Abstract
A new geometric scheme translates a diffraction grating along the straight central groove of an exponentially curved pattern. Lit by a stationary incident beam, the two-dimensional pattern scales isotropically, scanning wavelength without change to any angles, macroscopic distances, curvatures or aberrations. This is exemplified by a new class of self-focused grating monochromator, analyzed by rigorous light-path expansion and numerical raytracing. All spectral aberrations in pure meridional powers (including defocus, coma and spherical aberration) cancel for any angular deviation, magnification and translation range. The residual mixed powers yield Δλ/λ = 10−3 ~ 10−5 in the soft x-ray for plane and concave gratings at grazing incidence. Over the visible spectrum, Δλ/λ ~ 10−4 is shown for plane gratings mounted at Littrow and at normal incidence in reflection or transmission.
© 2016 Optical Society of America
1. Introduction
The author has recently presented a class of single-element high-resolution plane grating monochromator having fixed principal ray directions, object distance and image distance [1]. That solution correlates multiple-axis grating rotations to correct the geometrical aberrations over a broad range in scanned wavelength. As an alternative, the present work employs only translation of the grating along its surface, absent any rotations or multiple motions.
Previous geometries employing grating translation are few and limited to motions within a meridional plane normal to parallel grooves. Aspnes [2] proposed a sagittally curved cylindrical grating translating along its axis of symmetry and having an exponential variation in the groove spacings. This is the only fixed geometry design to date, and would exhibit both stigmatism and spectral resolution independent of the scan wavelength. However, the grating manufacture is challenging for use at grazing angles: a) the exponential spacing does not conform to patterns from interference lithography, thus is difficult to provide on the steeply curved surface, and b) spectral focusing to first and second powers determines the translation, requiring the grating ruled width significantly exceed its illuminated portion. Also, cylindrical symmetry constrains the magnification to be sin2 α/sin2β (where α and β are the graze angles of incidence and diffraction), which is undesirably far from unity and leaves insufficient freedom to eliminate third-power “spherical aberration”. Nonetheless, Aspnes’ insight offers maximum image brightness for a monochromator with moderate-high spectral resolution. Ishiguro et. al. [3] employed a large magnitude translation of a grazing incidence concave grating - plane mirror system. This corrects for first-power defocusing as the grating rotates about its pole to scan wavelength, while preserving the (parallel) optical axes entering and exiting the optics pair. Hettrick [4] eliminated the first and second power spectral aberrations by small translation of a varied line-space concave grating along its tangent plane. To scan wavelength, this motion was combined with rotation about an axis fixed in space. April and McCarthy [5] translated a near-normal-incidence plane grating (recorded by two-beam interference) in a direction which numerically minimized the net spectral aberration to third power. This trajectory, almost normal to the surface, changed both the object and image distances while the wavelength was scanned due to the varying angle of diffraction.
In the above four geometries, the critical tasks of scanning wavelength and controlling aberrations depend on the same optical parameters (e.g. the angles of incidence and diffraction, the varied spacing and the translation) specified in the meridional direction. This over-constrains the parameters and limits the spectral resolution. In the present work, these tasks are executed in orthogonal directions, an approach which is anticipated to improve performance and enable a practical design using existing technologies of grating manufacture.
2. Basic scheme
Consider an optical system employing a diffraction grating, wherein each of the macroscopic operational elements are fixed in position, angular orientation and curvature. The only variable in such an otherwise static geometry is the scale of the diffracting structure on the illuminated portion of the grating. In effect, the groove pattern may only uniformly expand or contract, to increase or decrease the scaled wavelength which is diffracted to a given image position. The technology to do this directly (by literally stretching an elastomeric grating [6] or constructing dynamically programmable meta-surfaces) at diffraction-limited accuracy is unlikely to be available soon, particularly for use at short wavelengths. However, a current practical means of effectively producing this result, albeit with residual aberrations, is illustrated in Fig. 1. Physically static grooves curve away from the line along which the grating translates, thereby placing a continuously-scaled groove pattern in the path of a stationary incident beam. This scheme provides the following inherent attributes: 1) the mechanical stability, accuracy and simplicity of rectilinear translation, 2) a fractional spectral resolution (∆λ/λ) which is independent of the wavelength, 3) a fixed principal ray (in-and-out) and 4) fixed angular apertures (in-and-out).
