Observation of the Magnetization Reorientation in Self-Assembled Metallic Fe-Silicide Nanowires at Room Temperature by Spin-Polarized Scanning Tunneling Spectromicroscopy

Round  1

Reviewer 1 Report

Paper can be accepted after the following corrections:

1.      Magnetization reorientation should be analyzed quantitatively. Later, the results of these analyses may be connected with measured macroscopic characteristics (e.g. current-bias)

2.      Abbreviations should be used in text and figures to increase overall quality of presentation (e.g. Current I (mA), etc.).

3.      Figure 5 is not clear. Please split it into separate figures, increase quality and provide suitable explanation.

4.      Conclusion should be stated in more quantitative way accordingly to the title, abstract and introduction.

Author Response

1. Magnetization reorientation should be analyzed quantitatively. Later, the results of these analyses may be connected with measured macroscopic characteristics (e.g. current-bias).

Þ (1) In 147^th line of our original manuscript, we have explained: “This magnetization modulation is only observed in rectangular-sectional nanowire with widths and heights larger than 36 nm and 5 nm. As shown in the SP-dI/dV profile in Figure 3d, these repeatedly arranged streak domains alternate with a magnetic period of 5.0±1.0 nm; the white and black streak domains have widths of 2.0±1.0 nm and 3.0±1.0 nm, respectively.”. To make the explanation of Figure 3d clearer, we added a mark in Figure 3d to indicate a magnetic period of ~5 nm.

As Referee’s suggestion, to make the analysis of magnetization reorientation more quantitative, we have tried to fit the measured line profile of dI/dV intensities using a standard wall profile of the domain wall, sin(ϕ(x)) = tanh((x − x₀)/(w/2)) [37], with a wall width of w ≈ 2 nm, where ϕ is the angle between the magnetization direction and the easy z axis, x is the position of the measured dI/dV signal and x₀ is the position of the domain wall center. The measured line profile of dI/dV intensities across a domain wall can be fitted by a 180^o Néel wall using a standard wall profile, as indicated by the arrows (i.e., the spin orientations) in Figure 6d. Figure 6 is separated from the original Figure 5, as Referee’s suggestion in Comment #3. To make the explanation of Figure 6d clearer, we added a mark in Figure 6d to indicate a wall width of 2.0±1.0 nm.

Based on the above explanations, we revised our original statements in 179^th line of our original manuscript: “The in-plane domains with a width of ~5.0 nm can be interpreted as domain walls. This interpretation is based on the comparison of the experimental SP-dI/dV profile in Figure 3d with the model calculation of magnetizations for Néel domain walls [31]. Therefore, the domain walls in the rectangular-sectional NW B are identified as Néel walls, as schematically represented in Figure 5d.” by the new statements: “Figure 6d shows an averaged line profile of dI/dV intensities, which was taken across a domain wall. The in-plane domains with a width of 2.0±1.0 nm can be interpreted as domain walls. On the basis of the micromagnetic model [37,38], we have tried to fit the measured line profile of dI/dV intensities using a standard wall profile of the domain wall, sin(ϕ(x)) = tanh((x − x₀)/(w/2)), with a wall width of w ≈ 2 nm (see the red line in Figure 6d), where ϕ is the angle between the magnetization direction and the easy z axis, x is the position of the measured dI/dV signal and x₀ is the position of the domain wall center. The measured line profile of dI/dV intensities across a domain wall can be fitted by a 180^o Néel wall, as indicated by the arrows (i.e., the spin orientations) in Figure 6d. Therefore, the domain walls of the magnetization reorientation in the rectangular-sectional NW B are identified as Néel walls.”.

(2) We thank Referee’s suggestion, our work needs a macroscopic characteristics (e.g. current-bias). However, the in-situ measurement of charge transport requires a four-point nanoscale probe [Nano Lett. 12, 938–942 (2012); Nano Lett. 13, 3684–3689 (2013)]. We haven’t such an expensive instrument. It is difficult to provide an experimental support for the in-situ charge transport in this work. Moreover, ex-situ macroscopic characteristics of transport properties of these parallel Fe-silicide nanowires needs to fabricate two-terminal electrical contacts of Fe-silicide nanowires using a photolithography followed by HF etching and evaporation of Ti/Au metal films [Rev. Mod. Phys. 71, 687 (1999)]. But we also haven’t such techniques and instruments. Nevertheless, in our present results, each representative I-V curve of an individual Fe-silicide nanowire is spatially averaged over 20 different positions taken along a single nanowire to confirm that every self-assembled Fe-silicide nanowire is metallic, as shown in Figure 4a.

2. Abbreviations should be used in text and figures to increase overall quality of presentation (e.g. Current I (mA), etc.).

