of the screened electron interaction and the volume of DOS participating in electron scattering are principally responsible for a faster thermalization time, not the LFE. On the other hand, many-body effects, which were shown in section B to have significant influence in Fe and Co, were neglected in the formulation of eq. (1)-(3). This suggests that the exclusion of the LFE could be responsible for the significant deviations between the experimental and calculated initial thermalization times. Additionally, it is possible that the disparity between calculated and measured thermalization time for Fe and Co potentially comes from the exclusion of many-body effects, such as the LFE, in calculating MeffM_{\mathrm{eff}}. This implies that the LFE might influence the thermalization time to a much lesser degree than the DOS volume and other factors used in calculating MeffM_{\mathrm{eff}}. We believe that the most significant contributions to the initial electron thermalization times are the basic metal properties of the DOS volume available for scattering and screened electron interaction and that many-body effects like the LFE contribute less, but still significantly to the initial electron thermalization time.

Measured therm. time (fs)ρ\rho (eV−1^{-1})∣Meff∣2\lvert M_{\mathrm{eff}}\rvert^2 (eV2^2)Calculated therm. time (fs)
Magnesium38±838 \pm80.460.0132
Platinum15±315 \pm30.270.1017
Iron4.2±14.2 \pm10.350.088.7
Cobalt2.0±0.32.0 \pm0.30.340.641.1

Table 1. The thermalization time was calculated for all four metals using equations 1-3. The parameter ρ\rho is equal to the calculated DOS at the central optical pump wavelength of 1.65 eV and the effective screened electron interaction, MeffM_{\mathrm{eff}}, for Fe and Pt come from Zhukov et al, Co comes from Knorren et. al and Mg was calculated using eq. 3.[30, 33]

There are several assumptions made in the formulation of eq. (1)-(3) that are worth commentary. First, they originate from the view of a single electron excited to a particular energy scattering with a sea of non-excited electrons to reach initial thermalization. The experimental reality contrasts with this simplified view. Experimentally the optical pump excites many electrons (102010^{20} – 102210^{22} electrons cm−3^{-3}) within an energy range dictated by the energy of the broad band optical pump. This results in the equations neglecting the interaction between excited electrons, which we showed contributes to the transient absorption line shape of Co and Fe. It also necessitates the use of an average energy of the ensemble of excited electrons.

Defining an electron temperature for the results also presents a challenge. Directly after optical excitation the excited electrons are in a highly non-equilibrium distribution with no defined temperature. The difficulty in reconciling the experimental reality of no electron temperature with the inclusion of an electron temperature in the equations has led to the commonly used approximation of taking the electron temperature of the hot Fermi-Dirac distribution (formed after initial electron thermalization) as the electron temperature of the non-excited electron bath directly after optical excitation.[9, 26, 28, 37, 38] Efforts have been made in the literature to circumvent this known inaccuracy by calculating the full Boltzmann collision integrals for laser excited metals, however this requires in-depth theoretical calculations that are dependent on both the metal properties and experimental conditions, which make this method difficult to generalize.[26] For the calculations presented here we follow the approximation that the electron temperature is equal to the hot Fermi Dirac distribution established after initial electron thermalization. For each metal kBTe≪Ek_{\mathrm{B}}T_{\mathrm{e}} \ll E which reduces eq.1 to 1τ≈AE2\frac{1}{\tau} \approx A E^{2}. This shows that the difference in electron temperature between the metals does not influence the measured lifetimes.

Despite the application of the single-electron equations to the laser-excited ensemble of electrons and the average values used for the excited electron energy, electron bath temperature, the density of states, and screened electron interaction, the calculated initial thermalization times of Mg and Pt agree well with the experimentally measured initial thermalization times and the general trend of τMg>τPt>τFe>τCo\tau_{Mg} > \tau_{Pt} > \tau_{Fe} > \tau_{Co} is reflected in these calculations. This shows that these simplified, single-electron equations can be used to describe the initial thermalization of an ensemble of laser-excited electrons. Because the quantities used to calculate the initial thermalization times are either basic material quantities (density of states, plasma frequency, Fermi level energy), readily available from measurements or calculations (MeffM_{\mathrm{eff}}), and experimentally controlled values (electron bath temperature and excited electron energy), these results can be easily extended beyond the measurements presented in this study, although the limits of the application have not been fully tested in this work.

Conclusion and Outlook

In summary, ATAS was used to determine an initial electron thermalization time of 38 ±\pm 8 fs, 15 ±\pm 3 fs, 4.2 ±\pm 1 fs, and 2.0 ±\pm 0.3 fs, for Mg, Pt, Fe, and Co, respectively. TDDFT decomposition of the contributions to the transient absorption lineshape of Mg and Co showed that the LFE is strong in Co, and by extension Fe, and that the LFE is weak in Mg, and by extension Pt. Single-electron equations developed from Fermi-liquid theory within the random-kk approximation were used to calculate initial thermalization times of 32 fs, 17 fs, 8.7 fs, and 1.1 fs for Mg, Pt, Fe, and Co, respectively.

There are two main outcomes from this study. First, due to the agreement between the theoretical and measured electron thermalization time and the simplicity of the model used here, this calculation can easily be applied to other metals and potentially to more complicated systems. Second, our results suggest that the initial electron thermalization time of metals is primarily affected by the volume of DOS participating in electron scattering and the screened electron interaction, and that many-body effects, such as the LFE, has a smaller, but significant affect. Ultimately the information gained through this study shows the unique view ATAS contributes to unraveling and monitoring electron dynamics their connection to many body effects in metals and beyond.

Acknowledgements

Investigations at the University of California, Berkeley were supported by the U. S. Air Force Office of Scientific Research (AFOSR) Grant No. FA9550-19-1-0314 (primary), FA9550-20-1-0334, FA9550-22-1-0451 (DURIP), and the W. M. Keck Foundation No. 046300. Lorenz Drescher acknowledges the European Union’s Horizon research and innovation program under the Marie Sklodowska-Curie grant agreement No. 101066334 – SR-XTRS-2DLayMat. Shunsuke A. Sato acknowledges the JSPS KAKENHI Grant Number JP20K14382.