Except when giving numerical examples, dimensionless quantities (in italic font) will typically be used, normalized to the physical object distance r. For in-plane diffraction, the straight central groove is in the sagittal direction (normal to the meridional plane of incidence) and having a length coordinate ℓ. This is the sum of the grating translation (s) and the illuminated aperture coordinate (σ) relative to a fixed point in space (P). By convention, the meridional plane ( in Fig. 2) is designated as “horizontal” and the sagittal plane as “vertical”, unrelated to the orientation of the physical instrument or the reader’s perspective of the figures drawn in this paper. At a given scan wavelength (and thus translation), the horizontal aberration ∆x′ is in the dispersion direction and is proportional to the fractional change in the meridional groove spacing dω over the illuminated aperture ∆σ. Furthermore, the vertical astigmatism aberration ∆z′ is proportional to ∆σ. Therefore, the image tilt angle ψ′ = arctan (∆x′/∆z′) is proportional to [∆dω(ω, ℓ)/dω(ω, ℓ)]/∆σ. To maintain a truly fixed geometry, this tilt angle (e.g. that of an exit slit) may not change as the grating is translated. This requires the 2D spacing function have the property that ∂dω/∂ℓ ∝ dω:
In the semi-classical case where dω(ω, 0) = do (a constant), meridional integration of the density (1/d) is trivial, determining the groove number N = e−cℓ ωr/do. Thus, each groove is curved along the pure exponential ω(N, ℓ) = ecℓ N do/r.
The exponential of Eq. (1) also provides equal changes to log (λ) per unit translation, facilitating a spectral scan over multiple octaves q. The exponential coefficient is:
where S is the total translation. In the above equations, ω ≡ w/r, s ≡ s/r and S ≡ S/r where w, s and S are physical dimensions. The required physical grating size is , in which the sagittal dimension is the sum of the translation and the sagittal illumination (typically ≪ S).2.1 Two-dimensional varied line-space gratings
The aberration correction provided by one-dimensional varied line-space (VLS) gratings has previously been determined by expressing the groove number as a power series in the grating ruled width coordinate [1]. An extension of that formalism to two dimensions is given here as:
where Nij are dimensionless coefficients, Noo is an arbitrary number assigned to the groove passing through the origin and n is the maximum value of the power-sum (≡ i + j − 1). For the present divergent groove grating, expansion of Eq. (1) about the translation coordinate s determines the full array of 2D coefficients over the (narrow) illuminated region of the sagittal aperture, based on the 1D coefficients (Ni) in the meridional plane at s = 0: where N1= 1. Derivatives of Eq. (3) provide the groove density, locally measurable in the two directions and thus convenient for testing and raytracing:where the dimensionless density coefficients are ωVij = (i + 1) Ni+1,j and σVij = (j + 1) Ni,j+1. Because they derive from the same scalar function (N), the “mixed” coefficients (i ≠ 0 and j ≠ 0) are related by σVij = ωVi+1,j−1 (i + 1)/j.As with prior 1D (straight and parallel groove) VLS gratings, any nonzero coefficients Ni (i > 1) cause a meridional variation dω(ω). The 2D VLS spacing function dω,(ω, ℓ) is set by Eqs. (4) and (5) with the substitutions j = 0 and σ = s = ℓ. Meridional integration of 1/d determines the groove number . The corresponding “varied exponential” curve of each groove is determined numerically, however including only the linear VLS coefficient (N2) yields an accurate closed-form (quadratic) approximation:
Expansion of the radical reveals that Eq. (6) reduces to the pure exponential in the limit of either a constant line-spacing (N2 = 0) or as ω → 0 near the central groove (N = 0).
