()Ultrafast magnetization dynamics of nickel has been studied for different degrees of electronic excitation, using pump-probe second-harmonic generation with 150 fs/800 nm laser pulses of various fluences. Information about the electronic and magnetic response to laser irradiation is obtained from sums and differences of the SHG intensity for opposite magnetization directions. The classical M(T)-curve can be reproduced for delay times larger than the electron thermalization time of about 280 fs, even when electrons and lattice have not reached thermal equilibrium. Further we show that the transient magnetization reaches its minimum ≈ 50 fs before electron thermalization is completed.PACS numbers: 42.65. Ky , 75.40.Gb , 78.47.+p Ultrafast spin dynamics in ferromagnets is of great interest from both theoretical and experimental points of view. In particular, the short-time dynamics of magnetism in transition metals, with many excited electrons not at equilibrium with the lattice, is a new area of physics. Such studies are important for developing a theory of transient magnetization behavior in the subpicosecond range. It seems that the only experimental data which can guide theoretical analysis are the ones reported by Beaurepaire et al.[1] on time-resolved demagnetization of Ni induced by femtosecond laser pulses of 620 nm at one specific fluence. The authors utilized the magneto-optical Kerr effect to detect hysteresis loops for different time delays between pump and probe pulses. By comparing the time-dependent remanence with the equilibrium temperature dependence of magnetization, M (T ), they derived the time evolution of the spin temperature within the framework of the phenomenological three-temperature model [2]. Clearly, it is of great importance to confirm whether or not M (T ) can be used to describe the transient magnetic response to electron excitations in itinerant ferromagnets and whether there is a time delay between electron thermalization and magnetization changes.In this Letter we present time-resolved data on the transient magnetization measured by pump-probe second-harmonic generation (SHG). The great advantage of this technique is that it allows to simultaneously follow electron-temperature relaxation and transient magnetization, without further need for additional calibration measurements. This is a consequence of the even and odd contributions to the nonlinear susceptibility [3]. The measurements were carried out for a large variety of pump fluences leading to different initial electron temperatures. After equilibration of the electron bath, we find the transient magnetization to be governed by the electron temperature T e via the classical M (T )-curve [4]. However, we observe a strong deviation of the data 1

Both theoretical and experimental results for the dynamics of photoexcited electrons at surfaces of Cu and the ferromagnetic transition metals Fe, Co, and Ni are presented. A model for the dynamics of excited electrons is developed, which is based on the Boltzmann equation and includes effects of photoexcitation, electron-electron scattering, secondary electrons (cascade and Auger electrons), and transport of excited carriers out of the detection region. From this we determine the time-resolved two-photon photoemission (TR-2PPE). Thus a direct comparison of calculated relaxation times with experimental results by means of TR-2PPE becomes possible. The comparison indicates that the magnitudes of the spin-averaged relaxation time τ and of the ratio τ ↑ /τ ↓ of majority and minority relaxation times for the different ferromagnetic transition metals result not only from density-ofstates effects, but also from different Coulomb matrix elements M . Taking MFe > MCu > MNi = MCo we get reasonable agreement with experiments.

No abstract

No abstract

Within a theoretical model based on the Boltzmann equation, we analyze in detail the structure of the unusual peak recently observed in the energy dependence of the relaxation time in Cu. In particular, we discuss the role of Auger electrons in the electron dynamics and its dependence on the d-hole lifetime, the optical transition matrix elements, and the laser-pulse duration. We find that the Auger contribution to the distribution is very sensitive to both the d-hole lifetime h and the laser-pulse duration l , and that it largely dominates the excited-electron distribution for realistic parameters. It is shown that it can be expressed as a monotonic function of l / h and that for a given h , it is significantly smaller for a short pulse duration than for a longer one. Our results indicate that the relaxation time at the peak depends linearly on the d-hole lifetime, but interestingly not on the amount of Auger electrons generated. We provide a simple expression for the relaxation time of excited electrons which shows that the shape of the peak can be understood by a phase-space argument and its amplitude is governed by the d-hole lifetime. We also find that the height of the peak depends on both the ratio of the optical transition matrix elements Rϭ͉M d→sp ͉ 2 /͉M sp→sp ͉ 2 and the laser-pulse duration. Assuming a reasonable value for the ratio, namely, Rϭ2, and a d-hole lifetime of h ϭ35 fs, we obtain for the calculated height of the peak ⌬ th ϭ14 fs, in fair agreement with ⌬ exp Ϸ17 fs measured for polycrystalline Cu.