Cation diffusion and hybridization effects at the Mn-GaSe(0001) interface probed by soft X-ray electron spectroscopies
S. Dash, G. Drera, E. Magnano, F. Bondino, P. Galinetto, M.C. Mozzati, G. Salvinelli, V. Aguekian, L. Sangaletti
aa r X i v : . [ c ond - m a t . m t r l - s c i ] M a y Cation diffusion and hybridization effects at the Mn-GaSe(0001) interface probed bysoft X-ray electron spectroscopies
S. Dash , G. Drera , E. Magnano , F. Bondino , P. Galinetto ,M.C. Mozzati , G. Salvinelli , V. Aguekian , and L. Sangaletti Interdisciplinary Laboratories for Advanced Materials Physics (I-Lamp) and Dipartimento di Matematica e Fisica,Universit`a Cattolica, via dei Musei 41, 25121 Brescia (Italy) IOM-CNR, Laboratorio TASC, S.S. 14, Km 163,5 I-34149 Basovizza (Italy) CNISM and Dipartimento di Fisica, Universit`a di Pavia, Via Bassi 6 (Italy) and Solid State Physics Department, V. Fock, Institute of Physics,Saint-Petersburg State University, Petrodvoretz, 198904 S.-Petersburg, Russia (Dated: July 16, 2018)The electronic properties of the Mn:GaSe interface, produced by evaporating Mn at room tem-perature on a ǫ -GaSe(0001) single crystal surface, have been studied by soft X-ray spectroscopies.Substitutional effects of Mn replacing Ga cations and Mn-Se hybridization effects are found bothin core level and valence band photoemission spectra. The Mn cation valence state is probed byXAS measurements at the Mn L-edge, which indicate that Mn diffuses into the lattice as a Mn cation with negligible crystal field effects. The Mn spectral weight in the valence band is probedby resonant photoemission spectroscopy at the Mn L-edge, which also allowed an estimation of thecharge transfer and Mott-Hubbard energies on the basis of impurity-cluster configuration-interactionmodel of the photoemission process. The charge transfer energy is found to scale with the energygap of the system. Competing effects of Mn segregation on the surface have been identified, andthe transition from the Mn diffusion through the surface to the segregation of metallic layers on thesurface has been tracked by core-level photoemission. PACS numbers: Valid PACS appear here
I. INTRODUCTION
The III-VI semiconductors GaSe, InSe, GaTe, andGaS have received considerable interest in the last fewyears because they show remarkable nonlinear opticalproperties and they are regarded as promising materi-als for photo-electronic applications , even in the formof nanowires . In the particular case of GaSe and GaS,the interest on these systems has been recently reneweddue to the possibility to obtain ultrathin layer transistorsbased on atomic-thin sheets .The magnetic properties of these systems doped withtransition metal ions (e.g. Mn , or Fe -doped GaSe) arealso under investigation with the aim to find out newclasses of diluted magnetic semiconductors (DMS) of theform A III − x M x B V I , where A
III B V I is a III-VI semicon-ductor and M is a transition metal ion. For instance,Mn has been incorporated into
GaSe (0001) in samplesgrown from the melt, and intriguing magnetic propertieshave been found . A short range anti-ferromagnetic or-dering has been invoked to explain the rather complexmagnetic behavior, but a clear identification of the shortrange coupling mechanisms related to these experimentalevidences is still missing. Moreover, a clear understand-ing of the interplay between magnetism and electronicproperties has not yet been reported so far, mainly dueto the difficulty of growing high quality Mn-doped singlecrystals and control both the doping level and possiblephase segregations or the creation of defects and vacan-cies upon doping. Also the local structure around Mnatoms at the Mn:GaSe interface has not yet been probed, being the mechanism of Mn diffusion in the lattice poorlyinvestigated. This can be important in order to relate theobserved magnetic behavior to either direct or superex-change interactions through M n − M n or M n − Se − M n bonds, respectively.Furthermore, recent studies on the Mn:Ga Se system have drawn the attention on this interface, thatis strictly related to the one we are currently studying.The magnetic properties of the Mn:Ga Se system havealso been reported , suggesting weak antiferromagneticcorrelations in the bulk crystal.Like the II-VI DMS, substitutional magnetic ions inthe III-VI DMS are found in a (distorted) tetrahedralenvironment. However, in sharp contrast to the II-VIDMS, the III-VI semiconducting host presents a two di-mensional (2D) nature, at the origin of the renewed in-terest in ultrathin layers of, e.g., GaSe and GaS . Theweak van der Waals bonding between the stacked fouratom thick layers (Se-Ga-Ga-Se) further enhances thetwo-dimensional nature of this crystal. Because of itsmarkedly nearly 2D structure, GaSe has been consideredin the past for angle-resolved photoemission (ARPES)experiments, and a recent study has refocused the in-terest on this aspect by providing high quality ARPESdata supported by band structure calculations of the bulkcrystal , while electronic structure calculations and op-tical spectroscopy experiments on few-layer GaSe sheetshave been recently published .The present study is focussed on the electronic prop-erties of the Mn-GaSe interface obtained by evaporatingMn ions on the (0001) surface of a ultra-high vacuumcleaved ǫ -GaSe single crystal. In the first part, we trackthe evolution of the Mn-GaSe interface by evaporating atroom temperature (RT) increasing quantities of Mn onthe GaSe surface. In this way we identify the interfacegrowth regimes, and in particular the balance betweencation diffusion through the surface and cation segrega-tion at the surface. Once the interplay between theseprocesses was assessed, we prepared a Mn:GaSe interfacewhere the Ga − x Mn x Se surface alloying through cationdiffusion is dominant over Mn segregation on the surface,and we studied the electronic properties through resonantphotoemission and X-ray absorption spectroscopies. Acomparison is drawn with the Mn:CdTe interface, whichdisplayed a similar transition from cation diffusion tocation segregation processes, as discussed in Ref. . Thisputs on a solid ground the early speculations on the sub-stitution of Ga by Mn in the GaSe lattice, and providesan experimental evidence of the capability of Mn to dif-fuse into the GaSe lattice. The similarity of the presentspectroscopic data with those found for Cd − x Mn x Te in-dicates that the diffusion process is rather efficient acrossthe surface of the GaSe system. Finally, a characteriza-tion of the magnetic properties is reported, based on astudy the temperature dependence of the magnetization.