2.2 Grating fabrication
This novel grating design may be manufactured by existing maskless optical lithography systems (e.g. Heidelberg Instruments). These write unconstrained two-dimensional patterns over large apertures (150–1400 mm), such combination being unfeasible by the methods of mechanical ruling, interference lithography (“holography”) or e-beam lithography. Blue lasers at 405 nm deliver a minimum feature size ~ 500 nm, allowing 1/d ≲ 1000 g/mm; while frequency-doubled Argon lasers at 244 nm have produced 3000 g/mm gratings with features below 160 nm using nonlinear resists [7]. Though a tilt stage [8] also accommodates steeply curved (30 mm sag height) non-planar substrates, auto-focus of the piezo-driven objective in real time will follow the shallow meridional curvature required of the gratings proposed here (< 0.1 mm sag height) without compromising the feature size. Plane gratings manufactured by the maskless method may also be employed as masks in projection lithography systems developed for the manufacture of microprocessors. At present, immersion reduction at 193 nm (e.g. Cymer Inc.) delivers feature sizes below 100 nm (1/d ≳ 5000 g/nm) on smaller (32 × 25 mm2) coherent plane apertures.
Optical lithography is not subject to the mechanical stresses of burnishing by a diamond tool, and maskless writing with lasers requires only minutes or hours (compared to days or months for mechanical rulings). This minimizes the groove positioning errors caused by environment variables (e.g. ground vibration, temperature and atmospheric pressure). Thus, given comparable systems for environmental isolation and interferometric control, the optically-written grating will exhibit smaller residual errors in the groove positions. Given a RMS random error ΔwRMS, the fraction of diffracted intensity which is scattered in the dispersion direction as focused stray light (“grass”) is [9]
The present commercially-available maskless pattern generators are equipped with translation and control systems suitable for the fabrication of integrated circuits (e.g. chip masks). Using the finest (10 nm) address grid for these systems, the 3σ specification for uniformity is 60 nm (ΔwRMS ~ 20 nm). From Eq. (7), this will result in FSL ≲ 10−3 for groove spacings d > 4000 nm (1/d < 250 g/mm). Upgrading these sub-systems to the accuracy of ΔwRMS < 1 nm achieved 40 years ago for mechanical ruling engines [10] will reduce the FSL to < 2.5 × 10−6 at 250 g/mm, 4 × 10−5 at 1000 g/mm and 2.5 × 10−4 at 2500 g/mm.
Grooves of rectangular profile (mesa width = a; ravine width = d − a) and constant depth (h) require the fewest post-processing steps. Amplitude diffraction is realized if the ravines are optically inactive. This would occur if the bottom surface is non-reflective or (at grazing angles and short wavelengths) it is shadowed by the adjacent walls [h/d > (1 − a/d)/(1/tanα + 1/tanβ). Using Kirchoff theory, the m’th order relative diffraction efficiency is then
In maskless lithography, both d and a may be specified as arbitrary functions of ω and ℓ. For example, if a/d is constant with ℓ, then εm is independent of the scan wavelength (as the grating translates by an amount s along axis ℓ, one has a ∝ d ∝ λ ∝ ecs). In the case of a d = 0.5, the nonzero even orders are eliminated and the odd orders provide εm = 1/(mπ)2 = 0.101 at m = ±1 and 0.011 at m = ±3. Note that Eq. (8) is also the efficiency of a transmission grating having gaps of width a between opaque bars of width d − a.
A more advanced option is to provide, independent of the scan wavelength, substantially higher first-order efficiency (ε1 ≳ 0.4) by phase diffraction. This requires groove profiles of fixed aspect (h/d) along their lengths, representing an extension of isotropy to the third dimension. Varying the chemical etching time along parallel grooves (constant d) has previously provided a factor four continuous variation in the depth of a laminar profile [11]. However, the present divergent grooves require only a fixed h/d, which may be more accurately and readily formed as triangular profiles of constant blaze angle (e.g. at the lattice planes of a crystalline substrate). A further simplification is that the new scan method already translates the grating along the center groove, so no additional motion is needed.
