Surface & Coatings Technology 265 (2015) 145–153

Contents lists available at ScienceDirect

Surface & Coatings Technology

j ourna l homepage: www.e lsev ie r .com/ locate /sur fcoat

Water content and high temperature influence on the oxidation behaviorof manganese and silicon thin films on iron matrix

Carina Hambrock a,b, Andrei Ionut Mardare a,c,⁎, Achim Walter Hassel a,c

a Institute for Chemical Technology of Inorganic Materials, Johannes Kepler University Linz, Altenberger Str. 69, 4040 Linz, Austriab voestalpine Stahl GmbH, Research & Development, voestalpine-Straße 3, 4020 Linz, Austriac Christian Doppler Laboratory for Combinatorial Oxide Chemistry at the Institute for Chemical Technology of Inorganic Materials, Johannes Kepler University Linz,Altenberger Str. 69, 4040 Linz, Austria

⁎ Corresponding author. Tel.: +43 732 2468 8702.E-mail addresses: carina.hambrock@voestalpine.com (

andrei.mardare@jku.at (A.I. Mardare), achimwalter.hassel

http://dx.doi.org/10.1016/j.surfcoat.2015.01.0470257-8972/© 2015 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 30 September 2014Accepted in revised form 20 January 2015Available online 30 January 2015

Keywords:High temperature oxidationThin filmsThermal evaporationWater content

Mn and Si thin films were deposited by thermal evaporation on different Fe-substrates. The use of a mobileshutter during deposition allowed a thickness gradient to be obtained along each sample favoring thicknessdependent annealing investigations of Mn and Si films on Fe. The annealing behavior of the prepared thinfilmswas investigated at temperatures ranging between 750 and 950 °C under different atmospheric conditions.Not only the temperature, but also the O2 partial pressure used as well as the film thickness influenced the oxi-dation behavior of the thin film samples during annealing. The Mn thin film deposited on the Fe substrate andannealed at varying conditions revealed a strong crystal growth. This depended on all annealing variablesresulting in a grain size decreasewith decreasing temperature, O2 partial pressure and film thickness. The Si coat-ed substrate showed a different oxidation behavior as compared to theMn case, revealing mainly a thickness in-dependent homogeneously oxidized surface. However, the heat treatment conditions were still evaluated asinfluencing factor for the Si thin film gradient because under specific conditions Fe diffusion through the Sithin film could be observed.

© 2015 Elsevier B.V. All rights reserved.

1. Introduction

The addition of Si andMn as alloying elements is very popular in thesteel industry. Both elements can induce specific properties to the steelwhen added within certain limits. A high amount of Si, for example, isadded to electrical steels in order to ensure excellent magnetic proper-ties, such as low core loss [1–4]. However, themaximumamount of Si insteel is limited to 6.5 wt.% because of the industrial production process,where more Si would result in a high brittleness of the steel [5–7].In order to provide best magnetic but also best metallurgical properties,not only Si but also Mn is added to electrical steels in small amounts(b0.5 wt.%). A similar situation occurs for high Mn-steels. Higherstrength and improved ductility during hot processing are only someof the properties that can be achieved by the addition ofMn [8,9]. There-fore, Mn in amounts of up to 30 wt.% is added e.g. to achieve TRansfor-mation or TWinning Induced Plasticity (TRIP or TWIP) steels, in orderto obtain both improved ductility and high-strength [10–13], or to pro-duce Hadfield steels with increased wear and abrasion resistance [14,15]. Additionally, Si is added to the TRIP and TWIP steels in order to pro-vide superior impact strength [11]. The oxidation behavior of steels

C. Hambrock),@jku.at (A.W. Hassel).

during annealing is highly relevant, possibly influencing the final prop-erties. Both kinds of steels mentioned (high Mn-steels and high Si-steels) are industrially produced under H2–N2 atmospheres at tempera-tures of up to 1000 °C. These annealing atmospheres are protective for Febut usually selective oxidation of Si andMn can still occur since their ox-ides are thermodynamically more stable as compared to Fe oxides [16].The dependence of the atmospheric conditions on the oxide layer forma-tionwas previously investigated [17–20]. This selective oxidation is det-rimental since it can induce failures during the industrial productionprocess which are more pronounced at an increased alloying content[5,21–26]. Especially the technology of coating the steel surfaces withZn is of highest relevance for the steel industry but it is still problematicfor these kinds of steels [13,23,27–29]. Previous investigations describedin literature are usually performed on industrial or laboratory alloys andtherefore limited by the abovementioned alloying contents.

An experimental approach via physical vapor deposition (PVD) waschosen in the present work for producing Fe–Si and Fe–Mn modelsamples. The preparation of both Fe–Mn and Fe–Si combinatorial thinfilms is described in literature, mainly by using complex chemicalvapor deposition (CVD) and magnetron sputtering methods [6,30–32].With those methods, the aforementioned alloying content restrictionsin the steel are achievedwithout facing the challenges presented duringhot-rolling. However, the alloying content restrictions are usually notexceeded.Overcoming these restrictionswith the help ofmodel samples

146 C. Hambrock et al. / Surface & Coatings Technology 265 (2015) 145–153

might be an advantage for increasing the visible effect of the alloyingcontent on the final material properties when considering the industrialproduction process with its specific annealing conditions. Therefore, anew approachwas chosen in this work by using pure Si and/orMn coat-ings deposited as thin films on the surface of Fe substrates. During thisinvestigation, the elemental thin filmswere obtained as thickness gradi-ents on Fe bulk substrates. The following step of annealing in differentatmospheres and at different temperatures induced various diffusionand oxidation phenomena. Those effects were not only dependent onthe annealing conditions but also on the thickness of each depositedthin film.