II. EXPERIMENTAL DETAILS
The GaSe single crystals have been grown by theBridgman method . The crystals were cleaved in ultra-high vacuum conditions prior Mn evaporation. Mn layerswere deposited at RT by in-situ electron beam evapora-tion from an outgassed Mo crucible loaded with metallicMn flakes. An Omicron EFM-3 triple evaporator wasused in all experiments. The deposition rate was prop-erly calibrated before evaporation on the GaSe cleavedsurface. Three interfaces have been produced during theexperiments. (i) The first was obtained by evaporatingMn with a constant flux of 500 nA measured across theexit slit of the evaporator. No post-growth annealingwas carried out. The data collected from this interfaceare reported in Section III.A. After an overall 180 sec-ond deposition at this rate, the amount of deposited Mnwas estimated to be 2 ML. The second sample (SectionIII.B) was obtained by evaporating 2.4 ML of Mn, and byannealing the interface at 400 ◦ C in ultra-high vacuumfor 10 minutes to favor the Mn diffusion process in theGaSe lattice. (iii) The third sample (Section III.C-III.E)was produced at the BACH beamline, by depositing asub-ML of Mn at room temperature. As in the previouscase, a 10 minute annealing in ultra-high vacuum at 400 ◦ C was carried out after the Mn evaporation.X-ray absorption and resonant photoemission spec-troscopy measurements were performed at the BACHbeam line of the Elettra Synchrotron Light Source.The X-ray photoemission (XPS) data have been col-lected at the Surface Science and Spectroscopy Lab ofthe Universit´a Cattolica (Brescia, Italy) with a non- monochromatized dual-anode PsP x-ray source and aSCIENTA R3000 analyser, operating in the transmis-sion mode, which maximizes the transmittance and workswith a 30 ◦ acceptance angle. The stoichiometry of theMn:GaSe interfaces produced during the experiments wasestimated by measuring the peak area of the Mn, Ga orSe atomic species, properly weighed by the photoemissioncross sections and the analyzer transmission. The surfacesensitivity (XPS probe depth) at the various kinetic en-ergies (KE) was evaluated by Monte-Carlo calculationsof the depth distribution function with the algorithm de-scribed in Ref. , in order to include inelastic as wellas elastic electronic scattering, in the so-called transportapproximation . We define the calculated probe depthfor photoemission as the maximum depth from which95% of all photo-emitted electrons can reach the sur-face. In case of exponential attenuation of the signal,this would correspond to three times the electron escapedepth.Magnetic measurements have been performed bymeans of a SQUID magnetometer. A 5 kOe magneticfield has been applied parallel to the sample plane tostudy the temperature dependence of the magnetizationin the range 2-360 K. A hysteresis cycle has been also col-lected at room temperature with magnetic field rangingbetween ±
10 kOe.