3. Light-path analysis of a self-focused grating
The simplicity and high reflection efficiency of a single-element system are of particular advantages at short wavelengths and grazing incidence. In this section, geometrical parameters (including the 2D VLS coefficients) will be used to analytically express the aberrations for a monochromator based on a self-focusing divergent groove grating (Fig. 2). Detailed results for the spectral resolution of a plane grating will be given in Secs. 4 (grazing) and 6 (non-grazing). Sec. 5 will treat a concave grating surface.
The horizontal and vertical lateral ray coordinates at the Gaussian image plane are expanded as a sum of powers in grating aperture coordinates and .
The principal ray horizontal image coordinate is fixed by the grating equation:
where ρ ≡ sinβ/sinα. In this work, a small-angle approximation is preceded by the “≃”sign and is valid when the effective reflection graze angle γ = (α + β)/2 is small.The principal ray vertical image coordinate obeys the law of sagittal reflection, expressed here as , where η ≡ r′/r is the image/object distance ratio for the principal ray.
3.1 General equations of a rigorous expansion method
The horizontal (ijx′) and vertical (ijz′) coefficients of the aberrant terms (i + j − 1 > 0) are provided by a mathematically consistent theory of light-path expansion [1, 12]:
for which a circular cylinder (meridional concave radius = R), correct to power ω6, yields: where b = [(ω/R + (ω/R)3/2 + 3(ω/R)5/8] is the surface slope at ω. For a plane grating, the last product in the numerators of Eqs. (12) and (14) simplifies to ω. The Taylor series for is 1 − t/2 + 3t2/8 − 5t3/16 + 35t4/128 + ⋯. Correct to power ω7, the lateral squared distances (t) are: where ∪ = ω[(ω/R)/2 + (ω/R)3/8 + (ω/R)5/16] is the depth of the pole relative to the surface at ±ω. For a plane grating, the last product in Eqs. (16) and (17) simplifies to ω2. The reference image coordinates ξij and ζij include all (paraxial and non-paraxial) power terms except those of the subject aberration ( and ):Numerical raytracings show that a sphere results in negligible change to the grating aberrations at grazing angles, however the above light-path equations employ a meridional cylinder on which the central groove remains straight in the translation (sagittal) direction.
3.2 Explicit expansion terms for a point object
Given xij = zij = 0 (a point object in the meridional plane), consider the 18 coefficients (ijx′ and ijz′) for which i + j − 1 = 1, 2 or 3. Due to the in-plane geometry of the present grating design, 0jz′ = 0 for all odd values of j, including 03z′. Also, numerical raytrace extractions determined the coefficients 22x′, 13z′, 13x′ and 04z′ to be negligible (~10−8). Equations (10)–(18) were used to derive the remaining 13 coefficients, employing substitute variables ijT ≡ ijx′ sinβ/(μ η), τ ≡ cosβ, Γ ≡ cosα/cosβ, Q ≡ μ/sin2β and κ ≡ Rsinβ. Note that τ ≃ Γ ≃ 1 at grazing incidence (α, β ≪ 1) and Q is of order unity at all angles:
These horizontal (x′) and vertical (z′) image plane positions of a monochromatic ray are projected to the spectral direction (x″), oriented at an angle ψ′ from the horizontal (see Secs. 3.3 and 3.4). The modulus of the full variation in this direction, as the rays wander over the rectangular illuminated grating aperture , is the extremum (Δ) aberration:
where pij = 21−(i+j) if i is odd and j is even (otherwise pij = 22−(i+j)). Differentiation of Eq. (9) converts this spatial aberration to wavelength at the principal ray position (x″ = 0):In the perpendicular direction, the extremum aberration is:
3.3 Horizontal image tilt (sagittally induced)
The lowest-degree mixed horizontal aberration created by the exponential groove pattern is , resulting from the nonzero value of N11 = −c in Eq. (4). Due to the linearity of with the sagittal pupil coordinate σ, this is simply a tilt of the astigmatic image by the angle relative to the horizontal. From Eqs. (19) and (20):