Thin Film Growth

Physical vapor deposition was used to create the metal films. Silicon nitride membranes 30 nm thick (Norcada), which are optically transparent to the XUV probe, were chosen as the deposition substrates. The base pressure in the deposition chamber was ∼10−7\sim10^{-7} Torr. For Pt and Co an ∼1\sim1 nm/hr growth rate was achieved by resistively heating a high purity wire (Kurt J. Lesker Company) to 1330 °C – 1348 °C and 1050 °C – 1100 °C, respectively. For Fe and Mg, high purity pellets (Kurt J. Lesker Company) were flash evaporated by a resistively heated sapphire basket that contained the pellet. This method achieved much higher deposition rates of ∼250\sim250 nm/hr and ∼12\sim12 nm/hr for Mg and Fe, respectively.

Calculating Film Thickness

The XUV absorption measurements without optical pump illumination were used to determine the film thickness of each metal. This is possible because absorbance is directly proportional to the amount of material being probed in the limit of weak absorption. In order to isolate the XUV absorption of the metal film alone, the XUV absorption from a clean 30 nm silicon nitride membrane was subtracted from the XUV absorption of the sample. For Co, Fe, and Mg the Center for X-ray Optics X-ray solid filter transmission database was used to determine the relationship between material thickness and absorbance at the relevant x-ray absorption edge.[39] For example, the transmissions of Fe around the Fe M2,3M_{2,3} – edge (52.7 eV) for Fe films of 5 nm, 10 nm, 25 nm, 50 nm, 75 nm and 100 nm were calculated using the Center for X-ray Optics X-ray solid filter transmission data base (Figure 6).[39] The transmission data were converted to absorbance by taking the negative log of the calculated transmission. A linear regression between the calculated absorbance and film thickness was then used to calculate an Fe film thickness of 18 nm (Figure 6, inset). The reported variance in film thickness was calculated from the standard deviation of the absorbance measurements using this same method. The same procedure was used for Mg and Co.

Image not extracted

Figure 6. Figure 6. The calculated absorption edge of Fe is from the Center for X-ray Optics X-ray solid filter transmission database and was used to determine the thickness of the thermally evaporated Fe film studied here. The insert shows the measured absorption (green line) with shaded area showing the standard deviation and the calculated absorption (pink line).

The Fano shape at the Pt N7N_7-edge is not well captured by the calculated spectrum (Figure 7A). For this reason, a well-calibrated, multilayer Co/Pt thin film was used to calculate the thickness of the Pt film measured in this study (Figure 7B). The Co/Pt film was grown by atomic layer deposition, where the thickness of each layer was monitored during the growth by reflectometery. There was a total of 21 nm of Pt in the film. By comparing the N7N_7 –edge absorbance for the 21 nm Pt standard to the Pt film studied here, it was determined that the Pt film studied here was 35 nm thick. The variance in film thickness reported in the main text comes from the standard deviation of the absorbance measurement of the Pt film.

Image not extracted

Figure 7. The Fano lineshape of the Pt N7N_7-edge is not well calculated through the Center for X-ray Optics and X-ray solid filter transmission database (A),[39] therefore the absorption from a previously measured Co/Pt film was used to determine the thickness of the Pt film studied here (B). The shaded area denotes the standard deviation in the absorption measurement.

Transient Absorption Apparatus and Methods

Transient absorption studies described here are accomplished through the apparatus outlined in Figure 8. A Coherent Legend Elite laser amplifier, which has a central wavelength of 795 nm, a 25 fs (FWHM) temporal duration, a 5 mJ energy, and a 1 kHz repetition rate, is the light source. The laser output is then broadened (Figure 9A) to 420 nm – 920 nm (Figure 9B) through self-phase modulation in a stretched, gas-filled, silica hollow core fiber that is 2 m long and has a 704 μm inner diameter. The hollow core fiber is differentially pumped with negligible pressure at the laser entrance and 0.003 psi of Ar at the laser exit of the fiber.

Image not extracted

Figure 8. The schematic of the experimental apparatus displays the major components that convert the 795 nm, 25 fs (FWHM) laser output to an attosecond pulse train of extreme ultraviolet light for the absorption probe and a few femtosecond broadband optical pump. It also shows how the optical pump and probe are recombined to perform transient experiments. Here, FS and ADP refer to fused silica and ammonium dihydrogen phosphate, respectively.

Image not extracted

Figure 9. The Coherent Legend Elite output is broadened from ~ 100 nm bandwidth (A) to ~ 400 nm bandwidth (B) through self-phase modulation in a gas-filled stretched hollow core fiber.

The process of self-phase modulation[40] induces a temporal delay between the spectral frequencies of the broadband optical light. Here, double angle chirp mirrors (Ultrafast Innovations PC1332, PC70) are used to correct first order dispersion and 2 mm of ammonium dihydrogen phosphate is used to compensate for higher order dispersion.[41] The beam is then split into the two arms, with 20% of the light reflected to be used as the optical pump, and the remaining 80% to be used in the generation of extreme ultraviolet light absorption probe. Both arms have anti-reflection coated, fused silica wedge pairs, with one wedge on a translation stage, to independently compensate for second order dispersion. After re-compression, the beam typically has a 4 fs pulse duration, as measured through dispersion scanning.

The XUV light is generated in the probe arm through the HHG process. This process repeats every half cycle of the 4 fs long, broadband laser field, which inherently imparts an attosecond time duration to each XUV burst. After generation, the residual driving laser field is blocked with a 100 nm thick aluminum foil filter and the diverging XUV light is focused at the sample position with a gold coated toroidal mirror. The transmission of the XUV light through the sample is then measured with an XUV spectrometer, which is constructed from a shallow angle XUV grating and an x-ray sensitized CCD camera. In the pump arm, the optical intensity is controlled with an iris. A double mirror delay stage enables the time delay between the optical pump and XUV absorption probe to be controllably varied. The optical pump is then focused to a spot size of 0.15 mm20.15\ \mathrm{mm}^2 at the sample. After the sample, the residual pump light is blocked with a 150 nm thick aluminum foil filter.

