1. Introduction

In order to achieve maximum dispersion in wavelength λ by use of reflection gratings at shallower angles of incidence a grating is usually operated in the so-called classical orientation [1]. In this case the dispersion plane coincides with the plane of incidence, which contains the incident beam and the normal to the surface. In such a configuration the grooves are extending perpendicularly to this plane of incidence. However, a reflection grating will also diffract the incident radiation, when the structure is simply rotated by 90° around the surface normal, such that the grooves are now presenting themselves parallel to the incident radiation beam. This configuration is referred to as the extreme off-plane orientation [2]. The related diffraction phenomenon is also called conical diffraction [3], as the different diffraction orders now line up on an arc, which is symmetrically oriented with respect to the direction of the grooves. In this case the incident beam illuminates the entire surface and no part of it remains in the shadow, as it is almost always the case in the classical configuration in combination with grazing angles of incidence. This off-plane configuration was proposed for sawtooth profiles with small blaze angle, i.e. with very shallow inclination angle between the surface and the longer tooth side. Intuitively one can expect very high diffraction efficiency, as now the diffraction into one particular order could be favoured, when the respective radiation beam is simply specularly reflected at the groove surface, like it was a mirror [2,4]. This will require to operate the object at grazing angles of incidence in the total reflection regime. Presently such objects are chosen for some special soft X-ray monochromators, when very short pulses from pulsed sources are not to be lengthened significantly upon diffraction [5]. This pulse lengthening in the diffraction process cannot be avoided in the classical orientation.

Differently to the asymmetric sawtooth profile the highly symmetric rectangular profile will redistribute the diffracted intensity symmetrically around the plane of incidence and it was thus proposed and successfully tested as an efficient beam splitter for soft X-ray and extreme ultraviolet radiation [6]. Even though it was predicted, that the performance would also be very promising for harder X-rays [7], the related experimental verification was not reported sofar. The present study addresses this point, showing that the expected highly efficient and symmetric beam splitting can be achieved in such a grating also for harder X-rays in the photon energy range 4 keV < E < 12.4 keV (X-ray wavelength 0.3 nm > λ > 0.1 nm), which is the range of particular interest at X-ray free electron laser sources [8].

2. Theoretical considerations

When a rectangular grating profile is operated in the extreme off-plane configuration, in which at very grazing angle of incidence the incident beam, i.e. ray A in Fig. 1, has its trajectory parallel to the grooves the situation looks very similar to the beam dispersion by use of a transmission grating with rectangular bars. Upon diffraction at the grating structure the X-ray beam is deflected away from the plane of incidence, which contains the incident beam and the normal to the diffracting surface. The purely specularly reflected beam will here be considered to be the zero order diffracted beam. Assuming hard X-ray wavelengths λ (fractions of nm) to be negligible compared to practical grating periods p (hundreds of nm) the diffraction angle ϕ_m for all other orders with index m away from the plane of incidence is then given as

 

Fig. 1 Orientation of a grating with grooves of rectangular profile for horizontal beam deflection in the extreme off-plane configuration. The inset at upper left presents the properties of the grating, in which rectangular bars of width a are separated in a structure with periodicity p by grooves of depth d. The drawing is not made to scale. The incident beam A arrives from the left and is diffracted to the right symmetrically around the plane of incidence as indicated by the two rays B and C. All diffraction peaks of different order will line up on half circles. The opening angle, under which such half circle is seen from the center of the grating, is identical for all sources, which are positioned on a half circle with identical opening angle upstream of the grating. The inset at right shows the dimensions of the tested grating. The angle of grazing incidence is denoted by θ. The two borders of 9 mm x 50 mm remained unetched and were coated completely like the ruled area. The shaded area represents the maximum footprint that the grating can accommodate for θ>0.46° in the horizontal direction.

Download Full Size | PDF

Unlike in the classical orientation here the diffraction angle does not depend on the chosen angle of grazing incidence. The ratio λ/p is of the order of 10⁻³, and thus adjacent diffraction orders are only little separated, which will then result in a large number of propagating orders.