III. RESULTS AND DISCUSSIONA. Cation diffusion vs. surface segregation
In Figure 1 a sequence of Mn 2p XPS spectra collectedafter each Mn evaporation at RT on the freshly cleaved ǫ -GaSe surface is shown. As can be noticed, the Mn 2p / and Mn 2p / spin-split features of the Mn 2p core lineare detectable. Each spin-orbit split component is com-posed of two broad peaks, denoted as A and B for the2p / component, and C and D for the 2p / component.These spectra are typical of Mn in DMS, as will be dis-cussed in the next Subsection. Furthermore, the overallMn 2p lineshape changes as the amount of deposited Mnincreases. In particular, starting from spectrum (f) (120seconds) a peak ascribed to metallic Mn is clearly de-tectable on the low BE side of the Mn 2p main line, witha BE of 638 eV. The Mn 2p XPS spectrum of a Mn thickfilm is also shown below spectrum (j) (shaded area), tohelp identifying the contribution of metallic Mn in thedata.In parallel with the Mn 2p core levels, also the Ga 2pcore lines have been collected. The integrated peak areaof the Ga 2p is shown in the top panel of Figure 2. TheGa 2p signal attenuation with deposition time is well de-tectable. The low kinetic energy of the Ga 2p electronsmakes the spectra more sensitive to the surface layers. Itis important to note that the Ga 2p attenuation seemsto follow two regimes. From 0 to about 50 seconds theattenuation is steeper than that measured after 50 sec- I n t en s i t y ( a r b . un i t s )
665 660 655 650 645 640
Binding Energy(eV)
Mn metal
ABCD
Mn 2p XPS
Mn 2pMn 2p (a) 20"(b) 40"(c) 60"(d) 80"(e) 100"(f) 120"(g) 140"(h) 160"(j) 180"
FIG. 1: (Color online) Sequence of Mn 2p spectra collectedwith Al K α photon source for Mn deposited in steps (20 sec-onds for each step) on ǫ -GaSe(0001) surface. The time labelindicated the total Mn incremental deposition time at eachstep. The spectra have been normalized to the peak height ofthe Mn 2p / component. onds. A similar behavior was found for Mn depositedon CdTe single crystals, and ascribed to Cd substitutionwith Mn . As in that case, we can ascribe the earlysteep decrease to Ga cation substitution with Mn (corre-sponding to Cd cation substitution with Mn in the CdTehost crystal), and the following slower decrease with anoverall screening of the Ga signal due to the growth ofa Mn overlayer on the crystal surface. Therefore, aftera determined Mn coverage, Mn diffusion to layers un-derneath the surface and Ga substitution processes arehindered, resulting in the build-up of Mn overlayers onthe surface.It should be noted that electrons from the Ga 2p corelevels are emitted with a kinetic energy of 361 eV. TheXPS probe in GaSe for these electrons is quite surfacesensitive, as the calculated probe depth is about 2.5 nm.Unfortunately, Se does not display core levels with a com-parable kinetic energy, as was the case of Cd 3d and Te3d , and it is not possible to evaluate differences in thesignal attenuation rate of Se surface sensitive emissionwith respect to the Ga 2p case. To overcome this limit,a set of shallow core levels (Se 3d, Mn 2p, and Ga 3d)with much higher kinetic energies (about 1450 eV) hasbeen collected. With this KE, the probe depth is about7.5 nm for the three shallow core level emissions. On G a P ea k A r ea ( a r b . un i t s ) Deposition Time (sec)Ga2p XPS peak area, measured KE @ 361 eV
Regime IDiffusion Regime IISurface Segregation P ea k a r ea r a t i o ( a r b . un i t s ) Deposition Time (sec) Se3d/Mn3p Ga3d/Mn3p
FIG. 2: (Color online) Top panel: Integrated intensities ofGa2 p core lines measured after each step of Mn deposition on ǫ -GaSe surface. Bottom panel: Se3d/Mn3p (filled squares)and Ga3d/Mn3p (filled triangles) peak area intensity ratiosvs. deposition time. this basis, the Ga3d/Mn3p and Se3d/Mn3p ratio vs. de-position time are shown in the bottom panel of Fig.2.As can be observed, Ga and Se signal have a steep de-crease down to about 100 sec Mn deposition time, andGa decreases more rapidly than Se, suggesting that Ga issubstituted by Mn in the lattice. Above this limit bothGa 3d and Se 3d shallow core levels display a much lowersignal attenuation, indicating that different mechanismsare at work to screen the photoemission signal, very likelythe prevalent growth of a Mn overlayer on the surface.Consistently with the higher probe depth, the change ofgrowth regime here is found around 100 seconds ratherthan at about 60 seconds. We rationalize this finding byassuming that the substitution of Ga with Mn is achievedat the early stages of deposition for the topmost Ga layers(preferentially probed by the Ga 2p attenuation), whilethe substitution mechanism in the underlying layers re-mains active also at further deposition steps, determinedby the kinetics of Mn diffusion processes.The build up of Mn on the surface leads to distinctfeatures of the sample surface, as shown in Fig.3, toppanel. In fact, while the freshly cleaved surface showsa quite flat profile with reduced roughness (Fig.3, bot-tom panel), the surface obtained after the last depositionstage is quite rough, with round-shaped protrusions thatcould be related to Mn segregation on the surface.The Mn stoichiometry x in the Ga − x Mn x Se alloy wasalso estimated by measuring the intensity of the Mn 2pcore level peak with respect to the Ga 3d and Se 3d corelines. After proper normalization to photoemission cross-sections and analyzer transmission, assuming a uniformdistribution of Mn ions through the topmost surface lay-ers, the amount of Mn diffused into the crystal after 60”evaporation (spectrum (c) of Fig.1, roughly at the bor-der between the two deposition regimes evidenced in thebottom panel of Fig.1) resulted to be x=0.056 ± − x Mn x Se.