2. Experimental

Thin films of Si and Mn were deposited as thickness gradients viathermal evaporation along industrial Fe substrates (99.6%purity, suppliedby voestalpine Stahl GmbH). The substratematerial was a cold rolled andannealed material with a grain size of around 20 μm and a thickness of1.4 mm. It was precut in a rectangular shape of 10 × 77 mm2 for servingas substrate for the Si and Mn vapor condensation in vacuum. Beforebeing used for thin film deposition, the surfaces of the substrates wereground using diamond paste with a 1 μm particle size. Both Si and Mnthin films were deposited from high purity materials (N99.9%, AlfaAesar) using a self-developed thermal evaporator designed for combina-torial thin film library preparation. The particularity of the system isfound in the evaporation geometry [33]. Three thermal sources can beused simultaneously, if desired, and each of them is positioned off-center with respect to the substrate. This leads to an accentuated thick-ness gradient along the substrate due to the cosine law of evaporation[34,35]. In the present study, only one source (Si or Mn) was used at atime for depositing thinfilms on the Fe substrates. In order to accentuatethe obtained thickness gradient, a slowly moving shutter (with a speedof 0.8 mm s−1) was used in front of the Fe substrates. A schematic crosssection of this process is displayed in Fig. 1. The crucible with the evap-oration powder (olive and green colored, respectively) is heated resis-tively. Even though the thickness of the deposited film depends on thedeposition angle (obeying a cosine law), a significant thickness gradientalong the substrate can be obtained only by using amoving shutter. Thispartly blocks vapor phase atoms from reaching the substrate due to theline-of-sight limitation of the evaporation process. The wedge-like

Fig. 1. Schematic illustration of the thermal evaporation method with the correspondingmoving shutter in order to obtain a thickness gradient (wedge like structure) on thesurface of the substrate.

structure is finally obtained on the substratematerial. Crystal quartzmi-crobalances (QCMs) positioned above each thermal source were usedfor in situ thicknessmeasurements. For both deposited species, evapora-tion rates ranging from 0.4 to 0.5 nm s−1 were used leading to finalthicknesses above 100 nm at the thicker end of the samples. The basepressure of the deposition chamber was in the range of 10−5 Pa andthe thermal heat delivered to the samples during deposition due to radi-ation resulted in a temperature increase up to values of 80 °C for Si and40 °C for Mn.

After their deposition, the thin films were annealed using an IR-heated furnace. Due to the particularities of the setup, within the fur-nace it was possible to vary the annealing conditions in terms of tem-perature, atmosphere and water content. The annealing temperaturesand atmospheres used are displayed in Table 1. For both conditions,the heating and cooling rates were in the range of 25 °C s−1 and−5 °C s−1, respectively. The regulation of the water content in theatmosphere is a common method for precisely defining the O2 partialpressure during an annealing procedure. Therefore, the annealing con-ditions indicated in Table 1 do not only give the content of H2O butalso the calculated partial pressure of O2. Fig. 2 displays the Ellingham-type diagram revealing the correlation between the partial pressure ofoxygen and the temperature. The equilibrium partial pressure for themetal-oxide equilibrium of the elements Fe, Mn and Si is given in thediagram. At the same time, the two annealing conditions 1 and 2(described in Table 1) are also visualized. All these data were obtainedusing specialized software Factsage 6.3. The calculation was performedusing the thermodynamically most stable phases within the databasedepending on the temperature.

The microstructures of the as-prepared and of the annealed thinfilms were evaluated by using scanning electron microscopy (SEM),energy-dispersive X-Ray (EDX) spectroscopy and glancing incidenceX-Ray diffraction (GIXRD). The surface characterization via SEM wasperformed using a Field Emission Gun Microscope Supra35 from ZEISSequipped with a secondary electron (SE) detector working at an accel-eration voltage of 2 kV. The chemical composition was determinedusing an EDX system from EDAX with a Si(Li)-detector of 10 mm2 andan acceleration voltage of 5 kV. The XRD characterization was accom-plished with an X'Pert Pro system from PANalytical equipped with aCo Kα anode providing a wavelength of 0.178 nm. A voltage of 40 kVand a current of 40 mA were used while the divergence slit was set to1/32° at an incidence angle of 0.5°.

3. Results and discussion

3.1. Thin film characterization before annealing

Both types of thin films prepared on the surface of polished Fesubstrates were characterized before annealing using SEM, EDX andGIXRD. Figs. 3 and 4 present the SEM images describing the surfacemorphology of the Mn and Si gradient thin films before annealing, re-spectively. In the SEM images, both thin films reveal a smooth andhomogeneous surface morphology, independent of the thickness ofthe film. The thickness (as previouslymeasured in-situ by QCMs) variedfrom 150 nm to 25 nm for the Mn thin film (see Fig. 3a) and b), respec-tively) and from100 nm to 25nm for the Si gradient (see Fig. 4a) and b),respectively).

The pattern of straight lines visible in all images is a result of thepolished substrate material where the scratch remnants from the sub-strate polishing procedure were not completely covered by the filmsdue to their low thickness. Therefore, the effect is better visible at thethin end of each film. At the thick ends of both Mn and Si thin films(Figs. 3a) and 4a), respectively) not only the homogeneous surfacemor-phology, but also small round grains were visible. Using EDX analysis ofthose grains it could be revealed that they contain high amounts of O2

together with Mn and Si, respectively, as compared to the nearby sur-face regions. The low deposition pressure used did not allow a natural

Table 1Annealing conditions for the thin film gradients of Si and Mn.