Transient absorption spectra are obtained by calculating the change in absorption for each time from the measured XUV transmission in the presence (Ipump onI_{\mathrm{pump\ on}}) and absence of the optical pump (Ipump offI_{\mathrm{pump\ off}}), dA=−log⁡Ipump onIpump offdA=-\log\frac{I_{\mathrm{pump\ on}}}{I_{\mathrm{pump\ off}}}. This measurement process is then repeated for each specified time delay position in a continuous loop, which enables equivalent time delay points to be averaged. To account for temporal drift, the position on the delay stage corresponding to temporal overlap of the pump and probe is determined in gaseous Ne before each series of time delay measurements is conducted on the metals film.

The temporal duration of the autoionizing one-electron states of Ne is used to account for drifts in temporal overlap throughout the duration of an experiment. XUV absorption of the one electron, 1s22s12p6 np1s^{2}2s^{1}2p^{6}\,np where n=3,4,5…n = 3, 4, 5\ldots, Ne states can be seen in Figure 10A. Coupling between the discrete, one electron states and the first ionization continuum, 1s22s12p61s^{2}2s^{1}2p^{6}, results in a Fano profile observed in the XUV absorption (Figure 10B).[42] At temporal overlap the broadband optical pump excites the one-electron states into the second ionization continuum (1s2s22p61s^{2}s^{2}2p^{6}), destroying the Fano profile (Figure 10C). As the delay between the optical pump and the XUV absorption probe is increased the Fano line shape is suppressed and then recovers (Figure 11A). Fitting the line shape response with equation (7) provides the position of temporal overlap (Figure 11B, dashed orange line) and the time duration of the instrument response, which is equal to the full width half maximum of the fit Gaussian component (Figure 11B). Fitting the example shown in Figure 11B shows that temporal overlap is located at 12.1 μ\mum on the delay stage and the instrument response function is 5.2 fs.

Image not extracted

Figure 10. Single electron autoionizing states in Ne couple to the second ionization state to characterize the temporal response function of the instrument. (A) The XUV probe promotes electrons from the ground state into the single electron states. Then the broadband (BB) pump promotes electrons from these single electron states into the ionization continuum. A coupling between the discrete states and the first ionization continuum leads to a Fano profile in the observation of the single electron states with XUV absorption (B). This profile is destroyed at temporal overlap, where the broadband pump excites the population of these single electrons states into the continuum (C).

Transient temporal scan showing the suppression of the Fano profile and its temporal dependence

Figure 11. The transient temporal scan for the 1s2^{2}2s1^{1}2p6^{6} 3^{3}p state shows the suppression of the Fano profile around temporal overlap (A). By integrating this spectrum between 45.7 eV – 45.8 eV, blue points, the temporal dependence of this lineshape change can be easily seen (B). Fitting this profile with an exponentially modified error function, solid blue line, shows temporal overlap occurs at 12.1 μ\mum (orange dashed line) on the delay stage and the temporal resolution is 5.2 fs.

Once the delay stage is calibrated, the transmission of the XUV light through the metal sample is collected on the CCD camera at a specified position on the delay stage, first with the optical pump blocked and then with the optical pump on. The XUV transmission through the metal sample with the optical pump blocked and then illuminating the sample is repeated for each specified position of the delay stage. The process of calibrating the delay stage by measuring the lineshape suppression in Ne followed by measuring the XUV transmission through the sample is repeated in order to increase the signal-to-noise ratio of the change of absorption in the sample by averaging the change in absorption at equivalent time delay points.

The change in absorbance (dAdA), which is shown in eq. (4), is calculated by taking the negative log of the XUV transmission with optical pump illumination (Ipump,onI_{\mathrm{pump,on}}) divided by the XUV transmission without the optical pump (Ipump,offI_{\mathrm{pump,off}})

DA=−log⁡Ipump,onIpump,off(4)DA = -\log\frac{I_{\mathrm{pump,on}}}{I_{\mathrm{pump,off}}} \tag*{(4)}

After the change in absorbance is calculated and averaged for each equivalent delay position, after calibration, (Figure 12, first column) edge referencing is used to account for intensity fluctuations in the XUV source between the recorded XUV transmission with and without optical pump illumination (Figure 12, second column).[43] The energy ranges used for the reference portion where there is no sample absorption features in the transmitted XUV light are 45 eV – 56 eV for Co, 45 eV – 50 eV for Fe, 58 eV – 68 eV for Pt, and 35 eV – 49 eV for Mg. Lastly, the transient absorption signal appearing before temporal overlap, which occurs from laser heating on the millisecond timescale, is subtracted. In the experiments presented here and in similar work on titanium metal, the temporal dependence of transient absorption features after subtraction was the same as those where the heat signal was not subtracted.[10].

Collected transient absorption data for Co, Fe, Pt, and Mg

Figure 12. The collected transient absorption data for Co (top), Fe (upper middle), Pt (lower middle), and Mg (bottom) with the time axis for each average corrected using the Ne temporal overlap scans are shown in the first column). Next, edge referencing was used to minimize shot-to-shot harmonic fluctuations (middle column). Finally, the latent heat signal, which gives a positive signal before temporal overlap, was subtracted (last column). mdOD is milli delta OD.