Þ We thank Referee’s suggestion. In order to explain the meanings of abbreviations in text (e.g. V_b and I_t), we added statements in the experimental section: “Each I–V curve was obtained by recording the tunneling current (I_t) while ramping the bias voltage (V_b) from −2.0 to 2.0 V with a voltage step of 0.05 V. The I−V curve was then numerically differentiated to derive a dI/dV curve (i.e., a STS spectrum).”.

3. Figure 5 is not clear. Please split it into separate figures, increase quality and provide suitable explanation.

Þ We thank Referee’s suggestion. Figure 5 was split into two separated Figures 5 and 6, i.e., the schematic sketches of the magnetization configurations of the triangular-section nanowire E and the rectangular-sectional nanowire B, respectively. Moreover, we also added the SP-dI/dV maps and the cross–sectional profiles of both the triangular-sectional nanowire E and the rectangular-sectional nanowire B in Figures 5 and 6, respectively, for the comparison of SP-dI/dV maps and the magnetization configurations of both nanowires E and B.

4. Conclusion should be stated in more quantitative way accordingly to the title, abstract and introduction.
Þ We thank Referee’s suggestion. As described in the reply to Comment #1, we revised the statements in the section of Conclusions: “Based on the noncollinear magnetism in a bulk Ni tip with a in-plane sensitivity and the micromagnetic mode of domain walls, these streak domains show alternating in-plane magnetization and out-of-magnetization with a periodicity of 5.0±1.0 nm and a domain wall of 2.0±1.0 nm, which can be fitted by 180° Néel walls using a standard wall profile, sin(ϕ(x)) = tanh((x−x₀)/(w/2)). Therefore, the domain walls of the magnetization reorientation in the rectangular-sectional NW B are identified as Néel walls.”.

Author Response File: Author Response.doc

Reviewer 2 Report

Review of “Observation of the magnetization reorientation inself-assembled metallic Fe-silicide nanowires at room temperature by spin-polarized scanningtunneling spectromicroscopy” by Hong and Liu

This is an interesting paper, certainly on a worthy and timely topic.  However, in its present form, it is highly speculative.

There is no structural or compositional analysis of the silicide islands, whatsoever.  Therefore the claim that the islands are Fe₅Si₃ is not corroborated by anything.  Metallicity of the STS spectra or reference to other reports where similar amounts of Fe deposited at a similar temperature resulted in the formation of Fe₅Si₃ are not sufficient arguments.  If not structure and composition determination by TEM and one of the electron spectroscopies (e.g. photoemission), respectively, the authors could at least use their STM for structural analysis.  Without it, the islands may as well have the magnetic Fe₃Si crystal structure, or even one of the non-magnetic disilicide variants.  There have been numerous reports on magnetic properties of disilicide nanocrystals (cf. He at al. JACS 137, 11419 (2015), or Goldfarb et al. Adv. Mater. 30, 1800004 (2018)).

Similarly, the authors hypothesis of a flat interface, has to be better substantiated.  While this may be the case, there are also plenty reports on faceted interfaces of silicide islands with silicon substrates, including those of Ti, Er, Co, and Fe (see, for example, Surf. Sci. 457, 147 (2000), Physica E 43, 176 (2010), Surf. Sci. 606, 1649 (2012), PRB 96, 045415 (2017)).  X-sectional TEM analysis could shed light on this issue.

The magnetic SP-STM contrast is very interesting, but its interpretation raises some questions, as well.  Have the authors modulated the tunneling bias for their spectroscopic SP-STM acquisition?  It is not mentioned in the experimental description.  The effect of stray fields from a ferromagnetic Ni tip on the contrast is not discussed.  Determination of the magnetization directions of the island domains and tip polarization is not obvious.  Polarization of the tip along its length could rather be expected.  Assignment of the Neel wall characteristics to the white stripes on the island top is not clear either.  The description of out-of-pane rotation in the z-y plane given in the text, is more consistent with Bloch wall.  Finally, no explanation is given on the difference between magnetic structure of the islands with rectangular section and those with triangular section.

Author Response

1. There is no structural or compositional analysis of the silicide islands, whatsoever.   Therefore the claim that the islands are Fe5Si3 is not corroborated by anything. Metallicity of the STS spectra or reference to other reports where similar amounts of Fe deposited at a similar temperature resulted in the formation of Fe5Si3 are not sufficient arguments. If not structure and composition determination by TEM and one of the electron spectroscopies (e.g. photoemission), respectively, the authors could at least use their STM for structural analysis. Without it, the islands may as well have the magnetic Fe3Si crystal structure, or even one of the non-magnetic disilicide variants. There have been numerous reports on magnetic properties of disilicide nanocrystals (cf. He at al. JACS 137, 11419 (2015), or Goldfarb et al. Adv. Mater. 30, 1800004 (2018)).