B. Core level photoemission of the Ga − x Mn x Sealloy
After the preliminary work so far described, wehave been able to identify the spectroscopic signaturesof the conditions where the alloying process, yieldingGa − x Mn x Se, was dominant over the Mn segregation onthe surface. The Mn evaporation was repeated on freshlycleaved GaSe surfaces and, after Mn evaporation, anneal-ing at 400 ◦ C in ultra high vacuum was also carried outto further induce the alloying process through diffusion.The amount of evaporated Mn (2.4 ML) exceeds the over-all amount of Mn of Section III.A by a factor 1.2, anda larger contribution of metallic Mn is expected beforeannealing. However, the larger amount of Mn allowedus to collect data with a a better statistics and obtaina good reference Mn 2p XPS spectrum of the Mn:GaSealloy after the annealing treatment.Figure 4 shows the Mn 2p core line of the Mn:GaSe in-terface prior and after annealing in vacuum. As observed,the four spectral features of the vacuum annealed inter-face present strong satellites (B and D) on the high BEside of the main lines (A and C) of the Mn 2p spin-orbitsplit components.We can exclude the presence of relevant oxygen con-taminations, as the measured Mn 2p XPS lineshape isquite different from that of MnO (Figure 4-d). Further-more, we did not observe any signal from oxygen withinthe sensitivity of our probe. The spectrum of the as-deposited film (Figure 4-b) shows two broad spin-orbitsplit components, suggesting the presence of several con-tributions that could be ascribed to both metallic Mnand Mn diluted in the GaSe lattice. In fact, a compari-son with the Mn 2p XPS core line from metallic Mn (Fig-ure 4-a) indicates that the M1 and M2 features (markedby dashed vertical lines) can be ascribed to metallic Mn.These features are progressively quenched with anneal-ing treatments (Figure 4-c), indicating that the anneal-ing is quite effective to prevalently induce a substitutionof Ga by Mn atom, rather than a clustering of Mn on theGaSe(0001) surface.The peaks A and C are separated by the spin-orbitinteraction and the width of these two peak is ascribed
FIG. 3: (Color online) AFM images (1 µm x 1 µm size) of thefreshly cleaved ǫ -GaSe(0001) surface (bottom panel), and ofthe Mn-GaSe(0001) interface after the last Mn evaporationstage at room temperature (top panel). I n t en s i t y ( a r b . un i t s )
665 660 655 650 645 640 635
Binding Energy (eV) (a) Mn metal(b) Mn:GaSe before vac. annealing(c) Mn:GaSe after vac. annealing(d) MnO
M1M2 ABCD
FIG. 4: (Color online) XPS spectra of the Mn 2p core linescollected from (a) a thick Mn metallic film, (b) the Mn:GaSeinterface before UHV annealing, (c) Mn:GaSe interface afterUHV annealing (d) a MnO single crystal to disorder effects, related to replacement of Ga atom byMn. On the high BE side of these peaks, two satellites arealso detectable (B and D), quite similar to those found inMn-based DMS, such as Cd − x Mn x Te, Zn − x Mn x S andGa − x Mn x As . They are ascribed to charge transfereffects from the ligand anions (Te, S or As, respectively)to the 3d levels of Mn cations. These effects are usuallyaccounted for in the frame of a configuration interactionmodel where the electronic states involved in the photoe-mission process are described by a linear combination ofseveral configurations (see, e.g. Ref and Refs. therein)such as 3d n , 3d n +1 L, 3d n +2 L ,where L represents a holein the ligand created by the charge transfer. The ligand-to-3d charge-transfer energy is defined by ∆=E(d n +1 )-E(d n ). The intensity of B and D satellites varies depend-ing on the charge transfer energy ∆, as well as on thehybridization strength (T) between the p and d orbitalsinvolved in the charge transfer process (here from Se 4 p to Mn 3 d ). Therefore, the line-shape analysis of the Mn2p core levels shown in Figure 4 provides an evidenceof Mn-Se hybridization effects for the Mn:GaSe system.In Subsection D, a detailed calculation of the Mn spec-tral weight in photoemission through CI models will becarried out for the 3d levels in the valence band region.Finally, it is rather important to compare the presentresults with those obtained on the Fe-GaSe interface .The analysis of Fe 2p XPS lineshape in the Fe-GaSe in-terface does not provide evidence of Fe-Se hybridization,being the Fe 2p XPS spectra quite similar to that ofmetallic Fe, while Fe clustering effects are found to bedominant. In turn, our measurements on the electronicproperties of the Mn:GaSe interface have shown the capa-bility of Mn to diffuse into the lattice with a remarkablehybridizations with Se anions.At this stage we cannot exclude that annealing in vac-uum can also trigger Mn desorption effects, but the re-sults of the first deposition (Section III.A) clearly indi-cate that alloying (i.e. Mn diffusion though the latticeand Ga substitution with Mn without resorting to anyannealing) is the dominant process. According to theGaSe layered crystal structure, the UHV cleaving shouldoccur between the two facing Se layers (i.e. through abreaking of the Se-Se weak van der Waals bonds). Theweakness of this bond should also favor the diffusion ofMn through the lattice between the Se layers, and even-tually the Mn hybridization with Se. C. X-ray absorption from the Ga − x Mn x Se alloy