Ideally a transmission grating is composed of transparent slits, which are separated by rectangular bars, in which the intensity of the transmitted beam is attenuated or eventually completely absorbed. When some intensity is still transmitted the respective wavefront will be subject to some phase retardation. Now in the present reflection grating with rectangular bars, the intensity attenuation will not be observed, when the reflectivity in the grooves is identical with the reflectivity at the tops. However, one will now find a global intensity attenuation due to the finite reflectivity (R<1) of the coating material. The phase retardation instead will be observed as a result of the optical path difference (abbreviated in the following as OPD) between the rays hitting tops and grooves.

Now for the perfect transmission grating the diffraction efficiencies into the different orders τ_m can be calculated analytically, as was shown by Schnopper et al [9]. In the present situation one would consider these diffraction efficiencies τ_m to be the structural efficiencies, which will have to be multiplied with the coating reflectivity R in order to obtain the practically measurable efficiencies

In the scalar approximation according to [9] the diffraction efficiencies for perfectly transparent slits are given in general by

and by

The two constants c₁ and c₂ refer to the relative intensity attenuation in the bars and to the related phase retardation of the transmitted radiation, respectively. In case of a transmission grating structure the unavoidable beam attenuation in the bars will lead to c₁<1. Instead in the reflection grating c₁ = 1 can be realised. This leads then to

c₂ is the phase retardation term, which in a rectangular profile grating with constant OPD is given as

Interestingly the efficiencies remain independent on the grating periodicity. And for m≠0 they are the result of a product of two uncorrelated factors. The only variable in the first factor $S_m={\left(\frac{\sin \left(m\pi a/p\right)}{m\pi }\right)}^2$ in Eqs. (3) and (5) is the ratio a/p, while the second factor 2(1-c₂) varies only with c₂.

Now in order to make the grating perform as an efficient perfect amplitude beam splitter it needs to concentrate the diffracted intensity primarily in only one pair of equivalent orders. This requires two diffraction peaks and thus m≠0. Maximum efficiency will then require S_m as well as 2(1-c₂) to be maximum. The maxima for S_m are obtained for a/p = 0.5 and for odd orders with $S_1={\left(1/\pi \right)}^2=0.101$, $S_3={\left(1/\left(3\pi \right)\right)}^2=0.0113$ and $S_5={\left(1/\left(5\pi \right)\right)}^2=0.00405$. Instead the second factor is maximum for c₂ = −1 with 2(1-c₂) = 4. Then for a/p = 0.5 and c₂ = −1 all even orders and the zeroth order vanish, while for c₂ = 1 all intensity will be contained in the zeroth order.

Figure 2 presents the theoretical dependence of the structural efficiencies τ_m on the ratio a/p and for c₂ = −1. One sees that ideally the two first orders with the absolute value of |m| = 1 can contain 81% of the diffracted intensity for a/p = 0.5. This efficiency varies little for |m| = 1, and a deviation from the ideal ratio a/p = 0.5 with ratios in the range 0.4<a/p<0.6, i.e. by up to 20%, can be tolerated. Such errors will then lead to a slightly reduced efficiency into the first orders by about 10% only, while it will reduce the diffraction efficiency into higher odd orders relatively more significantly, like e.g. for |m| = 3 as shown. The related intensity redistribution will therefore mostly feed even orders with indices m = 0 and |m| = 2.

 

Fig. 2 Dependence of the structural efficiencies τ_m on the ratio a/p of the grating structure for the orders with indices 0≤|m|≤ 3.

Download Full Size | PDF

The working point with c₂ = −1 is realized with OPD = λ/2. The latter varies with the angle of incidence θ and then according to Eq. (6) the appropriate angle is found via

In order to achieve high practical diffraction efficiency and little absorption in the grating substrate, the reflectivity of the coating material needs to be high, and thus the angle of grazing incidence needs to be smaller than the critical angle for total reflection θ_crit. This imposes a limitation for the choice of d to

In summary a rectangular profile reflection grating will provide efficient and symmetric amplitude beam splitting in the first order diffraction peaks with |m| = 1 for a/p = 0.5. The choice of the groove depth d has to be made by use of Eq. (8) and this depth will in turn determine the optimum working angle according to Eq. (7). In this condition all other orders will receive minimal intensity. The separation angle between the orders can formally be obtained from Eq. (1), however, it does not depend on the angle of grazing incidence θ.