Þ We agree Referee’s comment. It is necessary to use the structure and composition determination of TEM and electron spectroscopies to confirm the formation of Fe₅Si₃. However, we haven’t these instruments, we need to collaborate with other laboratories. It will take a long time to complete this experiment. Additionally, because the STM image corresponds to the special distribution of electronic densities of surface atoms, it is difficult to directly determine the atomic structure by using STM.

Nevertheless, the metallic Fe₃Si is only formed below ~500°C with similar amounts of deposited Fe like our work, as reported in the literatures [J. Appl. Phys. 105, 07B102 (2009); J. Appl. Phys. 104, 093707 (2008); Appl. Phys. Lett. 93, 132117 (2008)], and the disilicide is semiconducting [JACS 137, 11419 (2015)]. Thus, the metallic Fe-silicide nanowires formed at ~700°C in our work are likely to be Fe₅Si₃ according to the equilibrium phase diagram of Fe–Si in Refs. [10,11] and the metallic character shown in I-V curves of these parallel nanowires (Figure 4).

To make this point clear, we revised the statements in 111^th line of our original manuscript as the following: “RT magnetism of Fe-silicide NWs has been reported in previous studies on the magnetic properties of Fe₃Si [10,11], Fe₅Si₃ [12,13] and b-FeSi₂ [14] NWs. According to our growth conditions for Fe-silicide NWs, these NWs are likely to be iron-rich Fe₅Si₃ because the equilibrium phase diagram of Fe–Si shows that Fe₅Si₃ is formed at 500‒700°C [12,13], whereas Fe₃Si is formed below 500°C [10,11]. This result can be further confirmed by the metallic character shown in I-V curves of these parallel NWs (see Figure 4), which is consistent with that of Fe₅Si₃ NWs and rules out the possibility of semiconducting b-FeSi₂ NWs [14].”.

2. Similarly, the authors hypothesis of a flat interface, has to be better substantiated. While this may be the case, there are also plenty reports on faceted interfaces of silicide islands with silicon substrates, including those of Ti, Er, Co, and Fe (see, for example, Surf. Sci. 457, 147 (2000), Physica E 43, 176 (2010), Surf. Sci. 606, 1649 (2012), PRB 96, 045415 (2017)). X-sectional TEM analysis could shed light on this issue.

Þ We are very sorry for this mistaken presentation in the 91^th line of our original manuscript. In order to avoid readers’ misunderstanding, we replaced “a flat interface” by “a well-defined interface” in the text.

As shown in X-sectional TEM analysis in Figures 5‒8 of Referee’s suggested literature [Surf. Sci. 606, 1649 (2012)], all self-assembled silicide islands were observed to grow into the Si(001) surface with sufficiently well-defined interfaces. Also, the formation of a well-defined interface between the self-assembled silicide nanostructures and various Si substrates via an “endotaxial” growth mechanism has been demonstrated in Refs. [13,14] and the literature [Surf. Sci. 606, 1649 (2012); CrystEngComm, 20, 2916‒2922 (2018)].

In 88^th line of our original manuscript, we have addressed: “the formation of trenches indicates that self-assembled Fe-silicide nanowires grow into the Si(110) substrate via an endotaxial growth mechanism, as manifestly shown in the SP-STM image (Figure 3a) and its cross-sectional profile (Figure 3c). Therefore, there should be a well-defined interface formed between self-assembled Fe-silicide nanowires and the Si(110) substrate.”.  

3. The magnetic SP-STM contrast is very interesting, but its interpretation raises some questions, as well. Have the authors modulated the tunneling bias for their spectroscopic SP-STM acquisition? It is not mentioned in the experimental description.

Þ We thank Referee’s suggestion. As we have addressed in the experimental section: “SP-STM/STS measurements were acquired in the current imaging tunneling spectroscopy (CITS) mode at RT.”. In the CITS mode, the dI/dV curves (i.e., SP-STS spectra) were obtained by numerically differentiation of the I(V) curves, as reported by the literature [Microsc. Res. Tech. 66, 93–104 (2005)]. It is different from the lock-in technique with a modulation voltage of ~20 mV at 2k Hz added to the tunneling bias for their spectroscopic SP-STM acquisition (see the original Refs. [28‒31]).

   Thus, we added statements in the experimental section: “Each I–V curve was obtained by recording the tunneling current (I_t) while ramping the bias voltage (V_b) from −2.0 to 2.0 V with a voltage step of 0.05 V. The I−V curve was then numerically differentiated to derive a dI/dV curve (i.e., a STS spectrum).”, and also added a literature of the CITS measurement [Microsc. Res. Tech. 66, 93–104, (2005)].