Annealing Temperature/°C Atmosphere H2O/ppm p(O2)/Pa p(O2)/atm Time/s

Condition 1 950 10% H2 in N2 12,500 6.3 × 10−13 6.2 × 10−18 100Condition 2 950, 900, 800, 750 100% H2 35 4.9 × 10−20 to 3.7 × 10−24 4.8 × 10−25 to 3.6 × 10−29 200

147C. Hambrock et al. / Surface & Coatings Technology 265 (2015) 145–153

formation of those oxides. However, immediately after deposition thesenaturally formed oxide grains were identified on both samples. The factthat the density of the surface oxide grains was much higher on the Sithan on the Mn thin film gradient can be attributed to the lower O2 po-tential necessary for the formation of Si oxide as compared to that ofMn. In both cases presented in Figs. 3 and 4, the formation of nanopar-ticles wasmuch less pronounced at the thin side ofMn and Si films sug-gesting a Frank–van der Merwe thin film growth mechanism.

The GIXRD measurements on both thin films (not shown here) re-vealed the characteristic peaks of the Fe substrate material, as well asthe background signal. No additional characteristic peaks correspondingto crystalline phases were identified. These results suggest that the as-deposited Si and Mn films on Fe substrates are amorphous. This is gen-erally expected for thermal evaporation processes at room temperaturedue to the lowmobility of the surface atoms inhibiting crystalline struc-ture formation [36]. This lowmobility can also lead to defects occurringon the surface of the deposited thin films. These are energetically favor-able surface sites that promote the nucleation of oxides, especiallywhenthe partial pressure of O2 is raised. This occurred upon the removal ofthe samples from the vacuum chamber inducing the formation of natu-ral oxide grains at these favorable surface sites. The surrounding regionsrevealed ahigher oxidation resistance innatural ambient conditions [37,38].

The thickness gradients of both Si and Mn thin films were deter-mined before annealing by using an interference color recognitionmethod [39]. The thickness values obtained with the help of this meth-od correlated well to the thicknesses measured in-situ during deposi-tion using the quartz crystal microbalances fitted to the evaporationchamber.

3.2. Mn thin film after annealing

Two different sets of annealing parameters were applied fortreating the thin film samples. The first annealing condition (condition1) contained 10% H2 at a defined amount of water (see Table 1). As men-tioned before, Fig. 2 reveals the annealing conditions data described in

Fig. 2. Thermodynamic diagram revealing the partial pressure of oxygen vs. the tempera-ture in the range of 500 °C–1200 °C; lines with symbols correspond to the two differentannealing conditions (Table 1)while lineswithout symbols correspond to the equilibriumpartial pressure for the oxidation of Fe, Mn and Si depending on the temperature (valuesobtained by using Factsage).

Table 1 while additionally presenting the thermodynamic equilibriumof the metal-oxide reactions of Fe, Mn and Si. The first annealing condi-tion provided a high oxidation potential in the annealing atmosphere(significantly above the equilibriumpartial pressures ofMnand Si). How-ever, the oxidation potential was still sufficiently low to avoid the oxida-tion of the Fe substrate (see Fig. 2).

The surface appearance of the Mn thin film after annealing at condi-tion 1 is shown in Fig. 5, revealing its surface morphology for differentfilm thicknesses. On all four images, a clear crystal evolution on thesurface could be observed following an island growth mechanism. Thefacetted crystals occurred due to a surface energy minimization of thegrowing oxide in its cubic NaCl symmetry, allowing a dense packing.EDXmeasurements performed on individual crystals indicated a higheramount ofO andMn (corresponding to someMn-oxide) as compared tothe corresponding values measured nearby on the exposed substrateareas. The size of the crystals (as well as their surface coverage) de-creased with decreasing the thickness of the Mn thin film. Especiallyfor a thickness of 25 nm (Fig. 5d)), individual nanocrystals formed onthe surface exposing the Fe substrate underneath. At a thickness of150 nm, a dense layer of crystals could be observed with only fewvoids throughwhich pure Fe could be detected. At these highermetallicfilm thicknesses, the average size of the crystals is above 1 μm.

A new thin film was prepared via PVD in identical conditions andannealed at condition 2. Within this condition, the temperature variedfrom 950 to 750 °C at a constant gas atmosphere corresponding to100% H2 at low water content, as described in Fig. 2. This atmosphereis muchmore protective than the one used in condition 1 and it was ap-plied in order to provide less oxidizing conditions for the thinfilm.How-ever, it was not possible to prevent Mn and Si from oxidation duringthese experiments. Only an increase in temperature above 1100 °C atsimilar atmospheric conditions or a further decrease in dew point atsimilar temperatures could have prevented the oxidation of Mn. Fig. 6shows the SEM surface images of the Mn thin film after annealing atvarious temperatures under condition 2. Fig. 6a) reveals the regionwith a thickness of 200 nm and Fig. 6b) describes the region with thelowest thickness of 50 nm.

The surface appearance of the thin films annealed at different tem-peratures varied strongly. Evaluating the images of Fig. 6a) at all fourtemperatures, corresponding to the region of 200 nm thickness, theMn oxide crystals are clearly visible. At the highest temperature of950 °C the surfacemobility aswell as the partial pressure of O2 is higher,thus individual well-defined crystals can be observed on the surface.Those individual crystals were also visible at a temperature of 800 °C.However, upon lowering the temperature at the same atmosphericconditions (regarding water and H2 content), the O2 partial pressureof the atmosphere decreases. Less O2 as well as a lower surfacemobility(due to the lower temperature) is available for oxide crystal growth.Therefore, the size of the crystals formed upon annealing at 800 °Cwas much smaller as compared to the annealing at 950 °C and theiroverall appearance was much more homogeneous. The same could beobserved for the annealing at 750 °C, where the crystals are even small-er and less defined and the overall surface appears rather smooth. At atemperature of 900 °C the surface of the oxidizedMn thin film appearedto be different. Here, the smooth and homogeneous appearance ismuchmore defined and only the crystal planes but no individual crystal grains(or grain boundaries) were visible. This difference in surface morpholo-gy may be caused by a slightly higher than normal variation in dewpoint during the annealing procedure. This would define an

Fig. 3. Scaled tableau with SEM images of the Mn gradient thin film showing a) the thick region of the film (about 150 nm) and b) the thin region (about 25 nm).