3. Experimental details

3.1 Grating properties

The investigated structure is a holographically prepared reflection grating with rectangular grooves of period p = 820 nm (1220 lines/mm). The nominal width of the tops a = 370 nm corresponds to a ratio a/p = 0.45, which implies that according to Fig. 2, this grating could perform almost ideally as far as first order efficiency is concerned. The rectangular grooves are etched to a depth of about d = 7 nm into a silicon carbide substrate (50 mm x 50 mm) over a length of 50 mm with 32 mm wide grooves as shown to the right in Fig. 1. After the etching the structure, including the unruled border, was coated with a gold layer of 30 nm thickness on a thin chromium binding layer. The object can withstand the heat load at Elettra soft X-ray undulators [12] and is thus also suited for the operation at X-ray free electron lasers, when used in the extreme off-plane mount at an angle of grazing incidence in the total reflection regime of the coating material.

3.2 Experimental setup

The test was performed at the X-ray fluorescence beamline at Elettra, where the sample was installed in a 7 axis diffractometer, which is operated in a vacuum chamber [13]. The radiation beam (3.7 keV < E < 14 keV) is provided by a vertically dispersing double crystal monochromator, which is operated in a vertically collimated beam [14]. The relative spectral resolution corresponds to the intrinsic limit of the Si(111) reflex, which is about 0.07%. The monochromatised beam is refocused into an exit slit 10 m downstream and higher orders are filtered out by use of a double reflection higher order suppressor.

The sample was oriented as shown in Fig. 1 for horizontal beam deflection. At the sample position the beam has a top hat profile with width of 0.25 mm in the horizontal plane of incidence and with height of 0.12 mm perpendicular to it (i.e. in the vertical plane). The intrinsic beam angular spread in both directions is 0.13 mrad. The entire beam footprint, presented as a shaded area in Fig. 1, is then intercepted by the grooves of length 32 mm for angles of grazing incidence θ>0.46° (7.8 mrad).

The profiles of the incident and the diffracted beams could be inspected by use of a CCD camera (PCO Sensicam qe), in which a fluorescence screen was mounted upstream of a 1:1 transfer lens system with high numerical aperture. This system was mounted at a sample distance of 0.55 m. The equivalent pixel size was 6.7 μm. Besides a careful quantitative analysis could be undertaken with a photo-diode detector (Hamamatsu Si-photodiode S3590-09) at a closer distance of 0.14 m. In this case a vertical slit of opening 0.2 mm limited the beam acceptance in the plane of incidence. In the first place the angle of grazing incidence θ was varied. Furthermore the grating could also be rotated around the surface normal for fixed angle of grazing incidence. This degree of freedom will here be referred to as the rotation in yaw.

4. Experimental results and discussion

The grating alignment to the extreme off-plane configuration was made by use of the CCD camera. When ideally aligned the diffraction pattern is completely symmetric with respect to the plane of incidence of the grating, as far as intensities and diffraction peak positions are concerned. This situation is shown in the central image in Fig. 3.

 

Fig. 3 CCD images of the distribution of the diffracted intensity taken for 4 keV photon energy and for an angle of grazing incidence of θ = 0.64°, when the grating is perfectly aligned in the extreme off-plane orientation (center image), and when it is rotated by |ϕ| = 0.5° around the surface normal (yaw degree of freedom) in both directions away from this condition. The white scale bars in the central plot measure 0.5 mm in length. The grating diffracted the beam to the right. The colour coding refers to a linear intensity scale. The images in the left column cover the entire intensity range, while in the right column the gain for the intensity was increased 10 fold in order to reveal weaker peaks.