4. The effect of stray fields from a ferromagnetic Ni tip on the contrast is not discussed.

Þ Because Ni is a soft magnetic metal with a smaller magnetic moment (0.6 m_B), Ni produces a weaker stray field. The effect of stray fields from a Ni tip on the magnetic contrast of samples is weak, this makes Ni tips reliable as SP-STM tips, as reported in the literatures [Rev. Sci. Instrum. 90, 013704 (2019); J. Vac. Sci. Technol. B 32, 061801 (2014)].

5. Determination of the magnetization directions of the island domains and tip polarization is not obvious. Polarization of the tip along its length could rather be expected.

Þ The direct determination of the magnetization orientation of the tip requires utilizing an external magnetic field to in-situ detect the out-of-plane and in-plane magnetization of the samples via the measurement of the dI/dV hysteresis loop curves [Nano Convergence 4, 8 (2017); Jpn. J. Appl. Phys. 51, 0302089 (2012); Science 327, 843 (2010)]. However, we have no external magnet in our STM system. Although we can’t experimentally show the magnetization orientation of the tip, we utilized Ref. [27] and literatures [Appl. Phys. Lett., 84, 9 (2004); Phys. Rev. B 72, 214409 (2005); Surf. Sci. 600, 1586–1591 (2006)] to support that the bulk Ni tip exhibits the in-plane magnetization. In the following, we give our explanations.

(1) Because the magnetization directions of elongated island domains with the inclined surfaces typically exhibit the in-plane orientation [Appl. Phys. Lett., 84, 9 (2004); Phys. Rev. B 72, 214409 (2005); Surf. Sci. 600, 1586–1591 (2006)], the magnetization directions of the nanowires with the triangular section should also exhibit the in-plane orientation. Thus, we have addressed in 157^th line of our original manuscript: “there are two distinct domains with opposite in-plane magnetizations oriented along the individual triangular-sectional NW, as usually observed in the domain structures of Fe ribbons [26].”.

(2) Because the spin-dependent tunneling current in SP-STM depends on the relative orientation between the magnetizations of the measured region and the tip, another way is to use the well-known magnetization directions of the samples to confirm the spin polarization of the tip, as reported in the original Ref. [27]. Because the in-plane magnetization of the triangular-sectional nanowires exhibits the white contrast, this observation reveals that the magnetization of the triangular-sectional nanowires is parallel to the tip magnetization. Consequently, the bulk Ni tip in our SP-STM also exhibits the in-plane spin polarization, similar to the report of the bulk Cr tip with a in-plane magnetic sensitivity (i.e. the original Ref. [27] ).

   To make this point clear, we added statements in 162^th line of our original manuscript: “Because the tunneling current depends on the relative orientation between the magnetizations of the measured region and Ni tip, if the magnetization of the measured region is parallel to the tip magnetization, the dI/dV map is bright. While the magnetizations of the measured region and Ni tip are anti-parallel, the dI/dV map becomes darker.”.

6. Assignment of the Neel wall characteristics to the white stripes on the island top is not clear either. The description of out-of-pane rotation in the z-y plane given in the text, is more consistent with Bloch wall.

Þ The following two figures show the magnetization reorientation in the different types of domain walls (from https://iop.fnwi.uva.nl/cmp//qem/research_projects/domainwall.html). The type of domain wall in Figure (a) is a Néel wall, Figure (b) shows a Bloch wall. Figure (c) shows a 3-D view of a Néel wall and a Bloch wall. According to our discussions in Figure 6, the type of domain wall on the top of the wider rectangular-sectional nanowire B in Figure 3b should be a Néel wall.

(a)      Néel wall

(b)     Bloch wall

(c)    

7. Finally, no explanation is given on the difference between magnetic structure of the islands with rectangular section and those with triangular section.

Þ In 150^th line of our original manuscript, we have explained the origin of the magnetization modulation in the repeatedly arranged streak domains. This magnetization modulation originates from the competition between the magnetocrystalline polarization along the easy axis and the shape anisotropy energy along the wire axis [25]. In the triangular-sectional nanowire E, the shape anisotropy is stronger than the magnetocrystalline anisotropy, which tends to align the magnetization along the nanowire, as seen in the in-plane magnetization of the nanowire E in Figure 5c.

Author Response File: Author Response.doc

Reviewer 3 Report

The authors presented the synthesis of Fe-silicide self-assembled nanowires and observation of the quasi-periodic magnetic domains by STM and STS experiments.

The experimental design and characterization contain valuable interest and results to spintronic devices applications.