The Mn L-edge XAS spectra are shown in Figure 5. Inparticular, the data obtained after the Mn deposition (e)and after annealing at 400 o C and collected at RT (d) arepresented. The as-deposited Mn-doped GaSe (e) showsbefore annealing the presence of both metallic and a re-acted Mn-GaSe interface, as appearing from an overallsmooth XAS lineshape, with minor modulations that willultimately evolve into the post-annealing XAS lineshape(d). Indeed, after annealing at 400 o C, sharper features(labeled as A, B, C, D, and E) appear, and the compar-ison with multiplet calculations for a Mn → electric-dipole allowed transition (c) unambigu-ously shows that the measured spectrum can be ascribedto a Mn ion in the GaSe matrix. Similar transitioncalculated for Mn (b) and Mn (a) ions do not fit theexperimental data. The remarkable similarity with the XAS spectrum pre-dicted for the Mn calculation is particularly helpful forthe interpretation of the electron spectroscopy results.This will justify the assumption at the basis of CI calcu-lations for the Mn 3d spectral weight in the valence band(see next Section) where an Mn ion will be assumed asthe ionic configuration in the parameterized CI model.In the bottom panel of Figure 5, we have shown a setof calculated Mn XAS spectra in a tetrahedral T d sym-metry, starting from zero crystal field to 10Dq=2.5 eV,where 10Dq is the crystal field splitting of the 3d orbitals.It is important to note that crystal field effects seem tobe rather limited up to 10Dq=0.75 eV. At this energy,two features appear on the low photon energy sides ofthe calculated Mn L III and Mn L II edges, that have nocounterpart in the experimental data. Therefore, we as-sume that crystal field effects are negligible. This remarkwill also be at the basis of the CI model for the valenceband calculations, where crystal field splitting will be setto zero. D. Valence band resonant photoemission at theMn L-edge
The valence band spectra of the clean GaSe and of theMn-doped, annealed, GaSe single crystals are shown inFigure 6 (b) and (a), respectively. The photoemissionspectra have been collected with a photon energy of 797eV and have been normalized to the maximum of the va-lence band emission (peak A). Both spectra show a mainline with three features labeled A’, A, and B, and a peakC at higher binding energies. When Mn is evaporated onthe GaSe cleaved surface, the main changes that can beobserved are the appearance of a feature A” at the Fermiedge, and an increase of the spectral weight in the regionsbetween the peaks A and B and the peaks B and C. Thecurve displayed in Fig 6 (c) represents the difference be-tween the spectra (a) and (b). This difference confirmsthe increase of spectral weight upon Mn deposition inthe 3-7 eV binding energy range, and in the region justbelow the Fermi edge (BE=0-2 eV), while a decrease ofthe intensity is found below peak C after Mn depositionand annealing at 400 ◦ C.The main features of the present experimental datacan be interpreted on the basis of the calculated DOS sofar published . Indeed, the observed experimentalpeaks have a counterpart in, e.g., the DOS calculated byPlucinki et al. that identify three regions (I, II, and III)in the valence band (see Fig.3 of Ref. ). Region I corre-sponds to the observed peaks A and A’, region II to peakB and region III to peak C. From the analysis of the pro-jected DOS on Se s,p states and Ga s,p states, it is ratherinteresting to observe that Ga mainly contributes in theregion below peak C, i.e. in the region where a spectralweight decrease is observed upon Mn doping and anneal-ing. This is in agreement with the assumption of Gasubstitution with Mn atoms, as remarked in Subsection I n t en s i t y ( a r b . un i t s ) Mn:GaSe before vac.anneal Mn:GaSe after vac.annealMn (3d )Mn (3d )Mn (3d ) (a)(b)(c)(d)(e) A B C D E (p)(i)(j)(k)(l)(m)(n)(o) (g)(h)(f)
FIG. 5: (Color online) Top panel. Experimental Mn L-edgeXAS spectra of Mn deposited on GaSe surface before anneal-ing figure (e), after annealing figure (d). Calculated atomicMn L-edge XAS spectra (zero crystal field) for the +3 oxi-dation state (3d ) (a), +1 oxidation state (3d ) (b), and +2oxidation state (3d ) (c). Bottom panel. Comparison betweenthe experimental XAS spectrum of the annealed Mn:GaSe in-terface (shaded area) and the calculated XAS spectrum forthe atomic Mn configuration with a tetrahedral crystalfield 10 Dq ranging from 0.0 eV to 2.5 eV A. Finally, the states appearing at the Fermi edge (A”)can be ascribed to unreacted Mn at the surface.In order to enhance the Mn contribution to valenceband states, a ResPES study at the Mn 2p-3d absorp-tion edge has been carried out. The results are shownin Fig 7. As first, on the XAS spectrum (top panel) thephoton energies (a to j) selected to collect ResPES dataare indicated. The whole set of ResPES data is shown inthe bottom panel. The data span a photon range acrossthe Mn L