Carrier Density Calculation

Assuming that each photon absorbed by the metal films only excites one electron – hole pair out of the thermal distribution of electrons, the number of excited carriers (N) generated in the metal film is equal to the number of photons in the pump, scaled by how many photons are absorbed by the metal film.[17, 18]

N=FEd(1−R)(1−e−ad)(1+Re−ad)(5)N = \frac{F}{E d} (1-R) \left(1-e^{-a d}\right) \left(1+R e^{-a d}\right) \tag*{(5)}

Here, F is the pump fluence, E is the energy of the photons, d is the thickness of the film, R is the reflectivity, and a is the absorption coefficient. For each film the extinction coefficient (k) was used to calculate the absorption coefficient (a) through the following relationship.

a=4πkλ(6)a = \frac{4 \pi k}{\lambda} \tag*{(6)}
Thickness (nm)Extinction Coeff.Fraction of Reflectivity44Fluence (mJ cm−2{}^{-2})Carrier Dens. (cm−3{}^{-3})
Co9.47.620.71121.01.70×10221.70 \times10^{22}
Fe183.590.60529.21.88×10221.88 \times10^{22}
Pt357.610.95522.61.12×10211.12 \times10^{21}
Mg1327.240.95219.82.74×10202.74 \times10^{20}

Table 2. A list of the parameters used and the calculated carrier densities.

While the charge carrier density excited varies for the different metal films this does not significantly affect the electron thermalization times measured in this study. In Appendix § 15 the carrier densities reported here are used to calculate electron temperatures of 1300 K, 2300 K, 3700 K, and 3400 K for Mg, Pt, Fe, and Co, respectively (Appendix § 15). Equation (1) shows that when kBTe≪Ek_{\mathrm{B}} T_{\mathrm{e}} \ll E, then 1τ≈AE2\frac{1}{\tau} \approx A E^{2}. Although electron thermalization generally decreases with increasing electron temperature (carrier density), the variance in electron temperature in the present study does not significantly affect the measured electron thermalization times.

Measured Transient Absorption Spectra

The transient absorption spectra for Mg, Pt, Fe, and Co are shown, respectively, in panels A - D of Figure 13. Edge referencing is used to account for intensity fluctuations in the XUV source between the recorded XUV transmission with and without optical pump illumination (Appendix § 5 Fig. 12).[43] The energy ranges used as the reference portion are 45 eV – 56 eV for Co, 45 eV – 50 eV for Fe, 58 eV – 68 eV for Pt, and 35 eV – 49 eV for Mg. Lastly, the transient absorption signal appearing before temporal overlap, which occurs from laser heating on the millisecond timescale, is subtracted. In the experiments presented here and in similar work on titanium metal, the temporal dependence of transient absorption features after subtraction was the same as those where the heat signal was not subtracted.[10] The spectra shown in Figure 13 were integrated over the ranges of 60 – 70 fs, 35 – 45 fs, 15 – 25 fs, and 2 – 12 fs for Mg, Pt, Fe, and Co, respectively, to obtain the transient absorption line shapes shown in Figure 2.

Image not extracted

Figure 13. Panels A – D show the transient absorption data collected for Mg (A), Pt (B), Fe (C), and Co (D). The dashed lines correspond to the energy of the static absorption edges. In the color map, a positive change in absorption is displayed in red and a negative change in absorption is displayed as blue.

Fitting transient absorption increase over time

The positive feature shown in Figure 2 A - D was integrated with respect to energy in order to assess how this feature changes over time. The integration windows used were 49.2 eV – 49.7 eV, 70.5 eV – 71.3 eV, 51.8 eV - 52.6 eV, and 56.5 eV – 65.2 eV for Mg, Pt, Fe, and Co, respectively. These data were then fit with an exponentially modified Gaussian function of the form:

f(x,μ,σ,λ)=∫−∞aλ2exp⁡[λ2(2μ+λσ2−2x)]erfc⁡(μ+λσ2−x2σ) dx(7)f(x,\mu,\sigma,\lambda)=\int_{-\infty}^{a}\frac{\lambda}{2}\exp\left[\frac{\lambda}{2}\left(2\mu+\lambda\sigma^{2}-2x\right)\right]\operatorname{erfc}\left(\frac{\mu+\lambda\sigma^{2}-x}{\sqrt{2}\sigma}\right)\,dx \tag*{(7)}

Here, μ\mu is the mean of the Gaussian component, λ\lambda is the rate of the rise, and σ\sigma is the variance of the Gaussian component. In the context of the transient XUV absorption experiments performed here, μ\mu corresponds to the time where both the optical pump and XUV absorption probe arrive simultaneously, σ\sigma is the temporal instrument response function, which is measured in Ne as described below, λ\lambda is the rate of the rise of the signal due to the electron thermalization, and the integration bound (aa) represents the time points. The timescales reported in the main text are the inverse of the rate of rise.

Time Dependent Density Functional Theory and Density Functional Theory

Here, we describe the numerical methods for analyzing static electronic structure and electron dynamics from first-principles. In order to compute the electronic structures of metals, we employed an open-source package, Quantum ESPRESSO, which is based on density functional theory.[29] For practical calculations, we used the projector augmented-wave method to treat valence electrons, and we utilized the PBE functional as the exchange-correlation functional.[45, 46] We set the cutoff energy to 45 Ry for all metals investigated in this work. The partial density of states (projected density of states) was computed by projecting the computed Kohn-Sham orbitals onto atomic orbitals.

In order to describe electron dynamics in metals, we employed time-dependent density functional theory (TDDFT).[47] For laser-excited metals following pump irradiation, the absorption spectra were computed using the finite-electron temperature method, which models hot-electron states.[48] The alteration in absorption spectra due to an increase in electron temperature was further analyzed by decomposing it into the state-filling effect and the local field effect, employing the Fermi-Dirac distribution.[12]