I recommend this manuscript can be published in the Coatings after addressing several remarks in the below:

The authors presented that the IV characteristics in Fig. 4(a) represent metallic behavior. However, the IV curve is not linear but non-linear behavior, indicating not omhic ones. How can the authors address this non-linear behavior of IV characteristics by metallic ones?

In the legend of Fig. 4(b), please identify the measured points from Fig. 3(a) not marked by colors of domains.

In Fig. 5, the authors depicted the spin directions according to the STM-STS measurements. On the other hand, how can the authors distinguish or characterize the spin direction from the measurements? Please describe in detail for that.

There is a typo in 88th line : Figure 3c -> Figure 2c.

Author Response

1. The authors presented that the IV characteristics in Fig. 4(a) represent metallic behavior. However, the IV curve is not linear but non-linear behavior, indicating not omhic ones. How can the authors address this non-linear behavior of IV characteristics by metallic ones?

Þ (1) As we have addressed in 124^th line, all nanowires exhibit metallic character, whereas the adjacent substrate shows semiconducting behavior with a band gap of ~0.9 eV. This result reveals that a ferromagnetic Schottky junction is formed, i.e., no omhic junction. Consequently, the IV curve obtained in STM shows a non-linear behavior.

(2) Additionally, the tunneling current I between the sample and tip of the STM junction can be calculated on the basis of Fermi’s golden rule:

.

Here f is the Fermi-Dirac distribution function, V is the bias voltage applied between the electrodes, M denotes the tunneling matrix element and ρ_T,S is the density of states of the tip and sample.

   Therefore, the tunneling current I of a STM junction depends on the product of the densities of states of the tip and sample. It is different from the macroscopic IV characteristics, originated from the ohm law: V = IR. The non-linear behavior of IV characteristics can be attributed to the contribution from the product of the densities of states of the tip and sample at different energy positions (i.e. at different bias voltage V_b), as shown in the following I-V and dI/dV curves of a Ag(111) metal surface.

2. In the legend of Fig. 4(b), please identify the measured points from Fig. 3(a) not marked by colors of domains.

Þ We thank Referee’s suggestion. Because each representative I-V curve was averaged over 20 individual spectra measured along the corresponding nanowire, we added the SP-STS map (i.e. Figure 3b) in Figure 4b and indicated the measured regions by different marks.

3. In Fig. 5, the authors depicted the spin directions according to the STM-STS measurements. On the other hand, how can the authors distinguish or characterize the spin direction from the measurements? Please describe in detail for that.

Þ (1) As explained in the reply to Comment #5 of 2^nd Referee, in SP-STS image the spin-dependent tunneling current depends on the relative orientation between the magnetizations of the measured region and Ni tip, if the magnetization of the measured region is parallel to the tip magnetization, the dI/dV map is bright. While the magnetizations of the measured region and Ni tip are anti-parallel, the dI/dV map becomes darker.

Because the magnetization direction of elongated island domains with the inclined surfaces typically exhibits the in-plane orientation [Appl. Phys. Lett., 84, 9 (2004); Phys. Rev. B 72, 214409 (2005); Surf. Sci. 600, 1586–1591 (2006)], the magnetization direction of the nanowires with the triangular section should also exhibit the in-plane orientation. Thus, we have addressed in 157^th line of our original manuscript: “the two distinct domains with the white and red contrasts exhibit opposite in-plane magnetizations oriented along the individual triangular-sectional NW, as usually observed in the domain structures of Fe ribbons [26].”.

Because the white domain in the right half of the triangular-sectional nanowires is magnetized in-plane along the [] direction, the white streak domains in the rectangular-sectional NW B thus exhibit an in-plane magnetization parallel to [] as well, described in 169^th line of our original manuscript.

(2) Also, we have addressed in 172^th line of our original manuscript: “As reported in priori studies [28‒30], the intra-atomic noncollinear magnetism in magnetic tips leads to an enhanced (reduced) local density of electronic states (LDOS) over in-plane (out-of-plane) magnetized regions when imaged with an in-plane sensitive tip. The dI/dV intensities of these parallel streak domains on the rectangular NW B also result from the noncollinear magnetism, thereby revealing a contrast modulation with periodically alternating in-plane and out-of-plane magnetic domains in the rectangular NW B. Therefore, the out-of-plane domains appear as a darker area in the dI/dV map of NW B acquired at a sample bias of +0.50 V.”. Our proposed model for the magnetization reorientation in Figure 6c can be supported by the fitting of the measured line profile of dI/dV intensities by 180° Néel walls, shown in Figure 6d.