III threshold. The VB spectra show a clear en-hancement of the spectral weight with a photon energy I n t en s i t y ( a r b . un i t s )
12 10 8 6 4 2 0 -2Binding Energy (eV) (a)-(b), x 2
C B A A' A'' (a)(b)(c) (a) Mn:GaSe (b) GaSe (c) difference (a)-(b)
FIG. 6: (Color online) Valence band spectra of the clean GaSe(b) and of the Mn-doped, annealed, GaSe single crystal (a).Difference between the spectra of the doped and the cleansystem (c). of about 640 eV. At this energy a peak around BE= 4.5eV in the valence band shows a remarkable intensity en-hancement. The difference between the resonant (e) andoff-resonance (a) spectra (hereafter denoted as resonat-ing spectral weight, RSW) is shown in panel C. Here itis clearly seen that the RSW is determined by a peak at4.5 eV (R2) and by two broad features at about BE= 6-8eV (R1) and BE= 1-3 eV (R3).We have used spectrum (e) rather than (f), as it isknown that for spectra collected with photon energiesat the maximum of the absorption threshold [spectrum(f) in the present case] the weight of the Auger emis-sion channel is not negligible and the maximum of thevalence band emission is already shifted to higher BEwith respect to spectrum (e), where the resonant RamanAuger channel (RRAS) is still dominant.It is worth observing that the large R2 resonance isfound in the energy range where the difference spectrumof Fig.6-c shows the largest difference between the dopedand undoped surfaces, confirming that a large Mn contri-bution should be expected at least in this energy range.Further insight into the origin of the three spectralfeatures (R1, R2, and R3) in the RSW can be obtained atthe light of parameterized CI calculations for the valenceband.Impurity-cluster CI calculations of the Mn 3d spectralweight are shown in Fig.8. In the CI approach severalconfigurations, denoted as d n , d n +1 L ( L denotes a lig-and hole, here a hole on Te 4p orbitals) are used to de-scribe the open shell of the 3d transition metal ion dur-ing the photoemission process. The spectral weight ina photoemission experiment is calculated, in the suddenapproximation, by projecting the final state configura-tions ( | Ψ i,fs i ) on the the ground state, i.e. I XP S ( BE ) ∝ P i |h Ψ GS | Ψ i,fs i| δ ( BE − ε i ) where | Ψ GS i represents theground state (GS) wavefunction, and the sum is run I n t en s i t y ( a r b . un i t s ) Photon Energy (eV)
Mn 2p-3d XAS on Mn:GaSe(a) (b)(c)(d)(e)(f) (g)(h)(i) (j) I n t en s i t y ( a r b . un i t s ) R1 R3
10 8 6 4 2 0
Binding Energy (eV) (a) (b)(c)(d)(e)(f) (g)(h)(i)(j)
RSW (e)-(a)R2
FIG. 7: (Color online) Top panel: XAS spectrum of the Mn-GaSe interface collected at the Mn 2p-3d edge . The dotsindicate the photon energies chosen to collect the, resonant,valence band photoemission spectra. Botton panel: Set ofvalence band spectra collected at different photon energiesacross the Mn 2p-3d edge. Resonant spectral weight (RSW, x 1.2 ) obtained from the difference between the off-resonancespectrum (a) and resonant VB spectrum (e) collected at 635.5eV and 639.5 eV, respectively. Difference between the reso-nant and off-resonance spectra. TABLE I: Parameter values used for the parameterized CIcalculation of the Mn 3d spectral weight. The following valuesare specified: the charge transfer energy ∆, the d-d correlationenergy U, and the pd σ hybridization integralHost ∆ U pd σ Matrix (eV) (eV) (eV)GaSe 2.95 6.4 -1.25CdTe 2 5 -1.1 over all final state configurations | Ψ i,fs i with energy ε i .Where required, proper fractional parentage coefficientscan be used, as was done in the present case.The calculations have been carried out following thescheme presented by Fujimori et al. for several Mndoped semiconductors, based on a 3d initial state of thetransition metal atom . The results are shown in Fig.8,along with those obtained on a Mn-doped CdTe singlecrystal . The parameter set used in the calculation isreported in Table I.The comparison with the RSW detected under thesame conditions for the Mn-doped CdTe crystal is ratherinteresting. The gray spectrum (empty circles) in thetop panel represents the resonance spectrum collectedfrom an heavily doped CdTe single crystal (adapted formRef. ), which shows a contribution at the Fermi level ofmetallic Mn states similar to the states observed in theGaSe host crystal, though these states are less intense inthe GaSe host system. Also in the CdTe case, three