Here, we briefly describe the theoretical methods for analyzing the optical properties of laser-excited solids. The details of the methods were discussed elsewhere.[49] The dynamics of electrons are described by solving the following time-dependent Kohn-Sham equation for each electronic orbital:

iℏ∂∂tubk(r,t)=[12me(p+ℏk+eA(t))2+v^ion+vHXC(r,t)]ubk(r,t)(8)i\hbar\frac{\partial}{\partial t}u_{b\mathbf{k}}(\mathbf{r},t)=\left[\frac{1}{2m_e}\left(\mathbf{p}+\hbar\mathbf{k}+e\mathbf{A}(t)\right)^2+\hat{v}_{\mathrm{ion}}+v_{\mathrm{HXC}}(\mathbf{r},t)\right]u_{b\mathbf{k}}(\mathbf{r},t) \tag*{(8)}

where bb is the band index, and k\mathbf{k} is the Bloch wavevector, A(t)\mathbf{A}(t) is a spatially-uniform vector potential, v^ion\hat{v}_{\mathrm{ion}} is the ionic potential, and vHXC(r,t)v_{\mathrm{HXC}}(\mathbf{r},t) is the Hartree-exchange-correlation potential. In this work, we employ the local density approximation together with the adiabatic approximation for describing the electron dynamics in metals, and the Hartree-exchange-correlation potential becomes thus a functional of the instantaneous electron density as[50]

vHXC(r,t)=vHXC[ρ(r,t)](r,t)(9)v_{\mathrm{HXC}}(\mathbf{r},t)=v_{\mathrm{HXC}}[\rho(\mathbf{r},t)](\mathbf{r},t) \tag*{(9)}

Here, the electron density is defined by

ρ(r,t)=∑b∫BZdk fbkTe∣ubk(r,t)∣2(10)\rho(\mathbf{r},t)=\sum_b\int_{\mathrm{BZ}}d\mathbf{k}\,f_{b\mathbf{k}}^{T_e}\left|u_{b\mathbf{k}}(\mathbf{r},t)\right|^2 \tag*{(10)}

with the occupation factor fbkTef_{b\mathbf{k}}^{T_e} given by the Fermi-Dirac distribution. Hence, the occupation factor depends on the electron temperature TeT_e.

By employing the time-dependent orbital, ubk(r,t)u_{b\mathbf{k}}(\mathbf{r},t), the time-dependent current density can be computed as

J(t)=−emeΩ∑b∫BZdk fbkTe∫Ωdr ubk∗(r,t)vk(t)ubk(r,t)(11)\mathbf{J}(t)=-\frac{e}{m_e\Omega}\sum_b\int_{\mathrm{BZ}}d\mathbf{k}\,f_{b\mathbf{k}}^{T_e}\int_{\Omega}d\mathbf{r}\,u_{b\mathbf{k}}^*(\mathbf{r},t)\mathbf{v}_{\mathbf{k}}(t)u_{b\mathbf{k}}(\mathbf{r},t) \tag*{(11)}

where Ω\Omega is the volume of the unit cell, and vk(t)\mathbf{v}_{\mathbf{k}}(t) is the velocity operator.

In this work, to compute the absorption spectra of metals, we compute the electron dynamics by solving eq. 8 with a weak impulsive distortion, A(t)=−E0epΘ(t)\mathbf{A}(t)=-E_0\mathbf{e}_p\Theta(t), with the field amplitude of E0E_0 and the polarization direction along ep\mathbf{e}_p. We further compute the electric current with eq. 11 for evaluating the optical conductivity σ(ω)\sigma(\omega) as

σ(ω)=∫0∞dteiωt−γtep⋅J(t)E0(12)\sigma(\omega)=\frac{\int_0^\infty dt e^{i\omega t-\gamma t}\mathbf{e}_p\cdot\mathbf{J}(t)}{E_0} \tag*{(12)}

where γ\gamma is a damping factor introduced to reduce numerical noise due to the finite simulation time. In this work, we set γ\gamma to 0.5 eV/ℏ\hbar. By using the optical conductivity, one can further evaluate the dielectric function and the absorption coefficient as

ϵ(ω)=1+4πiωσ(ω)(13)\epsilon(\omega) = 1 + \frac{4\pi i}{\omega}\sigma(\omega) \tag*{(13)}
μ(ω)=2ωcℑ[ϵ(ω)](14)\mu(\omega) = \frac{2\omega}{c}\Im\left[\sqrt{\epsilon(\omega)}\right] \tag*{(14)}

For practical calculations for the optical properties of metals, we employed an open-source package, Octopus.[50] For bulk magnesium, we employ a hexagonal crystal structure, and we set the lattice parameter aa to 3.21 Å and the lattice constant ratio c/ac/a to 1.624.[51] The hexagonal lattice is discretized into 24×24×3924 \times24 \times39 real-space grid points, and the first Brillouin zone is discretized into 16316^{3} kk-points. The magnesium atom is described using a norm-conserving pseudopotential method, treating 2s2s, 2p2p, and 3s3s electrons as valence.[52] Likewise, for bulk cobalt, we use the hexagonal crystal structure, and we set aa to 2.51 and c/ac/a to 1.611.[51] The lattice is discretized into 24×24×3824 \times24 \times38 grid points, and the first Brillouin zone is discretized into 24324^{3} kk-points. The cobalt atom is described using a norm-conserving pseudopotential method, treating 3s3s, 3p3p, 3d3d and 4s4s electrons as valence.[52]

We compute the absorption spectra of Mg and Co in the equilibrium phase by setting the electron temperature TeT_e to 300 K, while we compute those of laser-excited systems by setting TeT_e to 0.25 eV/kBTk_{\mathrm{B}}T and 0.026 eV/kBTk_{\mathrm{B}}T for Mg and Co, respectively. Furthermore, the change in the absorption spectra by the increase in the electron temperature were decomposed into three different components: One is the state-filling effect, which originates from the temperature dependence of the Fermi-Dirac distribution via the occupation factor in eq. (11). Another one is the change of the local-field effect, which originates from the temperature dependence of the occupation factor in the time-dependent part of the potential in eq. (9). The other is the band structure modification, which originates from the temperature dependence of the static part of the potential in eq. (9). Note that, the band structure modification is negligible for both Mg and Co, in this work.

Calculating Final Electron Temperature

The electron temperature reached after initial thermalization (TthT_{\mathrm{th}}) can be estimated by the following relationship.

hcλN=∫T=300 KTthCe(T′) dT′(15)\frac{hc}{\lambda}N = \int_{T=300\ \mathrm{K}}^{T_{\mathrm{th}}} C_e(T')\,\mathrm{d}T' \tag*{(15)}

Here, NN is the number of photons absorbed by the sample (Appendix E), λ\lambda is the wavelength of the light, and Ce(T)C_e(T) is the electron heat capacity. The left hand side is the energy absorbed by the sample and the right hand side is the integrated electron heat capacity.