148 C. Hambrock et al. / Surface & Coatings Technology 265 (2015) 145–153

experimental error of the set partial pressure of O2 allowing a fastergrain growth, accelerating the coalescence of neighboring crystallites.In any case, the observed changes at 900 °C need to be reevaluated be-fore concluding upon a distinct oxidation behavior specific to this an-nealing condition.

The images in Fig. 6b) reveal thepart of the thinfilmwith a thicknessof about 50 nm. The annealing at 950 °C showed the formation of a dis-continuous layer of largeMn oxide crystals. The highest partial pressureof O2 and the highest temperature favored a high surface atommobility.This led to the formation of large Mn oxide crystals. Less Mn atoms areavailable for the crystal growth at the thin end of the film and thereforethe Fe substrate material from underneath was visible. The lower theannealing temperature, the higher is the degree of surface coveragewith oxide crystals due to the formation of a higher amount of smalleroxide crystals. After annealing at 800 °C there is still some Fe substratematerial visible whereas after annealing at 750 °C a complete coverageof the substrate was achieved. At this lowest temperature, the crystalsizes were in the range of 50 nm. After annealing at 900 °C the surfaceappearance changed, as already observed for the thick end of both asdeposited and annealed films. Here, the slightly higher variation indew point during the annealing procedure led to the formation of a ho-mogeneous layer with individual crystal planes but without individual

Fig. 4. Scaled tableau with SEM images of the Si gradient thin film showing a) the

crystals. At the same time, the Fe substrate material was not visiblefrom underneath because the layer of Mn oxide was completely cover-ing the surface.

Annealing of the Mn thin film at different temperatures revealed astrong correlation between the surface morphology, the temperatureand the O2 partial pressure of the annealing atmosphere. The higherthe film thickness, the partial pressure and the temperature, thebetter-defined the crystal structures are. When lowering the tempera-ture at the same annealing conditions (concerning H2 andH2O content)the oxidation potential as well as the surface mobility decrease. This isone main reason for the less defined crystals appearing on the surfaceof the samples annealed at lower temperatures. Another influencingfactor (for the less defined crystals) is the microstructure of the sub-strate which will be discussed later. The crystal size in general is a mea-sure of the surface free energy. The larger the facets the lower is thesurface energy of the crystals. The size of the facets depends on thethickness of the deposited film, thewater content aswell as on the tem-perature applied during annealing. Comparing different thicknessesranging from 150 to 25 nm (see Fig. 5), the crystal size decreases dueto less Mn atoms being available for crystal growth. The lowest possiblesurface energy could therefore only be achieved by the formation of ahigher amount of nano-sized crystals. When comparing the structures

thick region of the film (about 100 nm) and b) the thin region (about 25 nm).

Fig. 5. Scaled tableau with SEM images of the Mn thin film after annealing at condition 1 at varying thicknesses.

149C. Hambrock et al. / Surface & Coatings Technology 265 (2015) 145–153

of the MnO crystals in Fig. 5 (annealed at condition 1) and Fig. 6(annealed at condition 2) it was evident that the crystal structure isstrongly influenced by the partial pressure of O2 (through the watercontent) of the annealing atmosphere as well as by the annealing tem-perature. A higher water content results in a higher O2 partial pressureleading to better defined crystal growth with larger facets at similartemperatures. Additionally, higher temperature promotes the forma-tion of larger facets because of the increased surface mobility of theatoms. Additionally, when comparing condition 1 and condition 2, theannealing time can also have an influence on the crystal growth. Thesamples annealed at condition 2 have been annealed slightly longer,which could also lead to better defined crystals in this case. However,the influence of the annealing time is much lower at these low partialpressures of O2 as compared to the effect of partial pressure and temper-ature [38].

In order to provide information on the crystal structure of the oxidelayer grown on the surface of the Fe substrate, GIXRD measurementswere performed on each deposited/annealed thin film. Comparativemeasurements at the thin and thick sides confirmed the same structuralcomposition but with differing peak intensities. Therefore, only theXRD-measurements on the thick side are shown in Fig. 7.

Fig. 6. Scaled tableau with SEM images of the Mn thin film after annealing at different t

The oxidizedMn crystals which formed on the surface of the Fe sub-strate were identified as MnO due to the GIXRD peak positions (shownin Fig. 7). Comparing the peak intensities of theGIXRDmeasurements atvarying temperatures reveals that the intensities of the MnO peaks areincreasingwith decreasing annealing temperature. This is in accordancewith the amount of crystal grown on the surface. At higher temperature,a higher surfacemobility enables the atoms to form less butmuch largeroxide crystals resulting in a lower peak intensity in the GIXRDmeasure-ment. Upon annealing at lower temperatures more crystallization nu-clei were formed as can be observed by the amount of surface crystalsvisible in Fig. 6. A slightly different behavior could be observed for thesample annealed at 900 °C. Here, almost only the (111) plane of theMnO crystals was detected. This correlated well to the slightly differentappearance of the surface as shown in Fig. 6a). It could, therefore, beconcluded that the crystal growth of the MnO is very sensitive to slightvariations of the protective annealing atmosphere, especially when an-nealing at lowdewpoints. The individual crystalswere not visible in theSEM image in Fig. 6, but only the crystal planes which are parallel to the(111) plane (as evidenced from theGIXRDmeasurements) were identi-fiable. At the same time, the amount of α-Fe was lower on this sample.However, the other samples still revealed a high amount of α-Fe from

emperatures in condition 2 at varying film thicknesses of a) 200 nm and b) 50 nm.