peaksappear in the resonant spectrum, but the relative weightand width of these peaks are different from the Mn:GaSecase. Peak R2 is larger in the GaSe host, and the sep-aration between peak R1 and R2 is larger in the GaSehost with respect to the CdTe case. These differenceshave been considered as constraints in the calculations.In particular, the calculations for the GaSe case havebeen obtained by setting the crystal field to zero and byconsidering a larger charge transfer energy (2.95 eV) ascompared to the CdTe host (2.0 eV, Table I). The firstassumption is justified by the lack of relevant crystal fieldeffects observed in XAS, whereas the second is justifiedby the larger band gap of GaSe with respect to CdTe,as the charge transfer energy is usually assumed to scalewith the energy gap of the host crystal (E g =1.49 eV inCdTe, E g =2.02 eV in GaSe). This choice of the param-eter set resulted in a broadening of the calculated peakbelow R2 and in the intensity increase and BE shift of thecalculated spectral weight below R1. Also the calculatedspectral weight below R3 is increased, in agreement withthe measured data. We used the free ion Racah param-eter for Mn , B=0.126, C=0.421, and the crystal fieldwas set at 0.4 eV for the CdTe host and 0 eV for theGaSe host. I n t en s i t y ( a r b . un i t s )
15 10 5 0 -5
Binding Energy (eV)
R1 R2 R3
Mn:CdTe Mn:CdTe + metallic Mn Mn:GaSe RSW
Mn:GaSeMn:CdTe
FIG. 8: (Color online) Experimental data and CI calcula-tions for the Mn:CdTe (top panel, adapted form Ref. ) andthe Mn:GaSe (bottom panel) interfaces. The shaded areasrepresent the CI calculations, while the vertical bars indicatethe eigenenergies obtained by solving the Hamiltonian ma-trix. The height of each bar is proportional to the square ofthe projection of the eigenvector on the ground, initial, state.All energies are given in eV. E. Magnetic properties
Figure 9 shows the temperature dependence of themagnetic susceptibility χ , evaluated with respect to theunit-mass of the considered sample. The overall suscep-tibility is always negative disclosing a dominant diamag-netic behavior. This is confirmed by the negative slopeof the M vs. H curve measured at room temperature,reported in the inset on Fig. 9. The negative dominantbackground is not surprising as the GaSe host is dia-magnetic and the Mn:GaSe interface represents a smallfraction of the sample (few nanometers with respect tothe overall single crystal thickness, that is about 100 mi-crons). However, though negative, the magnetic suscep-tibility curve reveals a signature of paramagnetism, witha decrease of values with temperature that is fit by thecurve representing the Curie-Weiss law (continuous linein Fig. 9), with very small Weiss constant value (about-3 K).This effect can be ascribed to the topmost layers of thesample, i.e. those hosting the Mn dopant after the tem- perature induced diffusion. This behavior is quite differ-ent from that observed in bulk Mn:GaSe single crystals ,which show a complex behavior with a quenching of mag-netization at low temperatures, i.e., where we observe the -3.2-3.1-3.0-2.9 χ ( e m u / g O e ) x -202 M ( e m u / g ) x -10000 -5000 0 5000 10000H (Oe) FIG. 9: (Color online) Magnetic susceptibility vs. tempera-ture curve for the Mn-doped GaSe sample. Inset: magnetiza-tion M vs. external applied field H measured at 300 K. steady increase of paramagnetic signal. The magnetic be-havior is also different in many details from that observedfor other known bulk phases that can be regarded as pos-sible segregated phases in the growth of the Mn:GeSe in-terface, such as MnSe and MnSe , and MnGa Se .In turn, closer similarities, i.e. a virtually paramagneticbehavior, are found with respect to the case of monocliniccrystals of (Ga − x Mn x ) Se discussed in Ref. . IV. CONCLUSIONS
We have been able to prepare well characterizedMn:GaSe interfaces, with evidence of the formation ofa Ga − x Mn x Se alloy below the GaSe surface. Alloying isobtained already after evaporation at room temperature.The Mn deposition in the present study spanned tworegimes. In the early regime the Mn diffusion throughthe surface was the dominant mechanism, while in thesecond regime the segregation of Mn layers on the GaSesurface was the most likely process. Unlike Fe-GaSe in-terfaces, where iron clustering effects are dominant andno trace of Fe-Se hybridization is found upon the anal-ysis of Fe 2p XPS lineshape , our measurements onthe electronic properties of the Mn:GaSe interface haveshown the capability of Mn to diffuse into the lattice witha remarkable hybridizations with Se anions. Magneticmeasurements evidence a paramagnetic behavior for theMn-doped interface, while the dominant behavior is dia-magnetic, due to the bulk of the GaSe host crystal. K. Kato, N. Umemura, Optics Letters, , 746 (2011); A. Segura, J. Bouvier, M.V. Andres, et al.