For each metal, the electron heat capacity was estimated by using the linear relationship between heat capacity and temperature with the slope taken as the Somerfield electron heat capacity coefficient.[53, 54, 55] The final electron temperature was found to be 1300 K, 2300 K, 3700 K, and 3400 K for Mg, Pt, Fe, and Co, respectively (Appendix H). We assess the impact that the difference in electron temperature between the metals has on the initial electron thermalization time through eq. 1. For all the metals, kBTe≪Ek_{\mathrm{B}}T_e \ll E, which reduces eq. 1 to 1τ≈AE2\frac{1}{\tau} \approx A E^2. The result shows that the temperature differences between the metals does not significantly affect the measured thermalization time.

ρ\rho and MeffM_{\mathrm{eff}} Used in the Calculated Electron Thermalization Time

Table 1 presents the DOS (ρ\rho) and screened electron interaction (∣Meff∣2|M_{\mathrm{eff}}|^2) values used in equations 1 and 2 to calculate the electron thermalization time. Within the random-kk approximation theory for calculating thermalization times there have been several ways of including the number of states participating in scattering. The most accurate of these accounts for the volume of states with excited electrons by integrating up to the excitation energy of the unfilled states and accounts for the scattering partners by integrating the filled DOS up to the Fermi level.[6, 31, 32] While this is more accurate, we choose the more accessible simplification outlined by Zhukov and Chulkov where a constant DOS is assumed, which reduces the scattering integrals to cubing a singular DOS value at the excitation energy.[31] The DOS used here was calculated with an open-source package, Quantum ESPRESSO, with the methodology outlined in Appendix G. For Mg and Co there are two atoms per unit cell, therefore the DOS values at 1.65 eV are divided by 2, in order to get the per atom values.[32]

Although it is possible to calculate MeffM_{\mathrm{eff}} within the free electron gas (FEG) model, for most metals the FEG approximation is innacurate.[31, 32] However, it has been shown that fitting photoemission data with the random-kk approximation or directly calculating it with many-body scattering theory (GW, GW+T, STA-DOS) provide a more accurate MeffM_{\mathrm{eff}} value.[7, 31, 33, 35] For this reason, we use the MeffM_{\mathrm{eff}} values for Fe and Pt provided by Zhukov et. al and the MeffM_{\mathrm{eff}} value reported for Co by Knorren et. al.[30, 33] As the MeffM_{\mathrm{eff}} value for Mg has not been reported, we used the FEG approximation shown in equation 3 to calculate MeffM_{\mathrm{eff}} for Mg. While this is not accurate for most metals, because the DOS of Mg is similar to the FEG DOS, Mg is a good candidate for this simplification. The plasma frequency and Fermi level were calculated using the free electron model.[56] In this model ωp=47.1 eV(rsa0)−32\omega_p = 47.1\ \mathrm{eV}\left(\frac{r_s}{a_0}\right)^{-\frac{3}{2}}, where rsa0\frac{r_s}{a_0} is the reduced electron density parameter. For Mg, rsa0\frac{r_s}{a_0} is 2.66 which gives ωp=16.5\omega_p = 16.5 eV. For the Fermi level, EF=ℏ22me(3 π2 n)23E_F = \frac{\hbar^2}{2m_e}(3\ \pi^2\ n)^{\frac{2}{3}} where mem_e is the mass of an electron and nn is electronic density. For Mg, n=8.61×1022 cm−3n = 8.61 \times10^{22}\ \mathrm{cm}^{-3}, yielding EF=7.08E_F = 7.08 eV.[56]