where a positive value is counter-clockwise for an upstream observer. Therefore, the projection of these two aberrations in the spectral direction is made zero by use of an exit slit oriented at the fixed angle . In Fig. 2, a vertical sheet of incident rays (magenta) strikes the far edge of the illuminated rectangular aperture, at meridional coordinat . Upon diffraction (lighter magenta), this sheet of rays rotates clockwise to the angle ψ′, measured within the exit plane.3.4 Vertical image tilt (meridionally-induced)
The lowest-degree mixed vertical aberration would extend a perfect horizontal focus for meridional rays into a vertical line. This would be viewed by the tilted slit as a defocus normal to its length (multiplied by ). In effect, the vertical aberration from the meridional rays induces a second tilt angle . From Eqs. (21) and (22):
The solution is to purposely “defocus” the grating horizontally to force equality of the two tilts, as geometrically illustrated in Fig. 2. A horizontal sheet of incident rays (orange) strikes the grating at sagittal coordinate (left edge of the illuminated rectangular aperture). The diffracted sheet of rays (yellow) rotates counter-clockwise to the same tilt angle as given in Sec. 3.3 for the vertical (magenta) rays. Algebraically, this requires a simultaneous solution of Eqs. (35) and (36), yielding exact cancellation of spectral defocusing (0 = x″20 = x′20cosψ′ − z′11sinψ′). This is provided by the first-power VLS coefficient:
where the term in c2μ is effectively a correction for the sagittally-induced tilt of Eq. (35). Likewise, canceling the next higher-power aberration of “meridional coma” in the spectral direction determines the second-power VLS coefficient:Cancelation of “spherical aberration” in the spectral direction determines the third-power VLS coefficient:
Numerical raytracings confirm that Eqs. (37)–(39) result in removal of the , and lateral ray aberrations (numerical residuals < 10−13 radians). This cancelation may continue for indefinitely high powers of the pure meridional aberrations, to obtain for all i. In general, if , the following condition is noted from Eqs. (32) and (35):
4. Residual spectral aberrations of a grazing incidence plane grating
4.1 Bowtie aberration (Δλ21)
The above removal of Δλ11 and Δλi0 leaves Δλ21 as the dominant spectral aberration. Using Eqs. (23), (24), (32) and (33), and substituting , and c = q/S ln2, this aberration is expressed in parameters related to performance and construction, including the solid angle of collection (Ω = ϕmϕs) from the object:
A planar surface K = ∞ simplifies the grating fabrication and exhibits a focal length (η) which is nearly independent of γ ≲ 15° at fixed ρ [1, Secs. 2.4 and 8.8.5]. At these grazing angles, the approximations τ ≃ Γ ≃ 1, μ ≪ 1, Q ≃ (1/ρ2 − 1)/2, N20 ≃ (ρ2/η + 1)/(ρ2−1) and sinα ≃ α ≃ 2γ/(ρ+1) further simplify Eq. (41):
where M = η/ρ converts η to a performance parameter (the grating horizontal magnification) and L = (1 + η) r is the object-to-image path-length. Note that ρ, geometrically defined as sinβ/sinα [e.g. see Eq. (9)], is also the relative diffraction efficiency (or its reciprocal if ρ > 1) of a blazed triangular groove. The “sep” subscript denotes this is a measure of marginal resolution, defined as the center-to-center separation of a doublet whose scan profile exhibits a dip of 25%. This requires a form factor f21, which depends on the horizontal aberration (i = 2, j = 1) and the width of the exit slit. Numerical raytracings find f21~1/3 if the exit slit width transmits half of the intensity at either peak.Consider a small monochromator (L = 500 mm, S = 40 mm) scanning q = 3 octaves in the soft x-ray at γ = 3° and Ω = 8 × 10−6. The solid curves in Fig. 3 plot Eq. (42) at different magnifications (M) and as a continuous function of ρ. The aberration is near local minima at ρ ~ 1/2 and ρ ~ 2, where numerical raytrace extraction data (Sec. 4.2) are plotted at several magnifications, providing independent confirmation of the light-path calculations to within the software accuracy (64-bit). At M = 1 and ρ = 2, (Δλ21/λ)sep ~ 10−3, with other parameters as follows: η = 2, α = 2°, β = 4°, m = −1, μ = 0.0018283, r = 166.67 mm, r′ = 333.33 mm, c = 8.664 and ψ′ = −8.607°.
