Supplementary Material Detection of Spin-Resolved Electronic Structures from a Buried Ferromagnetic Layer utilizing Forward Mott Scattering In this supplementary section, we would like to describe the details of the efficiency enhancement by the use of the multi-channel detection, the numerical calculation of the effective Sherman function, multiple scattering problem in a standard Mott detector, and the Fe 2p core-level spectra for the thinner Au capped FeNi alloy film. Efficiency enhancement by the use of the multi-channel detector By the use of the multi-channel detector, the efficiency is enhanced by a factor 3,600 in comparison with the single channel detection. For the Fe 2p core-level HAXPES measurements, we have used an angle integrating transmission mode. The enhancement factor is given by 600 active channels in the x-axis multiplied by 6 channels (the energy step of the experiment divided by the minimum energy step) in the y-axis of Fig. 1, shown in the article. This enhancement sufficiently compensates the drawback of the low photoionization cross-section in HAXPES. The photoionization cross-section of the Fe 2p core-level at 10 keV is two orders of magnitude weaker than at 1 keV [1]. Numerical calculation To clarify the photoelectron intensity distribution created by the photoelectron scattering and to evaluate Seff, we have performed simplified numerical calculations, in which the single and forward scattering processes in the FeNi and Au layers were taken into account. To simplify the calculation, we assumed the isotropic scattering probability at a certain depth from the surface. The photoelectron intensity distribution is given by, I(q1,q 2 ) = IFeNi (q1,q 2 ) + IAu (q1,q 2 ) , IFeNi (q1,q 2 ) = å d1 ,d 2 ,q 1 ,q 2 Eq. (4) exp{ -(d1 - d2 ) / lFeNi cos q1 - (d2 - t Au ) / lFeNi cos(q1 + q 2 ) - t Au / lAu cos(q1 + q 2 )} n(d1 ) Eq. (5) 1 , IAu (q1,q 2 ) = å exp{ -(d1 - t Au ) / lFeNi cos q1 - (t Au - d2 ) / lAu cosq1 - d2 / lAu cos(q1 + q 2 )} n(d1 ) d1 ,d 2 ,q 1 ,q 2 , Eq. (6) where 1 is the emission angle of the generated photoelectrons with respect to the surface normal direction at a depth of d1 from the surface, 2 is the scattering angle of the photoelectrons at a depth of d2 (d2 d1), and n(d1) is the number of photoelectron trajectories generated at a depth of d1. The IMFPs of FeNi (FeNi) of 9.0 nm and Au (Au) of 6.9 nm for the 9.2-keV electrons were used in the calculation. An FeNi layer thickness (tFeNi) of 17.9 nm and a Au layer thickness (tAu) of 4.0 nm were used. The angular and depth steps in the calculation were set to 1°and 0.5 nm, respectively. I(1, 2), IFeNi(1, 2), and IAu(1, 2) correspond to the total intensity distribution, the intensity distribution of photoelectrons scattered in the FeNi layer, and the intensity distribution of photoelectrons scattered in the Au layer, respectively. take-off-angle (TOA) of photoelectrons is represented by 90-1-2. The We took the acceptance angle of the electron analyzer (6) into account: the central angle (1+2) IAu(1,2)/ I(1,2) = 0.75 was obtained from the calculation for was set to 65. 1+2 = 656. We thus see that most of the photoelectrons are scattered in the Au layer. We then calculated Seff due to scattering in the Au layer: å S(q ) ´ I 2 Seff = q2 åI Au Au (q1,q 2 ) (q1,q 2 ) , Eq. (7) q2 where S() is the calculated Sherman function for 9.2-keV electrons for Au point nuclei [2] as shown in Fig. 2(d) in the article. We also took into account the acceptance angle of the electron analyzer in the calculation of Seff. We obtained Seff = 0.07 from the calculation. Multiple scattering in standard Mott detector In the case of standard Mott detector using a Au target, the film thickness of the Au target is about 100 nm. The effect of the multiple scattering in a thick Au target is not negligible. Therefore the effective Sherman function decreases with increasing the Au 2 target film thickness. In addition the effective Sherman function in standard Mott detector is roughly less than half of the calculated Sherman function for gold atom using a single scattering formalism [3]. In contrast, the multiple scattering in our case is strongly reduced compared to standard Mott detector, since we used a 4-nm-thick Au film. Fe 2p core-level HAXPES spectra for the 2-nm-thick Au capped FeNi alloy film Figure S1 shows the Fe 2p core-level HAXPES spectra (I+ and I-) measured by switching the sample magnetizations of the 2-nm-thick Au capped FeNi alloy film. As seen in Fig. S1, the I+ spectrum was identical with the I- spectrum. Therefore we were unable to find any spin-dependent structures in the Fe 2p core-level region for the Au(2 nm)/FeNi alloy film. Fig. S1: Fe 2p core-level spectra (I+ and I-) for the Au(2 nm)/FeNi alloy film. [1] Scofield, J. H., Tech. Rep., Lawrence Livemore Laboratory Report No. UCRL-51326 (1973). [2] Sherman, N., Phys. Rev. 103, 1601 (1956). [3] See e.g. Qiao, S. & Kakizaki, A., Rev. Sci. Instrum. 68, 4017 (1997). 3