References

  1. [1]R. B. Wilson, J. Gorchon, Y. Yang, C.-H. Lambert, S. Salahuddin, J. Bokor, Ultrafast magnetic switching of GdFeCo with electronic heat currents, Phys. Rev. B (2017) 95 18040.DOI
  2. [2]M. Battiato, K. Carva, P. M. Oppeneer, Superdiffusive spin transport as a mechanism of ultrafast demagnetization, Phys. Rev. Lett (2010) 2 105.pubmed.ncbi.nlm.nih.gov/20867735
  3. [3]G. M. Choi, R. R. Wilson, D. G. Cahill, Indirect heating of Pt by short-pulse laser irradiation of Au in a nanoscale Pt/Au bilayer, Phys. Rev. B (2014) 89 064307.DOI
  4. [4]M. Battiato, K. Carva, P. M. Oppeneer, Theory of laser-induced ultrafast superdiffusive spin transport in layered heterostructures, phys. Rev. B (2012) 86, 024404.DOI
  5. [5]R. B. Wilson, S. Coh, Parametric dependence of hot electron relaxation timescales on electron-electron and electron-phonon interaction strengths, Commun. Phys. (2020) 3, 179.DOI
  6. [6]E. Zarate, P. Apell, P. M. Echenique, Calculation of low-energy-electron lifetimes, Phys. Rev. B (1999) 60 , 4.DOI
  7. [7]M. Baur, A. Marienfeld, M. Aeschlimann, Hot electron lifetimes in metals probed by time-resolved two-photon photoemission, Prog. Surf. Sci. (2015), 90 319 – 376.
  8. [8]M. Aeschlimann, M. Baur, S. Pawlik, Competing nonradiative channels for hot electron induced surface photochemistry, Chem. Phys. (1996), 127 – 141.DOI
  9. [9]H.-T. Chang, A. Guggenmos, S. K. Cushing, Y Cui, N. U. Din, S. R. Acharya, I. J. Porter, U. Kleineberg, V. Turkowski, T. S. Rahman, D. M. Neumark, S. R. Leone, Electron thermalization and relaxation in laser-heated nickel by few-femtosecond core-level transient absorption spectroscopy, Phys. Rev. B, 103 (2021) 064305.DOI
  10. [10]M. Volkov, S.A. Sato, F. Schlaepfer, L. Kasmi, N. Hartmann, M. Lucchini, L. Gallmann, A. Rubio, U. Keller, Attosecond screening dynamics mediated by electron localization in transition metals, Nat. Phys. 15 (2019) 1145–1149.
  11. [11]A. Niedermayr, M. Volkov, S. A. Sato, N. Hartmann, Z. Schumacher, S. Neb, A. Rubio, L. Gallmann, U. Keller, Few-Femtosecond Dynamics of Free-Free Opacity in Optically Heated Metals, Phys. Rev. X, (2022), 12 021045.
  12. [12]S. A. Sato, First-principles calculations for attosecond electron dynamics in solids, Comput. Mat. Sci. (2021) 194, 110274.arxiv.org/abs/2011.01677
  13. [13]P. H. Citrin, G. K. Wertheim, M. Schluter, One-electron and many body effects in x-ray absorption and emission edges of Li, Na, Mg, and Al metals, Phys. Rev. B, (1979) 20, 8.DOI
  14. [14]M. Fanfoni, S. Modesti, N. Motta, M. de Crescenzi, R. Rosei, Comparison between extended x-ray absorption and extended electron energy loss fine-structure results above the M₂,₃ edge of cobalt, Phys. Rev. B (1985), 32, 12.pubmed.ncbi.nlm.nih.gov/9936954
  15. [15]T. Shishidou, S. Imada, T. Muro, F. Oda, A. Kimura, S. Suga, Strong Fano effect in the magnetic circular dichroism of the Pt N₆,7 core absorption of ferromagnetic CoPt₃, Phys. Rev. B, (1997) 55, 6.
  16. [16]K. Yao, F. Willems, C. v. K. Schmissing, I. Radu, C. Struber, D. Schick, D. Engel, A. Tsukamoto, J. K. Dewhurst, S. Sharma, S. Eisebitt, Distinct spectral response in M-edge magnetic circular dichroism, Phys. Rev. B, (2020) 102, 100405.DOI
  17. [17]M. Zurch, H.-T. Chang, L. J. Borja, P. M. Kraus, S. K. Cushing, A. Gandman, C. J. Kaplan, M. H. Oh, J. S. Prell, D. Prendergast, C. D. Pemmaraju, D. M. Neumark, S. R. Leone, Direct and simultaneous observation of ultrafast electron and hole dynamics in germanium, Nat. Commun. 8 (2017), 1.DOI
  18. [18]S. K. Cushing, A. Lee, I. J. Porter, L. M. Carneiro, H.-T. Chang, M. Zurch, S. R. Leone, Differentiating Photoexcited Carrier and Phonon Dynamics in the Δ, L, and Γ Valleys of Si(100) with Transient Extreme Ultraviolet Spectroscopy, J. Phys. Chem. C 123 (2019), 3343.DOI
  19. [19]D. E. Aspnes, Local-field effects and effective-medium theory: A microscopic perspective, Am. J. Phys. 50 (1982), 704.DOI
  20. [20]F. Wooten, Optical properties of solids, (New Yok, London: Academic Press, 1972).DOI
  21. [21]J. M. Ziman, Principles of the theory of solids (Cambridge: University Press, 1964).
  22. [22]W. Hanke, The role of electron-hole interaction in the optical spectra of semiconductors and insulators, Advances in Solid State Physics (Springer, Berlin, Heidelberg, 1979).
  23. [23]E. E. Krazovskii, W. S. Schattke, Local field effects in optical excitations of semicore electrons, Phys. Rev. B, 60 (1999), R16251.
  24. [24]S. L. Alder, Quantum Theory of the Dielectric Constant in Real Solids, Phys. Rev., 126 (1961), 413.
  25. [25]N. Weiser, Dielectric Constant with Local Field Effects Included, Phys. Rev. 129 (1962), 62.
  26. [26]B. Y. Meuller, B. Rethfeld, Relaxation dynamics in laser-excited metals under nonequilibrium conditions, Phys. Rev. B (2013) 87, 035139.
  27. [27]R. H. M. Groeneveld, R. Sprink, Femtosecond spectroscopy of electron-electron and electron-phonon energy relaxation in Ag and Au, Phys. Rev. B (1995) 51, 17.pubmed.ncbi.nlm.nih.gov/9977873
  28. [28]M. Obergfell, J. Demsar, Tracking the Time Evolution of the Electron Distribution Function in Copper by Femtosecond Broadband Optical Spectroscopy, Phys. Rev. Lett. (2020) 124, 037401.DOI