The explanations of Figures 5 and 6 are in agreement with the previous studies for the magnetic domain walls of Fe films on vicinal W(110) as imaged with a Gd-coated tip providing sensitivity to the out-of-plane magnetization [Appl. Phys. A 78, 781–785 (2004); J. Magn. Magn. Mater. 305, 279–283, (2006); and the original Refs. [24,25] ], in which the monolayer (ML) Fe stripes appear black and exhibit the in-plane magnetization, the double-layer (DL) Fe stripes show the two different contrasts (white and grey) and are magnetized parallel or antiparallel to the tip magnetization, respectively. That is, the magnetization direction of the ML Fe stripes with a black contrast is perpendicular to that of the DL Fe stripes with the white/gray contrast. Accordingly, in Figure 3b of our manuscript, the magnetization direction of the black streak domains in NW B is perpendicular to that of other white/red magnetic domains in NWs B and E, as described in the explanations of Figures 5 and 6.

However, the black streak domain exhibiting a perpendicular magnetization may be oriented either up or down, two cases that cannot be distinguished with a tip exhibiting in-plane sensitivity, as reported in the literature [Phys. Rev. B 67, 020401(R), (2003)].

To make this point clear, we added the statements in 171^th line of our original manuscript as the following: “The black streak domains in NW B may represent an out-of-plane magnetization vertical to the NW surface, in agreement with the previous studies for the magnetic domain walls of Fe films on vicinal W(110) (i.e., the magnetization direction of the monolayer Fe stripes with a black contrast is perpendicular to that of the double-layer Fe stripes with the white/gray contrast).”.

4. There is a typo in 90th line : Figure 3c -> Figure 2c.

Þ Referee could misunderstand. It is not a typo, the cross-sectional profile of Figure 3c is used to show the formation of trenches adjacent to Fe-silicide NWs.

Author Response File: Author Response.doc

Round  2

Reviewer 1 Report

Paper was corrected in line with review report and can be accepted in the present state.

Author Response

Thank you for your positive comments. As suggested by Referee, our manuscript has undergone English language editing by MDPI. Please see the English-editing-certificate.

Author Response File: Author Response.pdf

Reviewer 2 Report

Response to 2^ndReferee’s comments:

1.There is no structural or compositional analysis of the silicide islands, whatsoever.   Therefore the claim that the islands are Fe5Si3 is not corroborated by anything. Metallicity of the STS spectra or reference to other reports where similar amounts of Fe deposited at a similar temperature resulted in the formation of Fe5Si3 are not sufficient arguments. If not structure and composition determination by TEM and one of the electron spectroscopies (e.g. photoemission), respectively, the authors could at least use their STM for structural analysis. Without it, the islands may as well have the magnetic Fe3Si crystal structure, or even one of the non-magnetic disilicide variants. There have been numerous reports on magnetic properties of disilicide nanocrystals (cf. He at al. JACS 137, 11419 (2015), or Goldfarb et al. Adv. Mater. 30, 1800004 (2018)).

Þ We agree Referee’s comment. It is necessary touse the structure and composition determination ofTEM and electron spectroscopiesto confirm the formation of Fe₅Si₃. However, we haven’t these instruments, we need to collaborate with other laboratories. It will take a long time to complete this experiment. Additionally, because the STM image corresponds to the special distribution of electronic densities of surface atoms, it is difficult to directly determine the atomic structure by usingSTM.

Þ2-nd Referee:  This is a rather strange argument.  I can understand the lack of ex-situ characterization equipment, though this is not a good enough line of defense if the authors insist to nominate the particular silicide structures they observe as Fe5Si3.  However, at least STM was invented exactly for the purpose of atomic structure determination, and nicely resolved images of a clean Si surface in Fig. 1 prove that their STM is certainly capable of it.

Nevertheless, the metallic Fe₃Si is only formedbelow ~500°Cwith similar amounts of deposited Felike our work, as reported in the literatures [J. Appl. Phys. 105, 07B102 (2009); J. Appl. Phys. 104, 093707 (2008); Appl. Phys. Lett. 93, 132117 (2008)], and the disilicide is semiconducting[JACS 137, 11419 (2015)]. Thus, the metallic Fe-silicide nanowires formed at ~700°C in our work are likely to be Fe₅Si₃according to the equilibrium phase diagram of Fe–Si in Refs. [10,11] and the metallic character shown in I-Vcurves of these parallel nanowires(Figure 4).

To make this point clear, we revised the statements in 111^thline of our original manuscript as the following: “RTmagnetismof Fe-silicide NWs has been reported in previous studies on the magnetic properties ofFe₃Si [10,11],Fe₅Si₃[12,13]and b-FeSi₂ [14]NWs. According to our growth conditions for Fe-silicideNWs, these NWs are likely to be iron-rich Fe₅Si₃because the equilibrium phase diagram of Fe–Si shows that Fe₅Si₃is formed at 500‒700°C [12,13], whereas Fe₃Si is formed below 500°C [10,11]. This result can be further confirmed by the metallic charactershown in I-Vcurves of these parallel NWs (see Figure4), which is consistent with that of Fe₅Si₃ NWs and rules out the possibility of semiconducting b-FeSi₂NWs [14].”.

Þ2-nd Referee:  As already mentioned in my 1-st report, these are not very convincing arguments, because the kinetics in epitaxial deposits often prevails over thermodynamics, and can differ between various experimental systems and growth events.

2.Similarly, the authors hypothesis of a flat interface, has to be better substantiated. While this may be the case, there are also plenty reports on faceted interfaces of silicide islands with silicon substrates, including those of Ti, Er, Co, and Fe (see, for example, Surf. Sci. 457, 147 (2000), Physica E 43, 176 (2010), Surf. Sci. 606, 1649 (2012), PRB 96, 045415 (2017)). X-sectional TEM analysis could shed light on this issue.

Þ We are very sorry for this mistaken presentation in the 91^thlineof our original manuscript. In order to avoid readers’ misunderstanding, we replaced “a flatinterface” by “a well-definedinterface” in the text.

As shown in X-sectional TEM analysis in Figures 5‒8 of Referee’s suggested literature [Surf. Sci. 606, 1649 (2012)], all self-assembled silicide islands were observed to grow into the Si(001) surface with sufficiently well-defined interfaces. Also, the formation of a well-defined interfacebetween the self-assembled silicide nanostructures and various Si substrates via an “endotaxial” growth mechanism has been demonstrated in Refs. [13,14] and the literature [Surf. Sci. 606, 1649 (2012);CrystEngComm, 20, 2916‒2922 (2018)]. 

In 88^thlineof our original manuscript, we have addressed:“the formation of trenches indicates that self-assembled Fe-silicide nanowires grow into the Si(110) substrate via an endotaxial growth mechanism, as manifestly shown in the SP-STM image (Figure 3a) and its cross-sectional profile (Figure 3c). Therefore, there should be a well-defined interfaceformed between self-assembled Fe-silicide nanowires and the Si(110) substrate.”.   

Þ2-nd Referee:  OK.

3.The magnetic SP-STM contrast is very interesting, but its interpretation raises some questions, as well. Have the authors modulated the tunneling bias for their spectroscopic SP-STM acquisition? It is not mentioned in the experimental description. 

Þ We thank Referee’s suggestion. As we have addressed in the experimental section: “SP-STM/STS measurements were acquired in the current imaging tunneling spectroscopy (CITS) mode at RT.”. In the CITS mode, the dI/dV curves (i.e., SP-STS spectra) were obtained by numerically differentiation of the I(V) curves, as reported by the literature [Microsc. Res. Tech. 66, 93–104 (2005)]. It is different from the lock-in techniquewitha modulation voltage of ~20 mVat 2k Hzadded to the tunneling bias for their spectroscopic SP-STM acquisition(see the original Refs. [28‒31]).

   Thus, we added statements in the experimental section: “Each I–V curve was obtained by recording the tunneling current (I_t) while ramping the bias voltage (V_b) from−2.0 to 2.0 V with a voltage step of 0.05 V. The I−V curve was then numerically differentiated to derive a dI/dV curve (i.e., a STS spectrum).”, and also added a literature of the CITSmeasurement [Microsc. Res. Tech. 66, 93–104, (2005)].

Þ2-nd Referee:  OK.

4.The effect of stray fields from a ferromagnetic Ni tip on the contrastis not discussed. 

Þ Because Ni is a soft magnetic metal with a smaller magnetic moment (0.6 m_B), Ni produces a weaker stray field. The effect of stray fields from a Ni tip on the magnetic contrast of samples is weak, this makes Ni tips reliable as SP-STM tips, as reported in the literatures [Rev. Sci. Instrum. 90, 013704 (2019); J. Vac. Sci. Technol. B 32, 061801 (2014)].

Þ2-nd Referee:  Still, AFM tips, such as Cr that the authors themselves mention in the next section, are preferred over FM tips (including Ni), because of the that reason, namely stray fields.

5.Determination of the magnetization directions of the island domains and tip polarization is not obvious. Polarization of the tip along its length could rather be expected. 

Þ The direct determination of the magnetization orientation of the tip requires utilizing an external magnetic field to in-situdetect the out-of-plane and in-plane magnetization of the samples via the measurement of the dI/dV hysteresis loop curves [Nano Convergence 4, 8 (2017); Jpn. J. Appl. Phys. 51, 0302089 (2012);Science 327, 843 (2010)]. However, we have no external magnet in our STM system. Although we can’t experimentally show the magnetization orientation ofthe tip, we utilizedRef. [27] and literatures [Appl. Phys. Lett., 84, 9 (2004);Phys. Rev. B 72, 214409 (2005);Surf. Sci. 600, 1586–1591 (2006)] to support that the bulk Ni tip exhibitsthe in-plane magnetization. In the following, we give our explanations.

(1)Because the magnetization directions of elongated island domains with the inclinedsurfaces typically exhibit the in-plane orientation [Appl. Phys. Lett., 84, 9 (2004);Phys. Rev. B 72, 214409 (2005);Surf. Sci. 600, 1586–1591 (2006)], the magnetization directions of the nanowires with the triangular section should also exhibit the in-plane orientation. Thus, we have addressed in 157^thlineof our original manuscript:“there are two distinct domains with opposite in-plane magnetizations oriented along the individual triangular-sectional NW, as usually observed in the domain structures of Fe ribbons [26].”.

(2)Because the spin-dependent tunneling current in SP-STM depends on the relative orientation between the magnetizations of the measured region and the tip, another way is to use the well-known magnetization directions of the samples to confirmthe spin polarization of the tip, as reported in the original Ref. [27]. Becausethe in-plane magnetization of the triangular-sectional nanowiresexhibits the white contrast, this observation reveals that the magnetization of the triangular-sectional nanowiresis parallel to the tip magnetization. Consequently, the bulk Ni tip in our SP-STM also exhibits the in-plane spin polarization, similar to the report of the bulk Cr tip with a in-plane magnetic sensitivity (i.e. Ref. [27] ).

   To make this point clear, we added statements in 162^thlineof our original manuscript: “Because the tunneling current depends on the relative orientation between the magnetizations of the measured region and Ni tip, if the magnetization of the measured region is parallel to the tip magnetization, the dI/dV map is bright. While the magnetizations of the measured region and Ni tip are anti-parallel, the dI/dV map becomes darker.”.

Þ2-nd Referee:  This whole discussion, as well as discussion in item 7 below, is highly speculative, where previous suppositions lay out the foundation for the next ones, e.g., “…if we assume this…then we can also assume that, and these lead to…” and so on.

6.Assignment of the Neel wall characteristics to the white stripes on the island top is not clear either. The description of out-of-pane rotation in the z-y plane given in the text, is more consistent with Bloch wall.

Þ The following two figures show the magnetization reorientation in the different types of domain walls (from https://iop.fnwi.uva.nl/cmp//qem/research_projects/domainwall.html). The type of domain wall in Figure (a) is a Néel wall, Figure (b) shows a Bloch wall. Figure (c)shows a 3-D view of a Néel walland a Bloch wall. According to our discussions in Figure 6, the type of domain wall on the top of the wider rectangular-sectionalnanowireBin Figure 3b should be a Néel wall.

Þ2-nd Referee:  OK.

(a)     Néel wall

(b)    Bloch wall

(c)    

7.Finally, no explanation is given on the difference between magnetic structure of the islands with rectangular section and those with triangular section.

Þ In 150^thline of our original manuscript, we have explained the origin of the magnetization modulation in the repeatedly arranged streak domains. This magnetization modulation originates from the competition between the magnetocrystalline polarization along the easy axis and the shape anisotropy energy along the wire axis[25]. In the triangular-sectional nanowire E, the shape anisotropy is stronger than the magnetocrystalline anisotropy,whichtends to align the magnetization along the nanowire, as seen inthe in-plane magnetization of the nanowire Ein Figure 5c. 

Þ2-nd Referee:  Why?  Both types seem to be similarly elongated, with the only real difference between them being the cross-sectional shape.  So, shape anisotropy should be similar, and so is the easy axis.  If both types of structure are also similarly oriented with respect to the substrate, than the magnetocrystalline direction should be identical, as well.  But to find out that, again, a crystallographic analysis is required.

In summary, the authors made am effort to address the comments.  Yet, some key questions remained unanswered. In particular, plethora of new magnetic properties in variously composed and shaped epitaxial iron silicide nanostructures is vast, including in bulk non-FM phases, e.g. disilicides, has been reported.  This info could be expected to be discussed or at least mentioned, before narrowing down to bulk-FM Fe5Si3 and Fe3Si, and eventually – and speculatively – to Fe5Si3. Because such selection is not as straightforward as the authors might have initially believed.  The reorientation of magnetization in one type of NW but not in the other is a very important point, perhaps a pivotal point of this paper, and should be better substantiated, as well as shape anisotropy vs. MAE argument. If the authors are prepared to go this extra mile, the paper can be published.     

Comments for author File: Comments.pdf

Author Response

Thank you for your comments. Please find attached the detailed response to your comments.

Author Response File: Author Response.pdf