Download Full Size | PDF

Here the grating dispersion is observed primarily in the vertical direction, however, one can also recognize a slight curvature in the line, which connects the centers of the diffraction peaks. This latter observation is expected for conical diffraction as well as the fact that only a few diffraction orders receive an appreciable intensity. The other two pictures are rotations in yaw to ϕ = −0.5° and to ϕ = 0.5° away from perfect alignment. In this case the orders remain oriented on circles as well, however, with the expected larger radii of curvature. The variation of the intensity in a given order remains small as long as the yaw angle remains small. The size and the shape of the diffracted orders are identical with the corresponding parameters of the incident beam. As only a few low index orders receive significant intensity in the aligned case the diffracted intensity remains almost on straight vertical lines, especially at the position photo-diode, which is at about ¼ of the CCD distance from the grating. Consequently in a specular reflectivity scan, i.e. in a θ-2θ scan, in which the diode detector moves by twice the angular steps compared to the change in angle of grazing incidence, one will now register the total diffracted intensity from the grating structure. As long as the ratio a/p is of the order of 0.5 and c₁ = 1, the integration of the structural diffraction efficiencies according to Eqs. (3) and (4) over all orders yields

Thus for constant incident intensity the total diffracted intensity is expected to be identical with the reflected intensity from a plane mirror with an identical coating. As is shown in Fig. 4, in which both intensities are normalised to the incident intensity, this is indeed measured depending on the angle of grazing incidence θ at the ruling and at the unruled stripe. In order to measure this latter stripe the grating was rotated in yaw by 90°. The data is presented for a photon energy of 6 keV (λ = 0.207 nm), and it is plotted only for those angles, at which the respective sample size can intercept the incident intensity completely. The unruled stripe of length 50 mm, i.e. the mirror, intercepts the entire beam crossection already for angles starting at 0.3° (5 mrad), while the 32 mm long grating structure can intercept it only for angles of grazing incidence above θ>0.46° (7.8 mrad). Then in the overlap region for θ>0.46° both reflectivity curves are essentially identical, which confirms the validity of the assumption c₁ = 1. The shoulder measured in both reflectivity curves around θ = 1° is characteristic for the nominal coating of 30 nm of Au on a thin Cr layer considering a rather large surface roughness of 1.8 nm rms.

 

Fig. 4 Specular reflectivities R depending on angle of grazing incidence θ (displayed linearly at left and logarithmically at right) measured in θ-2θ scans for a photon energy of 6 keV from the border of the substrate (red symbols) and from the grating structure, when aligned perfectly in the extreme off-plane orientation (black symbols).

Download Full Size | PDF

The intensity distribution into the orders was registered by use of the CCD camera. For this purpose the CCD camera was kept at a fixed position and the intensity distributions were registered in a multi exposure image presented in Fig. 5, in which the angle of grazing incidence was varied with a constant increment of 0.04° between θ = 0.36° (6.3 mrad) and θ = 0.76° (13.3 mrad).

 

Fig. 5 Multi-exposure CCD image of the distribution of the diffracted intensity for a photon energy of 4 keV taken, when the grating is perfectly aligned in the extreme off-plane orientation and the angle of grazing incidence is scanned in steps of 0.04° starting at left at an angle of θ = 0.36°. The white scale bar measures 0.5 mm in length. The colour coding refers to a linear scale. The upper image covers the entire intensity range, while in the lower image the gain for the intensity was increased 10 fold. The distribution of the intensity into orders with indices −6≤m≤6, as labelled, is displayed.

Download Full Size | PDF

In this Fig. 5 for a photon energy of 4 keV (λ = 0.31 nm) it is immediately obvious, that all diffraction orders move on horizontal lines parallel to the plane of incidence. Consequently and as predicted in Eq. (1), the diffraction angles for the different orders do not depend on the angle of grazing incidence θ. Additionally and as expected for conical diffraction, the radius of curvature on which the orders line up increases from left to right linearly with the angle.

In Fig. 5 one also sees immediately that in most of the positions the two first orders with |m| = 1 receive symmetrically most of the diffracted intensity. Only in the first registration does the zeroth order receive similar intensity. Note that the zeroth order is wider in the first 3 exposures than the other orders, as part of the beam is still intercepted and specularly reflected at the unruled border, as presented in Fig. 1. This zeroth order receives minimum intensity between the seventh and the eighth exposure at an angle of grazing incidence of thus about θ = 0.66° (11.5 mrad). The intensity into the second and the third orders is very similar and is increasing very slowly with increasing angle. The same holds in the comparison between the orders with index |m| = 4 and |m| = 6, which receive significantly less intensity. Instead the opposite is found for the fifth order, which vanishes progressively towards larger angles of grazing incidence. Negligible or no signal was observed for all orders with |m|≥7.

The quantitative analysis will now deal with the structural efficiency. This evaluation can be made in a relative way, as the total diffracted intensity from the grating was found to be identical to the reflected intensity from the mirror with the same coating. Then the structural efficiencies sum up to unity, i.e. ${\displaystyle \sum _{m=-\infty }^{m=\infty }{\tau }_m=}1$. Now in Fig. 5 independently in each exposure the registered counts were averaged in the largest window of fixed size, which covered the flatter part of each diffraction peak such that the counts from the adjacent orders would not affect the result. From these averaged counts S_m one can then obtain the structural efficiencies τ_m for the observable diffraction orders with indices |m|≤6 via ${\tau }_m=S_m/{\displaystyle {\sum }_{i=-6}^{i=6}{S}_i}$. The related data is presented to the right in Fig. 6, where it is compared to the expectations for the nominal ratio a/p = 0.45 and for the nominal groove depth of d = 7 nm, which are shown at left. As the summation was limited to the observable diffraction peaks, this procedure will systematically slightly overestimate the derived structural efficiencies. The error is rather small as less than 2% of the diffracted intensity was found to be contained in higher orders with |m|≥7.

 

Fig. 6 Left: Calculated structural efficiencies according to Eqs. (3) and (4) depending on the angle of grazing incidence θ for a grating with rectangular grooves and with a ratio a/p = 0.45. c₁ = 1 is assumed. Right: Experimental structural efficiencies for the single orders derived from the total diffracted intensities for a photon energy of 4 keV.

Download Full Size | PDF

Then in Fig. 6 one sees that the measured zeroth order efficiency is minimum almost as predicted at an angle of about θ = 0.66°. Now all other orders should theoretically provide their efficiency maxima or minima at the same angle. This is not really observed here. However, as the goal of this contribution is the verification of the symmetric and highly efficient diffraction into the two first orders, a discussion of the observed discrepancies compared to the expectations for the higher orders, goes beyond the scope of this contribution. Nevertheless a few observations can be made. The orders with indices |m|≥4 present experimentally rather small efficiencies almost in agreement with the expectations. Instead the lower index orders with |m|≠1 were measured in the vicinity of θ = 0.66° with higher than expected efficiencies. As a consequence the efficiency for |m| = 1 is reduced in this condition. Hence when used as a beam splitter the test grating will not provide the ideally expected performance. Nevertheless the relative efficiency for this first order is rather large in excess of 30%, which is 75% of the ideally expected efficiency of 40.5%. Moreover, as predicted this high efficiency is maintained in a large angular range.

In order to discuss the practically available efficiency in this case one can deduct from Fig. 7, that the reflectivity of the coating for the optimum operation angle of θ = 0.66° is about R = 0.45. In the experiment it was found that 60% of the diffracted intensity were concentrated into the two first orders, and thus the practical efficiency for the beam splitting into two symmetric orders according to Eq. (2) is in this case 27%.

 

Fig. 7 The normalized total diffracted intensity is presented as a black line depending on angle of grazing incidence θ for the grating in the off-plane orientation and for a photon energy of 4 keV. The red circles present the results for the angles, which were chosen for the exposures as reported in Fig. 5. For comparison also the ideal theoretical reflectivity for a thick gold mirror is presented as a blue line, while the green line refers to a Ni coating. The latter results were calculated by use of [16].

Download Full Size | PDF

The registered diffraction orders in Fig. 5 are well separated in the CCD pictures for 4 keV photon energy. According to Eq. (1) this separation will not be found anymore for larger photon energies. Infact in pictures taken at 6 keV photon energy (data not shown), the orders overlapped significantly. Nevertheless the angle for minimum diffraction into the zeroth order could be identified to be θ = 0.43° (7.5 mrad). This is in agreement with the expectation for the nominal groove depth of d = 7 nm, and it is almost exactly the angle, at which the grooves intercepted the incident beam completely. In the corresponding diffraction pattern the first order structural efficiency τ₁ was derived to be about 30% also in this case. And similarly to the observations at 4 keV this diffraction efficiency decayed towards larger angles and it remained roughly constant towards smaller angles. Now in Fig. 4 one finds at the optimum operation angle for a photon energy of 6 keV a relatively larger reflectivity of almost R = 0.6, and thus in this case the practical efficiency for the beam splitting amounts to about 36%.