Fig. 7. GIXRD-results of the Mn thin film after annealing at condition 2 at the 200 nm thick end of each condition shown in Fig. 6.

150 C. Hambrock et al. / Surface & Coatings Technology 265 (2015) 145–153

the substrate material despite the glancing incidence condition of theGIXRD measurements. Starting from 800 °C, additional γ-Fe peakscould be detected. These peaks developed since the non-alloyed (andtherefore non-stabilized) Fe partly transformed to austenite upon inter-action with Mn, stabilizing the austenitic region [40,41]. At lower tem-peratures, the ferritic region is still stabilized because only a smallamount of Mn atoms is assumed to form the solid solution especiallyat these rather low annealing times. The change in microstructure ofthe substrate material depending on the annealing temperature canalso influence the crystal structure of the growing MnO. No preferencein grain orientation was visible, but when comparing the crystalsgrown on the thick side at 750 °Cwith the crystals grown at higher tem-peratures, a much less defined structure is visible. On the one hand, asalready mentioned before, these less-defined crystals can be attributedto the lower temperature and the lower oxidation potential. On theother hand, also the higher amount of ferritic substrate material maycontribute to the less defined MnO crystals. The lattice parameters forthe crystals of MnO are much more similar to austenite and not verywell matching the ferrite [42,43]. Therefore, the austenitic substratemi-crostructure can enhance the growth of well-defined oxide crystals.

The industrial grade materials, like TRIP and TWIP mentioned in theIntroduction section, have already been investigated in literatureconcerning several topics. The influence of the content of H2O and thepartial pressure of O2 in the annealing atmosphere was discussed, aswell as the influence of temperature and H2 content. All these variablesinfluence the selective oxidation processes on the surface of the steels,therefore it is difficult to perform a well-defined industrial process.However, the results reveal an enrichment of Mn close to the surfaceas well as the formation of MnO crystals at the surface. A comparabletendency of a crystal size decrease with a decrease in temperature anddew point was previously observed [13,28,29,44–46]. The observed de-pendencies on the model samples were therefore comparable to those

of the industrial grade materials. Additional studies are still necessaryto prove the significance of the results in order to transfer these to theindustrial process.

3.3. Si thin film after annealing

The Si thin filmwas annealed in two different atmospheres, identicalto those already described for the Mn thin film. Both conditions wereprotective for Fe but not for the Si thin film (see Fig. 2). Therefore, theSi thin film deposited on the Fe substrate was oxidized upon annealing.Fig. 8 shows the SEM images of the surface morphology of the Si thinfilm after annealing at condition 1 depending on the thin film thickness.Fig. 8a) reveals the surface appearance of the 100 nm thick region ofthe film, whereas Fig. 8d) displays the Si film with a thickness of25 nm. At higher thicknesses, the Si oxide formed a continuous andvery homogeneous layer on the surface of the substrate. Despite the ho-mogeneous oxide layer also round-shaped spots were visible. Thesespots appear rather bright which could be related to an elemental con-trast in the SEM image in comparison to the homogeneously coveringsurface layer. Only the elements Fe and Si were able to participate inthe reaction upon annealing. The elemental composition of the twopreviously mentioned areas (determined using EDX analysis) was di-rectly compared in spite of potentially high quantification errors. How-ever, assuming similar penetration depths, these measurements (notshown here) suggest a high concentration of Fe on the bright spotswhile the homogeneous surface layer revealed a high amount of Siand O upon direct comparison to each other. The occurrence of Fe atthe surface was attributed to a thermally activated diffusion of the sub-strate material and not to a self-diffusion of the Si atoms at the surface,thus exposing the substrate from underneath. The diffusion of Fe in Sifollows an interstitial mechanism which is enhanced at higher temper-atures [47,48]. On some parts of the film, the Fe is not completely

Fig. 8. Scaled tableau with SEM images of the Si thin film with varying thickness after annealing at condition 1.

151C. Hambrock et al. / Surface & Coatings Technology 265 (2015) 145–153

diffused because the energy barrier necessary to be overcome is higherwhere less surface defects or grain boundaries are available for enhanc-ing the diffusion phenomena. Therefore, the partly diffused Fe was vis-ible as less bright round spots in the SEM image. The diffusioncoefficients of Fe in Si and SiO2 indicate that the Fe diffusion is muchstronger in Si than in SiO2 [48,49]. Therefore, it is assumed that only asmall part of the deposited thin film was oxidized upon annealingunder these conditions and for these short annealing times.

Comparing Fig. 8b) (presenting a slightly lower film thickness) withFig. 8a), larger round spots could be observed. On the one hand, thedecrease in thickness leads to a stronger diffusion of Fe (revealed bythe size of the bright round spots). On the other hand, with a furtherdecrease in thickness (starting from about 50 nm, as shown in Fig. 8c)),the Fe is no longer visible in round spots but more in long-shapedareas. Especially Fig. 8d) indicates a large amount of long-shaped Feregions visible on the surface. Those long-shaped areas suggest thatless Fe diffusion and more Si self-diffusion took place during oxidation.With decreasing thickness and increasing temperature, the surface mo-bility of the Si becomes high enough for activating a dewetting phenom-enon. This dewetting phenomenon could only be observed for very thin

Fig. 9. Scaled tableau with SEM images of Si thin film after annealing at conditi

layers of oxides resulting in Fe substrate exposure. Additionally, at lowerthicknesses also the Fe grain boundaries are better visible (see Fig. 8d))as compared to higher thicknesses of the same thin film (see Fig. 8a)).Although these boundaries are not well visible in the SEM image of thethicker end of the film, they are expected to occur on this side of thefilm as well. These defects can (as already mentioned above) introducediffusion processes due to their lower activation energy as compared tothe surrounding regions.