  29. [29]P. Giannozzi, O. Andreussi, T. Brumme, O. Bunau, M. B. Nardelli, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, M. Cococcioni, N. Colonna, I. Carnimeo, A. Dal Corso, S. de Gironcoli, P. Delugas, R. A. DiStasio Jr., A. Ferretti, A. Floris, G. Fratesi, G. Fugallo, R. Gebauer, U. Gerstmann, F. Giustino, R. Gorni, J. Jia, M. Kawamura, H.-Y. Ko, A. Kokali, E. Kucukbelb, M. Lazzeri, M. Marsili, N. Marzari, F. Mauri, N. L. Nguyen, H.V. Nguyen, A. Otero-de-la-Roza, L. Paulatto, S. Ponce, D. Rocca, R. Sabatini, B. Santra, M. Schlipf, A. P. Seit­sonen, A. Smogunov, I. Timrov, T. Thonhauser, P. Umari, N. Vast, X. Wu, S. Baroni, Advanced capabilities for materials modelling with Quantum ESPRESSO, J. Phys. Condens. Matter (2017) 29, 465901.
  30. [30]R. Knorren, K. H. Bennemann, R. Burgermeister, M. Aeschlimann, Dynamics of excited electrons in copper and ferromagnetic transition metals: Theory and experiment, Phys. Rev. B (2000) 61, 9427 - 9440.DOI
  31. [31]V. P. Zhukov, E. V. Chulkov, Lifetimes of low-energy electron excitations in metals, J. Phys.: Condens. Matter (2002) 14, 1937 - 1947.DOI
  32. [32]V. P. Zhukov, F. Aryasetiawan, E. V. Chulkov, P. M. Echenique, Lifetimes of quasiparticle excitations in 4d transition metals: Scattering theory and LMTO-RPA-GW approaches, Phys. Rev. B (2002) 65, 115116.DOI
  33. [33]V. P. Zhukov, E. V. Chulkov, P. M. Echenique, Lifetimes and inelastic mean free path of low-energy excited electrons in Fe, Ni, Pt, and Au: Ab initio GW + T calculations, Phys. Rev. B, (2006) 73, 12105.DOI
  34. [34]V. P. Zhukov, E. V. Chulkov, The femtosecond dynamics of electrons in metals, Phys.-Usp., (2009) 52, 105.DOI
  35. [35]C. Chen, Z. Tao, A. Carr, P. Matyba, T. Szilvasi, S. Emmerich, M. Piecuch, M. Keller, D. Zusin, S. Eich, M. Rollinger, W. You, S. Mathias, U. Thumm, M. Mavrikakis, M. Aeschlimann, P. M. Oppeneer, H. Kapteyn, M. Murnane, Distinguishing attosecond electron- electron scattering and screening in transition metals, PNAS (2017) 114, 1-8.
  36. [36]E. Campillo, V. M. Silkin, J. M. Pitarke, E. V. Chulkov, A. Rubio, P. M. Echenique, First-principles calculations of hot-electron lifetimes in metals, Phys. Rev. B (2000) 61, 13484.DOI
  37. [37]D. Pines and P. Nozieres, The Theory of Quantum Liquids (Benjamin, New York, 1966).
  38. [38]W. F. Krolikowsky, W. E. Spicer, Photoemission studies of noble metals. 1. Copper, Phys. Rev. (1969) 185, 882.
  39. [39]B. L. Henke, E.M. Gullikson, and J.C. Davis. X-ray interactions: photoabsorption, scattering, transmission, and reflection at E=50-30000 eV, Z=1-92, Atomic Data and Nuclear Data Tables Vol. 54 (no.2), 181-342 (July 1993).
  40. [40]S. C. Pinault, M. J. Potasek, Frequency broadening by self-phase modulation in optical fibers, J. Opt. Soc. Am. B (1985) 2, 1318.DOI
  41. [41]H. Timmers, Y. Kobayashi, K. F. Chang, M. Reduzzi, D. M. Neumark, S. R. Leone, Generating high-contrast, near single-cycle waveforms with third-order dispersion compensation, Opt. Lett. (2017) 42, 811.DOI
  42. [42]K. Colding, R. P. Madden, D. L. Ederer, Resonances in the Photo-Ionization Continuum of Ne I (20 - 150 eV), Phys. Rev (1967) 155, 26.
  43. [43]R. Geneaux, H.-T. Chang, A. Schwartzberg, H. J. B. Marroux, Source noise suppression in attosecond transient spectroscopy by edge-pixel referencing, Opt. Express (2021) 29, 951.DOI
  44. [45]J. R. Chelikowsky, N. Troullier, K. Wu, Y. Saad, Higher-order finite-difference pseudopotential method: an application to diatomic molecules, Phys. Rev. B (1994) 50 (11355).pubmed.ncbi.nlm.nih.gov/9975266
  45. [46]J. P. Perdew, K. Burke, M. Ernzerhof, Generalized gradient approximation made simple, Phys. Rev. Lett. (1996) 77, 3865.pubmed.ncbi.nlm.nih.gov/10062328
  46. [47]E. Runge, E. K. U. Gross, Density-functional theory for time-dependent systems, Phys. Rev. Lett. (1984) 52, 997.DOI
  47. [48]S. A. Sato, Y. Shinohara, T. Otobe, K. Yabana, Dielectric response of laser-excited silicon at finite electron temperature, Phys. Rev. B (2014) 90, 174303.DOI
  48. [49]J. P. Perdew, Y. Wang, Accurate and simple analytic representation of the electron-gas correlation energy, Phys. Rev. B, (1992) 45, 13244.pubmed.ncbi.nlm.nih.gov/10001404
  49. [50]N. Tancogne-Dejean, M. J. T. Oliveira, X. Andrade, H. Appel, C. H. Borca, G. Le Breton, F. Buchholz, A. Castro, S. Corni, A. A. Correa, U. De Giovannini, A. Delgado, F. G. Eich, J. Flick, G. Gil, A. Gomez, N. Helbig, h. Hubener, R. Jestadt, J. Jornet-Somoza, A. H. Larsen, I. V. Lebedeva, M. Luders, M. A. L. Marques, S. T. Ohlmann, S. Popolo, M. Rampp, C. A. Rozzi, D. A. Strubbe, S. A. Sato, C. Schafer, I. Theophilou, A. Welden, A. Rubio, Octopus, a computational framework for exploring light-driven phenomena and quantum dynamics in extended and finite systems, J. Phys. Chem. (2020) 152, 12.
  50. [51]K. Hermann, Crystallography and surface structure: an introduction for surface scientists and nanoscientists (John Wiley & Sons, 2017).
  51. [52]M. J. Oliveira, F. Nogueira, Generating relativistic pseudo-potentials with explicit incorporation of semi-core states using APE, the Atomic Pseudo-potentials Engine, Comput. Phys. Commun. (2008) 178, 7.DOI
  52. [53]T. I. Kakushadze, Electron heat capacity in transition metals and alloys. II, Annalen der Physik, (1961) 463, 7.DOI
  53. [54]C. Kittel, Introduction to Solid State Physics (Wiley, New York, Ed. 5th, 1988).
  54. [55]D. H. Dye, J. B. Ketterson, G. W. Crabtree, The Fermi Surface of Platinum, J. Low Temp. Phys. (1978) 30, 5.
  55. [56]N. W. Ashcroft and N. D. Mermin, Solid State Physics (Saunders, Philadelphia, 1976).

Paper details

vlnc:W_75xupqd9
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents