1. Introduction

Multilayer mirrors operating at the Brewster incidence angle (about 45° in our case) are commonly used as reflective polarizers/analyzers for the soft X-ray (SXR) and extreme ultraviolet radiation. By now, multilayer polarizers (MPs) of different compositions have been fabricated, studied, and put to practical use in different spectral regions. Among them, we can indicate Mo/Si MPs (working wavelength range λ ~12.5–20 nm) [1], Mo/Y MPs (λ ~8–13 nm) [2], La/B₄C MPs (λ ~6.6–8 nm) [3], and Sc/Cr MPs (λ ~3.1 nm) [4].

Nowadays, SXR polarimetry has attracted considerable interest of astrophysicists. There are manifold cosmic objects, such as pulsars, active galactic nuclei, and binary black holes, whose emission spectra lie in the SXR range, the radiation being partially polarized with the degree of polarization varying from 10% to 30%. Detection and analysis of the polarization degree can gain a deeper insight into the emission mechanisms of these sources.

Recently, a satellite mission concept called the Lightweight Asymmetry and Magnetism Probe (LAMP) has been suggested to study the emission of pulsars and the synchrotron-like emission produced by relativistic jets in blazars [5]. The key element of the LAMP is the SXR polarimeter, operating at about 250-eV photon energy (the wavelength λ ~5 nm) and 45° angle of incidence. Evidently, development of high-performance MPs is of extreme importance for this project. Note that the response of the LAMP is determined by the integrated reflectivity ${\displaystyle \int \left[R_s(E)-R_p(E)\right]}dE\approx {\displaystyle \int R_s(E)dE}$, rather than the peak reflectivity value [5]. Here, R_s and R_p are the reflectivity of s- and p-polarized radiation, respectively, and integration is performed over the photon energy. R_p is negligible compared with R_s because the incidence angle is close to the Brewster's angle (about 45° in the SXR region).

Carbon-based multilayers demonstrate the highest reflectivity at the λ ~4.5–6.5 nm wavelength, because of the minimal absorptivity of carbon in this spectral interval lying above the K-edge of absorption (λ ~4.4 nm). Fabrication technology of Cr/C multilayers has been likely the most extensively studied topic to date. In particular, we demonstrated in the previous paper [6] that the reflectivity of Cr/C multilayers could be as high as 21.8% at the 250-eV photon energy and 45° angle of incidence (s-polarized radiation case). However, as chromium is a relatively light material, variation in the dielectric constant at Cr–C interfaces is rather low, and thus, the penetration depth of SXR radiation into the multilayer structure is large. The result is a narrow spectral bandwidth of reflection and a small-integrated reflectivity of only 0.8 eV [6].

We can imagine two possible ways to broaden the reflectivity curve. First, we may use the depth-graded multilayer structure with the period varying with depth, while the deposition technology becomes more complicated. Second, we may use the conventional periodic multilayer structure, but with a heavier absorbing material providing a larger variation in the dielectric constant at interfaces and, hence, smaller penetration depth and a wider reflectivity band.

One of the possible candidates is the Co/C periodic multilayer structure. The polarizability of cobalt is about 33% more than that of chromium, which results in an approximately 45% increase in the dielectric constant variation at the Co–C interfaces as compared with the Cr–C interface. Therefore, the theoretical reflectivity of the Co/C multilayer mirror (~40% at the 250-eV photon energy) exceeds the reflectivity of Cr/C one by about 3% despite the fact that the absorptivity of Co exceeds that of Cr by a factor of 1.5. The integrated reflectivity is expected to rise to 1.2 eV, because of the increasing reflectivity bandwidth, if we assume that the effects of interfacial roughness and interlayers are the same as for Cr/C multilayers. One more advantage of the Co/C multilayer is that the required number of bi-layers is decreased by 1.5 times as compared with the Cr/C structure.

However, as it was demonstrated by Bellotti and Windt [7], strong interdiffusion between neighboring layers was observed in Co/C multilayer structure. The diffusion results in essential smoothening of interfaces, thus decreasing variation in the dielectric constant. One possible way to stabilize interfaces in a multilayer structure is through the addition of a certain amount of nitrogen to a working gas (typically Ar) during magnetron sputtering deposition. This approach was successfully used for fabrication of different multilayers [8–13]. Bellotti et al. [7] and Bai et al. [14] applied this method to stabilize interfaces in the Co/C multilayer structure by adding nitrogen of 6% and 25% partial pressure, respectively, while the achieved reflectivity in the SXR region was not indicated in their results.

It should be noted that a comprehensive study of variations in the internal structure and chemical composition of Co/C multilayers caused by adding of nitrogen to the working gas as well as the determination of an optimal amount of nitrogen providing the highest reflectivity in the SXR region has not been performed to date. Therefore, the present paper presents such a study.

Fabrication of the samples and characterization of their internal structure by the grazing incidence hard X-ray reflectometry and transmission electron microscopy (TEM) are described in Section 2. The compositional and chemical analysis of Co/C multilayers by X-ray photoelectron spectroscopy (XPS) and energy dispersive X-ray spectroscopy (EDX) are discussed in Section 3. The results of the SXR reflectivity measurements are presented in Section 4. The main results of the paper are summarized in Section 5.

2. Co/C multilayer fabrication and structure characterization

Due to an interest in the application of Co/C multilayers as polarizers in the LAMP project, the geometrical parameters of Co/C multilayers were optimized to obtain the maximal integrated reflectivity at the 45° angle of incidence. Assuming the bulk density of both materials and neglecting the formation of interlayers, the optimal period of multilayers was found to be d = 3.55 nm, the carbon layer thickness d_C = 0.6d, and the total number of bi-layers N = 100.

Direct current magnetron sputtering was used to fabricate the samples [15]. The Co/C multilayers were deposited onto super-polished silicon wafers (10 mm × 10 mm size) with a root mean squared (rms) surface roughness of 0.2 nm, and the last value being determined from a 2 × 2 μm atomic force microscope scan. The base pressure in the technological chamber before deposition was equal to 1.0 × 10⁻⁴ Pa. The spin motion of substrates during deposition was used to improve coating uniformity. The working gas was either pure Ar (99.999%) or Ar with a small proportion of pure nitrogen (99.999%) added, namely, Ar + 4% N₂, Ar + 6% N₂, Ar + 10% N₂, and Ar + 15% N₂, where the nitrogen content is defined as the ratio of the partial pressure of N₂ to the total pressure in the deposition chamber, with the last being equal to 0.266 Pa (~2 mTorr).

First of all, the fabricated multilayers were characterized by the grazing incidence X-ray reflectometry (GIXR) with the use of Cu-K_α radiation (λ = 0.154 nm) and the lab-based diffractometer (D1 system, Bede Inc.). The shift of the Bragg peaks’ position shown in Fig. 1 demonstrates that the multilayer period is varied from 3.44 nm to 3.59 nm for different samples. The main problem with fabricating multilayers with precisely the same period and thickness ratios is that the deposition rate of C increases significantly with an increasing proportion of N₂, while the deposition rate of Co is changed only slightly. A similar phenomenon was observed earlier in [9] in the studies of W/B₄C multilayers.

 

Fig. 1 GIXR measurements of Co/C multilayers at the 0.154 nm wavelength. The multilayers were deposited with the use of pure Ar and mixtures of Ar + 4% N₂, Ar + 6% N₂, Ar + 10% N₂, and Ar + 15% N₂ as the working gas.

Download Full Size | PDF

Figure 1 shows that only three Bragg peaks are observed for the Co/C multilayer fabricated with pure Ar, while the fourth order peak arises after adding 4–6% nitrogen. The first to third Bragg peaks’ reflectivity was also enhanced. These facts indicate that the addition of a small proportion of nitrogen to the working gas reduces the intermixing of neighboring layers and/or reduces interfacial roughness as compared with the structure fabricated with pure Ar.

Further increase in the nitrogen content up to 10% and 15% results in a widening of the high-order Bragg peaks and decreasing peak reflectivities, which signifies that the periodicity of the multilayers worsened because of the increased instability of the sputtering process after the addition of nitrogen to the working gas.

An internal structure of multilayers was studied with TEM measurements that were performed at the Materials Analysis Technology Inc. The samples for the TEM analysis were prepared by focused ion beam milling with the use of an FEI Nova600 instrument. The high-resolution transmission electron microscopy (HRTEM) and the selected area electron diffraction (SAED) analyses were performed using an FEI Tecnai G2 F20 instrument operating at 200 keV with 0.1 nm resolution.

The HRTEM images of internal structure of the samples are shown in Fig. 2, where the carbon layers appear as the light areas and the cobalt layers as the dark areas. Well-pronounced interlayers formed due to interdiffusion between Co and C layers are observed in Fig. 2(a) for the multilayer mirror deposited with pure Ar. Adding 4% nitrogen increases the sharpness of interfaces and thus enhances the reflectivity, while narrow interlayers are still observed at the C-on-Co interface seen in Fig. 2(b). Further increase in the ratio of nitrogen results in an essential reduction in the interlayer thickness. As seen in Figs. 2(c) and 2(d), the multilayers exhibit extremely smooth and sharp interfaces.

 

Fig. 2 Cross-sectional HRTEM images and SAED patterns (insets) of Co/C multilayers deposited with different working gases: (a) pure Ar; (b) Ar + 4% N₂; (c) Ar + 6% N₂; and (d) Ar + 15% N₂.

Download Full Size | PDF

The insets in Fig. 2 show the SAED pattern images. Contrary to the results reported in [7, 16], only faint and broadened diffraction ring indicated by arrow in Fig. 2 is observed, which can be attributed to tiny grains of cubic (111) or hexagonal (220 or 002) cobalt. We suggest thus that the layers of both materials are in an almost amorphous state. This result was confirmed by the X-ray diffraction measurements, where no diffraction from crystalline grains was observed.

3. XPS and EDX compositional and chemical analysis

To study the chemical composition variation with the depth, Co/C coatings deposited with pure Ar and a mixture of Ar + 6% N₂ as the working gas were investigated using X-ray photoelectron spectroscopy (XPS) and energy dispersive X-ray spectroscopy (EDX).

The XPS measurements were carried out with the AXIS Ultra DLD spectrometer and Al-K_α characteristic emission line (photon energy E = 1486 eV). An Ar ion beam was used to etch the samples with an estimated etching rate of 0.4 nm/min. The C 1s, Co 2p, O 1s, N 1s, and Ar 1s core level spectra were recorded in different layers of the samples. The binding energy scale was calibrated by using the argon photoelectrons spectrum emitted from deep layers. Taking into account the errors of measurements, calibration and fitting procedure, the uncertainty of the binding energy was estimated to be ± 0.2 eV.

The average atomic concentrations of C and Co in the multilayer fabricated with pure Ar proved to be 61.0 at.% and 35.8 at.%, respectively, with a concentration ratio of about 1.7. This value is in agreement with the ratio of 1.8 calculated on the basis of the designed layer thicknesses assuming the bulk density of both materials. Only an insignificant amount of oxygen (<0.5 at.%) and nitrogen (<0.3 at.%) was observed inside the structure except in the top bi-layer, where the concentration of oxygen was several times higher. After the addition of 6% N₂ to the working gas, the average concentration of nitrogen was increased to 7.8 at.%. Unfortunately, the depth distribution of the nitrogen is not well defined in the XPS profile because of the finite depth resolution of XPS dictated by the inelastic mean free path length of photoelectrons in matter (~1–2 nm at the excitation energy) and possible intermixing of layers under the Ar etching of the short-period multilayer structure resulting in the smoothening depth distribution of atomic concentrations.

To obtain a clear knowledge of nitrogen depth distribution after nitridation of the Co/C multilayer structure, we used the EDX technique, which was performed with the same FEI Tecnai G2 F20 instrument used for TEM. A well-focused electron beam (1~2 nm size) fell normally onto the multilayer cross section and moved along it perpendicular to interfaces. Characteristic fluorescent radiation of different chemical elements was recorded, allowing the composition distribution of chemical elements in depth of a multilayer structure to be determined.

The EDX profiles of the Co/C multilayers fabricated with the use of pure Ar and a mixture of Ar + 6% N₂ as the working gas are shown in Fig. 3. Experimental points were interpolated by the cubic splines. An essential shortcoming of the EDX technique is the impossibility to determine the absolute value of atomic concentration without comparison of the registered signal with that of an etalon sample. At the same time, relative variation in the concentration of the same chemical element is adequately determined by the EDX, which is sufficient for the purposes of our study. Nevertheless, to provide more evident physical meaning of the curves shown in Fig. 3, we normalized the averaged EDX signal to the average atomic concentration found by the XPS.

 

Fig. 3 EDX depth profiles of the relative atomic concentration of the dominant chemical elements composed of Co/C multilayers fabricated with (a) pure Ar and (b) a mixture Ar + 6% N₂ as the working gas.

Download Full Size | PDF

Comparison of the concentration profiles in Figs. 3(a) and 3(b) allows us to reach the following conclusions. First, the amplitude of the Co and C concentration oscillations increases essentially after the addition of 6% nitrogen to the working gas, demonstrating increasing sharpness of the Co/C multilayer structure interfaces that agrees with the TEM images. Second, the most interesting and unexpected result is that the depth distributions of nitrogen and carbon concentration are congruent, that is, concentration of nitrogen is maximal inside carbon layers rather than in cobalt layers.

Figures 4(a) and 4(b) show the C 1s core-level photoelectron spectra of two Co/C samples deposited with Ar or Ar + 6% N₂ as the working gas, respectively. The core peaks were analyzed using a nonlinear Shirley-type background, and all components were fitted with a product of 70% Gaussian and 30% Lorentzian line-shapes. To fit the spectrum from the sample deposited with pure Ar, three components were introduced: A1 at the binding energy of 283.7 eV (48.2% content), A2 at 284.6 eV (33.4%), and A3 of a small contribution at 285.6 eV (18.4%). The peak A2 can be attributed to graphite-like or amorphous carbon and the peak A3 is assigned to sp³-hybridized carbon [17–19]. The peak A1 is assigned to Co–C bonds according to the results reported in [20]. The lower energy of this component is attributed to an electron charge transfer from Co to C. The high content of Co–C bonds indicates a strong intermixing of Co and C layers, which is consistent with the results of TEM shown in Fig. 2(a).

 

Fig. 4 C 1s photoelectron spectra obtained at the excitation energy of 1486 eV from Co/C multilayers deposited (a) with pure Ar and (b) a mixture of Ar + 6% N₂ as the working gas.

Download Full Size | PDF

For the sample fabricated with a mixture of Ar + 6% N₂, the binding energy of the peaks A1 and A2 remains unchanged. However, the feature of A3 shifts to higher energy, and two peaks are needed to fit this shift: A3* (285.9 eV, 15.7%) and A4* (287.7 eV, 9.0%), which correspond to two different binding states between carbon and nitrogen [19–21]. The content of the peak A1 decreases from 48.2% to 39.9%, which confirms the decreasing interdiffusion between Co and C layers due to the formation of C–N bonds. When fitting, the A3 component, which gives a small contribution to the spectrum, was not used to simplify the model, while an existence of C-C sp3 bonds in the sample prepared with Ar + 6% N₂ cannot be excluded.

The N 1s core-level spectra of the sample deposited with Ar + 6% N₂ working gas is shown in Fig. 5. Three components are used for fitting: B1 at the 398.1 eV binding energy, B2 at 400.2 eV, and B3 at 402.8 eV. The peak B3 is attributed to N–O or N–N bonds. The other two peaks, at 398.1 eV and 400.2 eV, are identified as originating from N–C bonds, which correspond to the two N–C binding states shown in Fig. 4 [20]. According to the available data [19, 21, 22], peaks B1 and B2 are attributed to nitrogen bonded to sp³- and sp²-hybridized carbon, respectively. This is consistent with the peaks A3* and A4* in the C 1s photoelectron spectrum in Fig. 4(b). Co–N contribution cannot be found in the fitting.

 

Fig. 5 N 1s photoelectron spectra obtained at the excitation energy of 1486 eV from Co/C multilayers deposited with Ar + 6% N₂ working gas.

Download Full Size | PDF

We also measured the Co 2p core-level spectra shown in Fig. 6. The Co 2p peaks of the two samples are quite similar. Three components were used for fitting the Co 2p3/2 peak: C1 at 778.4 eV corresponding to the Co–Co bonds, C2 at 778.8 eV corresponding to the Co–C bonds, and a small contribution of C3 peak at 781.6 eV, attributed to a slight cobalt oxidation [23]. As shown in Fig. 4, the content of Co–C bonds decreases from 38.1% to 32.2% due to formation of C–N bonds after adding nitrogen to the working gas.

 

Fig. 6 Co 2p core-level spectra from Co/C multilayers deposited with (a) pure Ar and (b) mixture Ar + 6% N₂ as the working gas.

Download Full Size | PDF

Thus, XPS and EDX measurements allow us to reach the two following conclusions concerning the samples fabricated with the addition of nitrogen to the working gas. The first conclusion is an unexpected result: adopted nitrogen is mainly contained in carbon layers, rather than in metallic cobalt layers. The second conclusion is that the increasing nitrogen proportion in the working gas results in an increasing content of C–N bonds in carbon layers, thus decreasing mobility of carbon atoms and deterioration of interdiffusion between neighboring layers.

Moreover, one question still remains: the decrease of the content of Co–C bonds in Fig. 4 is not as significant as we expected from the HRTEM image. This is probably caused by the effect of Ar ions etching, which was investigated in [24]: for pure carbon, the C 1s peak shows no shift in position but was significantly broader after the etching of ion beam etching. For the CN_x layer, a shift of peak position to lower energy was observed. Therefore, if we were to consider the effect of ion beam etching in our XPS analysis, the C 1s core-level spectra in Fig. 4(a) remains unchanged, and the one in Fig. 4(b) shifts to higher energy, which leads to a decreasing content of Co–C bonds in the fitting.

4. Soft X-ray polarization measurements

The SXR reflectivity of the Co/C multilayers was measured for both s-polarized (R_s) and p-polarized (R_p) radiation at the 45° angle of incidence. The measurements were performed at the Bending Magnet for Emission Absorption and Reflectivity (BEAR) beamline of the ELETTRA synchrotron [25]. Figure 7(a) shows the dependence of Rs on the photon energy E for multilayers fabricated with different nitrogen content in the working gas. The shift of the Bragg peaks in respect to each other is caused by a slightly different period of the multilayer mirrors and different chemical composition of layers, that is, different optical constants influencing the Bragg peak position.

 

Fig. 7 (a) The soft X-ray reflectivity of the Co/C multilayer versus the photon energy for s-polarized beam at the 45° angle of incidence; (b) The peak value of the reflectivity R_s and the integrated reflectivity R_int versus the pressure ratio of nitrogen in the working gas.

Download Full Size | PDF

The reflectivity of the multilayer fabricated with pure Ar achieves the peak value R_s = 15.6% at E = 256.7 eV. As the pressure ratio of nitrogen increases to 4% and 6%, the reflectivity increases to 16.8% and 19.0%, respectively, owing to the stabilization of interfaces due to nitridation of carbon layers. However, a further increase in the nitrogen content in the working gas up to 10% or 15% reduces the reflectivity peak to 15.2% or 13.7%, respectively. Among the physical reasons for the reflectivity drop, we can recognize the following. First, high nitrogen content in the working gas results in increasing instability of the sputtering process, thus increasing random variations in layer thicknesses from the prescribed values. Second, the increasing proportion of nitrogen in carbon layers enhances the SXR absorption in the Co/C multilayer structure. Third, high nitrogen content in cobalt layers may decrease their density and, hence, reduce the dielectric constant contrast at interfaces.

The integrated reflectivity $R_{\mathrm{int}}={\displaystyle \int R_s(E)dE}$ of Co/C multilayers is shown in Fig. 7(b) in dependence on the partial pressure of nitrogen in the working gas. As with the peak reflectivity, the integrated reflectivity achieves the maximal value of 1.1 eV at 6% nitrogen content in the working gas. This value essentially exceeds the integrated reflectivity of the Cr/C multilayer (0.8 eV) reported in the previous paper [6], and it is close to the predicted value of 1.2 eV. Note that the optimal nitrogen content in the working gas providing maximal reflectance of the studied Co/C multilayers is similar with that found earlier for multilayers of other types: 6.4% for W/B₄C [9] and 6% for Pd/Y [13] mirrors.

Figure 8 demonstrates the reflectivity of the Co/C multilayer (s-polarized radiation case) fabricated with Ar + 6%N₂ as the working gas dependent on the photon energy for different angles of incidence. The peak reflectivity is increased to 23% when approaching the carbon edge of absorption (E ~284 eV), where the absorptivity of carbon is minimal.

 

Fig. 8 The measured reflectivity (s-polarized radiation) of the Co/C multilayer fabricated with Ar + 6% N₂ as the working gas dependent on the photon energy for different angles of incidence.

Download Full Size | PDF

The measured reflectivity of p-polarized radiation at the 45-degree angle of incidence was found to lie in the interval from 0.1% to 0.3%. In agreement with Fig. 7, this fact demonstrates the high potentialities of Co/C multilayers as polarizers in the 4.5–6.5-nm wavelength range.

5. Conclusion

The effects of nitridation on the internal structure and the SXR reflectivity of Co/C multilayer mirrors were studied. The multilayers were fabricated by the direct current magnetron sputtering technique, and the nitridation was carried out by adding a small proportion of nitrogen (4–15% partial pressure) to the working gas (Ar).

First of all, we demonstrated using HRTEM that nitridation caused deterioration of interdiffusion of carbon and cobalt layers inherent to multilayers fabricated with pure Ar as the working gas. As a result of the essential decrease in the interlayer thickness, the multilayer mirror reflectivity was increased from 15.6% to 19.0% at the 251-eV photon energy and 45° angle of incidence.

The physical reasons for the enhancement stability of interfaces were established by XPS and EDX analyses. We demonstrated that added nitrogen was mainly contained in carbon layers, rather than in metallic cobalt layers. This resulted in a great quantity of C–N bonds in carbon layers, thus decreasing mobility of carbon atoms and causing deterioration of interdiffusion between neighboring layers.

The optimal partial pressure of nitrogen providing the maximal value of both peak reflectivity and integrated reflectivity was found to be about 6%, with the reflectivity values being equal to 19% and 1.1 eV, respectively, at 251 eV photon energy and a 45° angle of incidence of s-polarized radiation. The reflectivity of p-polarized radiation did not exceed 0.3%, demonstrating high potentialities of Co/C multilayers for their use as polarizers in the 4.5–6.5-nm wavelength range.

Further increases in the nitrogen pressure retained abrupt interfaces, but resulted in decreasing reflectivity for the following reasons: (a) enhancement of the SXR radiation absorption in the nitrided carbon layers; (b) possible decrease in the cobalt layer density due to incorporation of nitrogen and thus the reduction in the dielectric constants contrast at interfaces of the multilayer structure; (c) degradation of the multilayer structure periodicity due to increased instability of the sputtering process after addition of nitrogen to the working gas.