The surface appearance of the Si thin film after annealing at condi-tion 2 is shown in Fig. 9. Depending on the change in temperature, var-ious effects could be observed. For all temperatures used during theexperiments, the Si thin film was oxidized forming a continuous oxidelayer. However, depending on the temperature and thickness, the oxi-dized Si thin film spalled upon oxidation. Oxidation of Si stronglydepends on the temperature and leads to an increase in the intrinsicstress of the film. This stress is described as compressive due to thestrong volume increase during Si oxidation, and it is additionallyincreased by a higher amount of H2 in the annealing atmosphere [50].At the highest temperature and the highest thickness of the film (seeFig. 9a) at 950 °C) the forming Si oxide film spalled compressively.

on 2 at a) the thick end (200 nm) and b) the thin end (75 nm) of the film.

152 C. Hambrock et al. / Surface & Coatings Technology 265 (2015) 145–153

This is according to the expectations because the spallation effect is gen-erally more severe at higher film thicknesses and higher annealingtemperatures. The higher the thickness, the higher is the stress thatcan build up. With a higher oxidation temperature the mobility anddiffusivity of Si and O atoms, respectively, are increased. At the sametime, with an increase in temperature the partial pressure of O2 also in-creases resulting in a higher oxidation potential at given atmosphericconditions (concerning H2 and H2O content). All three influencing fac-tors (the temperature, the partial pressure and the film thickness)cause an excessive compressive spalling upon oxidation at the highesttemperature and the highest thickness in these experiments. Whentwo parameters were kept constant and one was changed (in this casethe thickness, see Fig. 9b)), a different behavior could be observed.The stress formed at the highest temperature but lower thickness is ten-sile leading to the formation of cracks exposing the Fe substrate fromunderneath. This indicates that the thickness is highly influencingthe spallation behavior when themobility of the Si atoms and the sup-ply of O2 are kept constant. In order to keep other kinds of stresses, e.g.thermal expansion stress (upon cooling), as small as possible thecooling was performed much slower as compared to the heating(−5 °C s−1 and +20 °C s−1, respectively) [51]. Additionally, the Fegrain boundaries were visible at the thin end of the film after the hightemperature annealing at 950 °C (see Fig. 9b)). Those boundaries ofthe substrate material were uniformly covered by the thin film andtherefore remained observable.

For a temperature of 900 °C, similar phenomena could be observedfor the thick and also thin ends of the film. The thick side of the filmspalled compressively while the thin side of the film revealed cracksresulting from tensile stress. However, starting from a temperature of800 °C, the spallation behavior was different. On the thick end of thethin film, some cracks developed that could be correlated to a tensilestress. This indicates that not only the thickness but also the tempera-ture is a main influencing factor for the oxidation behavior of the Sithin film. Upon lowering the temperature at similar film thickness, theoxidation process cannot proceed as well-defined as before due to lessenergy available for surface diffusion and lower oxygen potential. Simi-larly, on the thin end of the Si film annealed at 800 °C the oxidation be-havior differed when compared to higher temperatures. For the lowfilm thickness, no spalling or crack formation could be observed afterannealing. When decreasing the temperature further to 750 °C, the ef-fects of temperature and oxidation potential were even less visible.Here, the thin film did not reveal any spallation neither for the thickend of the film nor for the thin end. After annealing at both tempera-tures, the grain boundaries of the Fe substrate material were no longervisible. This can be attributed on the one hand to the lower temperaturewhich influences the substrate material less severe and on the otherhand to the lower mobility of the surface atoms.

Despite the three mentioned factors influencing the oxidation be-havior of the Si thin film, another factor has to be differentiated. Bynow only the partial pressure of O2 was mentioned as an influencingfactor for the oxidation process. In Table 1, however, the correspondingwater content which was added to the atmosphere is given. This wasvery low for the annealing procedure described as condition 2. Previousinvestigations showed that this so-called wet oxidation is still fasterthan the completely dry oxidation. This can be related to the better ox-idant solubility in the case of the wet annealing atmosphere [52]. How-ever, usually these annealing procedures are performedmuch longer inorder to see these effects. Therefore it is assumed that this effect maynot cause significant deviations during the experiments performedhere [50].

Despite the higher annealing time, the Fe diffusion through thefilm could not be observed. This was explained by the fact that thewater content and therefore the partial pressure of O2 were muchlower during annealing in condition 2 as compared to condition 1.Only this higher water content combined with the low thickness ofthe film gave rise to the effect of Fe diffusion.

GIXRD measurements were performed also for the Si thin films.These, however, did only reveal the significant peaks of the α-Fe sub-strate material. No additional, characteristic diffraction peaks could bedetected as in the case of the Mn thin film. An interaction of the Sithin film with the Fe substrate material was expected to occur at theinterface similar to that observed for Mn (formation of austenite).Here, however, no austenite formed upon interaction of Fe with Sisince Si is a ferrite stabilizing agent. Therefore, the interaction layercould not be detected with the GIXRD because the diffraction peaksare overlapping with the peaks from the substrate material. Upon an-nealing under the applied conditions (see Fig. 2), the thermodynamiccalculations predict the formation of a stable Si oxide phase with themost stable being SiO2. As the signals for crystalline SiO2 were missingin the GIXRD spectra, it may be concluded that amorphous SiO2 formsupon annealing. This is well in accordance with previous studies,where a comparable oxidation behavior of the Fe–Mn to the Mn bulkmaterial is described while the Si bulkmaterial behaved strongly differ-ent as compared to the Fe–Si diluted alloy [53].

4. Conclusions

The analysis of Mn and Si elemental thin films deposited on Fe sub-strates revealed different effects upon annealing in varying conditions.The Mn thin film formed oxide crystals (identified via XRD as MnO).The structure and especially the size of the crystals were stronglydependent on the temperature, the film thickness and the O2 partialpressure of the annealing atmosphere. The higher the partial pressure,the temperature and the film thickness, the better defined were theobserved structures. A diffusion of Fe could not be clearly identifiedbecause of the strong crystal growth upon oxidation.

The Si thin film formed a continuous oxide layer upon annealingunder both conditions used. Annealing at condition 1 gave rise to thephenomenon of Fe diffusion through the partly oxidized Si thin filmwhich was dependent on the temperature and the O2 partial pressureas well as on the thickness of the film. The lower the thickness, thehigher was the influence of the self-diffusion of Si oxide on the surfacewhich resulted in a creeping of the film. At the second annealing condi-tion, the partial pressure of O2 of the annealing atmosphere was muchlower and therefore other phenomena occurred. High temperatures,high thicknesses and high partial pressure of oxygen led to spallationof the film by compressive stress because the volume increased uponoxidation. Annealing at lower temperatures and lower thicknesses ledto the formation of a continuous and homogeneous Si oxide layer. At in-termediate ranges of temperature and thickness, the film revealedcracks due to tensile stresses formed due to creeping of the Si atomsin order to provide a compact oxide layer. A diffusion of Fe throughthe amorphous oxide layer could not be observed because the oxidationpotential was probably too low to induce these diffusion phenomena.

Some of the phenomena occurring on the surfaces of themodel thinfilm samples upon annealing could in principle be correlated to the pro-cesses occurring at the surface of the industrial grade AHSS or electricalsteel materials. Despite the fact that these industrial gradematerials arediluted alloys and that they usually behave different in terms of oxida-tion, comparable phenomena were observed. Especially for the Mnthin film as compared to the Mn diluted alloys, the formation of MnOcrystals could be observed. The dependencies of both materials on thetemperature and the partial pressure of O2 were similar, forming largercrystals at higher temperatures and higher partial pressures. For the Sithin film a lower comparability to the industrial grade material couldbe observed. The diffusion and spallation processes occurring on thethin film were not directly comparable to the processes occurring onthe industrial gradematerial. Therefore, new approaches e.g. depositionof combinatorial libraries of Fe–Si have to be chosen in order to obtain ahigher comparability between the industrial grades and the model thinfilm samples.

153C. Hambrock et al. / Surface & Coatings Technology 265 (2015) 145–153

Generally, it is possible to determine influencing factors and depen-dencies with the help of thin filmmodel samples. These model sampleshave the great advantage of being easily produced by thermal evapora-tion without any restrictions on the alloying content. Therefore, theapproach of providing model thin film samples for the determination ofinfluencing factors or dependencies is quite promising. In future investi-gations, additional annealing conditions combined with the preparationof binary and ternary combinatorial libraries have to be followed inorder to obtain a deeper insight into the oxidation and diffusion process-es and the corresponding dependencies.

Acknowledgments

The authors gratefully acknowledge Abdellatif Jerrar and JohannaPöhmer (voestalpine) for their help with the XRD and SEM measure-ments. Thefinancial support by theAustrian FederalMinistry of Economy,Family and Youth and the National Foundation for Research, TechnologyandDevelopment through the ChristianDoppler Laboratory for Combina-torial Oxide Chemistry (COMBOX - CDLA 13170001) is gratefullyacknowledged.

References

[1] K.M. Lee, S.Y. Park, M.Y. Huh, J.S. Kim, O. Engler, J. Magn. Magn. Mater. 354 (2014)324–332.

[2] M.F. Littmann, IEEE Trans. Magn. 7 (1971) 48–60.[3] M.A. da Cunha, S. da Costa Paolinelli, J. Magn. Magn. Mater. 320 (20) (2008)

2485–2489.[4] Y. Oda, M. Kohno, A. Honda, J. Magn. Magn. Mater. 320 (20) (2008) 2430–2435.[5] D. Steiner Petrovic, Mater. Technol. 44 (6) (2010) 317–325.[6] H. Haiji, K. Okada, T. Hiratani, M. Abe, M. Ninomiya, J. Magn. Magn. Mater. 160

(1996) 109–114.[7] T. Ros-Yanez, Y. Houbaert, O. Fischer, J. Schneider, J. Mater. Process. Technol. 141

(2003) 132–137.[8] J.R. Davis, Alloying: Understanding the Basics, First edition ASM International,

United States of America, 2001. 134 ff.[9] K.H. Kwon, B.C. Suh, S.I. Baik, Y.W. Kim, J.K. Choi, N.J. Kim, Sci. Technol. Adv. Mater.

14 (2013) 014204.[10] H. Aydin, E. Essadiqi, I.-H. Jung, S. Yue, Mater. Sci. Eng. A 564 (2013) 501–508.[11] O. Grässel, L. Krüger, G. Frommeyer, L.W. Meyer, Int. J. Plast. 16 (2000) 1391–1409.[12] J. Nakano, Sci. Technol. Adv. Mater. 14 (2013) 014207.[13] H. Liu, Y. He, S. Swaminathan, M. Rohwerder, L. Li, Surf. Coat. Technol. 206 (2011)

1237–1243.[14] Kutz, Handbook of Materials Selection, First edition John Wiley & Sons, New York,

2002.[15] O. Aponbiede, U. Shehu, J. Emerg. Trends Eng. Appl. Sci. 1 (2) (2010) 174–177.[16] D.R. Gaskell, 5th ed.CRC Press, Inc., Florida, 2008, pp. 364–365.[17] C. Hambrock, K. Vincze-Minya, S. Hild, A.W. Hassel, Phys. Status Solidi A 211 (2014)

1429–1438.

[18] S.C. Paolinelli, M.A. da Cunha, J. Magn. Magn. Mater. 304 (2006) e599–e601.[19] I. Parezanovic, M. Spiegel, Surf. Eng. 20 (2004) 285–291.[20] X. Vanden Eynde, J.P. Servais,M. Lamberigts, Surf. Interface Anal. 33 (2002) 322–329.[21] K. Yanagihara, S. Suzuki, S. Yamazaki, Oxid. Met. 57 (2002) 281–296.[22] L. Cho, G.S. Jung, B.C. De Cooman, Metall. Mater. Trans. A 45 (2014) 5158–5172.[23] X.S. Li, S.-I. Baek, C.-S. Oh, S.-J. Kim, Y.-W. Kim, Scr. Mater. 57 (2007) 113–116.[24] S. Swaminathan, M. Rohwerder, Defect Diffus. Forum 309–310 (2011) 203–208.[25] V. de Freitas Cunha Lins, L. Madeira, J.M. Carneiro Vilela, et al., Appl. Surf. Sci. 257

(2011) 5871–5878.[26] J. Rehrl, K. Mraczek, A. Pichler, E. Werner, Mater. Sci. Eng. A 590 (2014) 360–367.[27] R. Sagl, A. Jarosik, D. Stifter, G. Angeli, Corros. Sci. 70 (2013) 268–275.[28] L. Cho, M.S. Kim, Y.H. Kim, B.C. De Cooman, Metall. Mater. Trans. A 44 (2014)

5081–5095.[29] E.M. Bellhouse, J.R. McDermid, Metall. Mater. Trans. A 42 (2011) 2753–2768.[30] T. Gebhardt, D. Music, M. Ekholm, et al., Acta Mater. 59 (2011) 1493–1501.[31] G. Tian, X. Bi, J. Alloys Compd. 502 (2010) 1–4.[32] J.-M. Ting, S.-W. Hung, Diam. Relat. Mater. 15 (2006) 1834–1838.[33] A.I. Mardare, A.P. Yadav, A.D. Wieck, M. Stratmann, A.W. Hassel, Sci. Technol. Adv.

Mater. 9 (2008) 035009.[34] S.O. Klemm, A.G. Martin, L. Lengsfeld, J.C. Schauer, B. Schumacher, A.W. Hassel, Phys.

Status Solidi A 207 (2010) 801–806.[35] M. Voith, A.I. Mardare, A.W. Hassel, Phys. Status Solidi A 210 (2013) 1000–10005.[36] S. Ianerjee, S. Banerjee, A.K. Yagi, Functional Materials: Preparation, Processing and

Applications, Elsevier, London (England), 2012. 335.[37] K. Tanaka, E. Maruyama, T. Shimada, H. Okamoto, Amorphous Silicon, John Wiley &

Sons Ltd, West Sussex (England), 1999.[38] K.R. Lawless, Rep. Prog. Phys. 37 (1974) 231–316.[39] F.A. Jenkins, H.E. White, Fundamentals of Optics, McGraw-Hill, New York (USA),

1981.[40] M. Durand-Charre, Microstructure of Steels and Cast Irons, Springer-Verlag, Berlin

(Germany), 2004. 265–266.[41] E. De Moor, D.K. Matlock, J.G. Speer, M.J. Merwin, Scr. Mater. 64 (2011) 185–188.[42] M. Onink, C.M. Brankman, F.D. Tichelaar, E.J. Mittemeijer, S. van der Zwaag, J.H. Root,

N.B. Konyer, Scr. Metall. Mater. 29 (1993) 1011–1016.[43] O. Madelung, U. Rössler, M. Schulz, MnO. Phase diagram, crystal structure, lattice

parameters (part of Landolt–Börnstein — Database, Group III Condensed Matter,Chapter-DOI: 10.1007/10681735_449).

[44] M. Blumenau, A. Barnoush, I. Thomas, H. Hofmann, H. Vehoff, Surf. Coat. Technol.206 (2011) 542–552.

[45] M. Norden, M. Blumenau, T. Wuttke, K.-J. Peters, Appl. Surf. Sci. 271 (2013) 19–31.[46] H. Liu, Y. He, L. Li, Appl. Surf. Sci. 256 (2009) 1399–1403.[47] W.D. Callister Jr., Materials Science and Engineering: An Introduction, 7th ed. John

Wiley & Sons, Inc., New York (US), 2007. 112.[48] J. Fisher, Diffusion in Silicon— 10 Years of Research, Scitec Publications, First edition,

1998. p. 389ff.[49] H.G. Francois-Saint-Cyr, F.A. Stevie, J.M. McKinley, K. Elshot, L. Chow, K.A.

Richardson, J. Appl. Phys. 94 (12) (2003) 7433–7439.[50] A. Szekeres, P. Danesh, J. Non-Cryst. Solids 187 (1995) 45–48.[51] E. Kobeda, E.A. Irene, J. Vac. Sci. Technol. B 7 (1989) 163.[52] C. Hollauer, Modeling of Thermal Oxidation and Stress EffectsPhD Dissertation TU

Wien, 01/2007.[53] A.S. Khanna, Introduction to High Temperature Oxidation and Corrosion, 1st ed.

ASM International, Materials Park, 2002. 92–128.