Phys. Rev. B, , 4075 (1997); S. Nusse, P.H. Bolivar, H. Kurz, et al. , Phys. Rev. B, ,4578 (1997); H. Ertap, G.M. Mamedov, M. Karabulut, A. Bacioglu,Journ. of Luminescence, , 1376 (2011) P.A. Hu, Z.Z. Wen, L.F. Wang, P.H. Tan, K. Xiao, ACSNano, , 5988 (2012) H. Peng, S. Meister, C.K. Chan, X.F. Zhang, Y. Cui, NanoLetters, , 199 (2007) D.J. Late, B. Liu, J.J. Luo, A.M. Yan, H.S.S.R. Matte, M.Grayson, C.N.R. Rao, V.P. Dravid, Advanced Materials, T. M. Pekarek, B. C. Crooker, I. Miotkowski and A. K.Ramdas, Journ. of Appl. Phys. T. M. Pekarek, C.L. Fuller, J. Garner , B.C. Crooker, I.Miotkowski, A.K. Ramdas, Journ. of Appl. Phys. , 7030(2001) T.C. Lovejoy, E.N. Yitamben, S.M. Heald, F.S. Ohuchiand M.A. Olmstead, Appl. Phys. Lett. , 241907, (2009). T.C. Lovejoy, E.N. Yitamben, S.M. Heald, F.S. Ohuchiand M.A. Olmstead, Phys. Rev. B , 155312 (2011). S.F. Dubinin, V.I. Maksimov, and V.D. Parkhomenko,Crystallography Reports, , 1165 (2011) L. Plucinski, R.L. Johnson, B.J. Kowalski, K. Kopalko,B.A. Orlowski, Z.D. Kovalyuk, G.V. Lashkarev, Phys. Rev.B D.V. Rybkovskiy, N.R. Arutyunyan, A.S. Orekhov, I.A.Gromchenko, I. V. Vorobiev, A.V. Osadchy, E.Y. Salaev,T.K. Baykara, K.R. Allakhverdiev, E.D. Obraztsova, Phys.Rev. B, , 085314 (2011) L. Sangaletti, S. Pagliara, I. Dimitri, F. Parmigiani, A.Goldoni, L. Floreano, A. Morgante, V.F. Aguekian, Sur-face Science
508 (2004) M. Yamamoto, H. Mino, I. Akai, T. Karasawa, V.F.Aguekian, Journ. of Luminescence, , 275 (2000) W. S. M. Werner, Surf. Int. Anal. , 141 (2001). A. Jablonski, Phys. Rev. B , 16470 (1998). L. Sangaletti, A. Verdini, S. Pagliara, G. Drera, L. Flore-ano, A. Goldoni, A. Morgante, Phys. Rev. B , 245320(2010) L. Sangaletti, M.C. Mozzati, G. Drera, V. Aguekian, L.Floreano, A. Morgante, A. Goldoni, G. Karczewski Appl.Phys. Lett. , 142105 (2010) J. Okabayashi, A. Kimura, O. Rader, T. Mizokawa, A.Fujimori, T. Hayashi, M. Tanaka, Phys. Rev. B R4211(1998) A.R. de Moraes, D.H. Mosca, W.H. Schreiner, J.L.Guimaraes, A.J.A. de Oliveira, P.E.N. de Souza, V.H. Et-gens, M. Eddrief, Journal of Magnetism and Magnetic Ma-terials , 1551 (2004) A.R. de Moraes, D.H. Mosca, N. Mattoso, J.L. Guimaraes,J.J. Klein, W.H. Schreiner, P.E.N. de Souza, A.J.A. deOliveira, M.A.Z. de Vasconcellos, D. Demaille, M. Eddrief,V.H. Etgens, J. Phys.: Condens. Matter , 1165 (2006) Zs. Rak, S.D. Mahanti, Krishna C. Mandal, N.C. Fernelius,Journ. of Phys. Chem. of Solids, , 344 (2009) M.O.D. Camara, A. Mauger, I. Devos, Phys. Rev. B ,125206 (2002) See, e.g., L. Sangaletti, S. Dash, A. Verdini, L. Floreano, A.Goldoni, G. Drera, S. Pagliara, and A. Morgante, J. Phys.:Condens. Matter , 235502 (2012), and Refs. therein T. Mizokawa and A. Fujimori, Phys. Rev. B A. Fujimori, M. Sacki, N. Kimizuka, M. Taniguchi, S. Suga,Phys. Rev. B Q. Peng, Y. Dong, Z. Deng, H. Kou, S. Gao, Y. Li, Journ.Phys. Chem. B M. Morocoima, M. Quintero, E. Quintero, J. Gonzlez, R.Tovar, P. Bocaranda, J. Ruiz, N. Marchn, D. Caldera andE. Calderon, Journ. of Appl. Phys. , 053907 (2006) J.B.C Efrem Dsa, P.A. Bhobe, K.R. Prilkar, A. Das, P.S.R.Krishna, P.R. Sarode, R.B. Prabhu, Pramana-Journal ofPhysics,63