The photon energy of 12.4 keV, i.e. a wavelength of λ = 0.1 nm, is usually considered as a reference photon energy and the goal for X-ray free electron laser operation [8]. For this photon energy the optimum working point is now predicted for a smaller angle of grazing incidence of θ = 0.22°, at which the grooves cannot intercept the entire beam crossection anymore. Consequently the throughput with respect to the incident beam is expected to be reduced. Nevertheless from the analysis of the related experiment (data not shown) one finds a practical efficiency for the beam splitting into the two first orders of still about 20% of the incident intensity. This indicates even for this photon energy a still high structural efficiency of the present grating.

5. Possible performance improvements

It is obvious from Fig. 2, that the mismatched present ratio of a/p = 0.45 will not affect the grating performance very much. So this ratio is not the parameter, which needs to be improved. Instead improvements need to be made in the grating smoothness and in the coating material. As far as the range of interest around the optimum working angle is concerned, the comparisons in Fig. 7 show that the present rather high surface roughness of 1.8 nm rms, which can explain the measured reflectivity curves, reduces the reflectivity by almost 30% compared to smoother state-of-the-art surfaces. A similar gain of 30% in diffracted intensity is thus feasible. A further improvement is possible by choosing the optimal coating material. Gold with its high atomic number Z is a rather absorbing material for harder X-rays. As a consequence even in the total reflection regime the reflectivity of a gold coating decays rapidly with increasing angle of grazing incidence, as is evident in the related simulation presented in Fig. 7. A relative reflectivity increase with then reduced absorption is also very desirable from the power load point of view. For both aspects other metals with lower Z, like Ni or Rh, could provide better performance in the total reflection regime (see Fig. 7). Ni would be suggested for operation at photon energies below 8 keV and Rh for operation at larger photon energies. The related critical angle for total external reflection of x-rays at a low absorbing material is given by ${\theta }_{crit}=\sqrt{2\left(1-n\right)}$, where n is the real part of the refractive index of the material [15]. This real part of the refractive index is given by [15] $1-n=\frac{r_e{\lambda }^2{N}_e{f}_1}{2\pi }$, here r_e = 2.818 10-15 m is the classical electron radius, N_e is the number of atoms in the coating material per unit volume and f₁ is an atomic scattering factor, which is tabulated [16]. For the example of Ni the scattering factor f₁ is almost a constant between the absorption edges at photon energies of 1.008 keV and at 8.333 keV. Consequently in this range holds ${\theta }_{crit}\propto \lambda$ or ${\theta }_{crit}\propto E^{-1}$. By use of the tabulation [16] one obtains a critical angle for Ni for a photon energy of 4 keV (see Fig. 7) of ${\theta }_{crit}=0.84°$. This critical angle is twice smaller for a photon energy of 8 keV, while it is twofold larger for 2 keV, respectively.

Considering the discussed possible improvements, it seems reasonable to assume that the practical efficiency of optimized rectangular grating profiles employed as amplitude beam splitters for hard X-rays could exceed 50% with respect to the incident intensity.

6. Conclusion

It was shown that reflection gratings with rectangular groove profiles can be used as efficient and symmetric beam splitters for hard X-rays, when operated in the extreme off-plane configuration. The test grating directed 27% of the incident intensity of 4 keV photon energy symmetrically into two well separated diffraction orders. For 6 keV photon energy the efficiency was slightly better with 36%. It is expected that values in excess of 50% are possible. Such objects could then be particularly efficient for the amplitude beam splitting at presently commissioned X-ray free electron lasers. As the entire surface of the grating will be operated in the total reflection regime of the coating material, a high incident power can be tolerated. Presently for the purpose of beam splitting at free electron lasers transmission gratings are used [17] as are mirror systems for beam division [18]. The first are rather fragile and are not compatible with higher power load, as some power will always unavoidably be absorbed in the bars. Instead in a beam divider a mirror, which is shaped properly up to its border, will intercept only part of the beam and will deflect it away from the passing beam. In this case both beams will be subject to unwanted edge diffraction. This will introduce additional features, i.e. artefacts, when both beams are recombined. Such “artefacts” will not be observed when amplitude beam splitters are employed.