4.2 Numerical raytracings of a compact soft x-ray monochromator
The above light-path equations have been verified by the independent calculations of numerical raytracing, employing a recent 2D enhancement of the VLS coefficients included in the open source code BEAM4 [13]. In particular, the basis Eqs. (19)–(31) agree exactly with the results of the numerical raytrace extraction method given by equations (75)–(83) and (86) in a recent paper [1]. The discrepancies are < 10−13 radians, corresponding to a lateral image size error of < 0.001 nm at a grating focal length of 10 meters. This sub-atomic scale is many orders of magnitude smaller than the diffraction limit of soft x-rays.
In the case of the mixed aberrations given by Eqs. (23), (24), (30) and (31), the 21-ray extraction method (see Sec. 7.3 in [1]) initially confirmed their accuracy to 10−11~10−12 radians. This was improved to 10−13~10−14 radians (comparable to the 64-bit floating point accuracy of the software) by adding the following rays:, , and . Using this enhanced (25-ray) extraction, the equations
for the 2nd power mixed terms remove contamination by the 6th power mixed terms, and for the 3rd power mixed terms remove contamination by the 7th power mixed terms.If the small monochromator described in Sec. 4.1 has a square grating format (aspect g = 1) of 40 mm × 40 mm, then ϕm = 8.4 mrad and ϕs = 0.96 mrad. However, the solid angle and hence the spectral resoluton are unchanged if ϕm is decreased and ϕs increased in proportion. For example, ϕm = 2.83 mrad and ϕs = 2.83 mrad reduces the grating meridional dimension to 13.5 mm (g = 0.3375). Though the vertical astigmatism triples, this does not significantly increase the overall image length, due to the aberration component having been comparable and is now reduced by a factor of 3. In fact, the 2D image brightness increases downstream of the exit slit [Fig. 4(c)], due to the reduced angular aperture in the horizontal direction. It is noted that the fewer number of grooves illuminated is not an issue for this moderate resolution design, as only a few thousand are required on the basis of physical diffraction (see Eq. (66)). An elliptical iris (13.5 mm × 0.47 mm ≃ 2.83 mrad × 2.83 mrad) is used to remove the most offensive (21.5%) rays at the corners of the rectangular region shown in Fig. 2. Figure 4(a) shows the 2D spot diagram (on the exit plane) of a point object whose spectral output consists of a line doublet. Figure 4(b) is a wavelength scan, where the indicated intensity values are the geometrically-transmitted fraction (through iris and exit slit) of the photons emitted by the object over a solid angle of 8 × 10−6 sr at each of the two (equal strength) lines. The 26.5% dip in the transmitted intensity between these lines is slightly better than that of the standard Rayleigh criterion, and confirms the marginal resolution of (Δλ21/λ)sep = 10−3 plotted in Fig. 3 from Eq. (42).
For a minimum physical wavelength of 1 nm, the groove density is 1828/mm near one end of the sagittal dimension and decreases by the required 3 octaves (factor of 8) to 228/mm over the translation range of 40 mm (for a maximum wavelength of 8 nm). As there need not be coherence over more than the illuminated groove length (0.472 mm), the grating may be assembled from three square segments. each 13.5 mm × 13.5 mm and covering one octave. Alternatively, an array of 7 such sub-gratings can scan the entire grazing incidence region extending from ~ 0.75 nm in the deep soft x-ray to 96 nm at the border of the far ultraviolet.
Doubling S/L (to 1/6.25) in Eq. (42) would halve the limiting aberration (to 1/2000). The slit widths required by Eq. (65) would remain unchanged if the length of the monochromator was also doubled (to L = 1000 mm). The grating size would increase to 27 mm × 160 mm (or five individually coherent segments produced by DUV projection lithography, each 27 mm × 32 mm) with an elliptical illumination of 27 mm × 0.94 mm. At the same grating format, illumination and slit widths, each subsequent 2-fold improvement in spectral resolution requires a 2-fold increase in the length and a 4-fold decrease in the solid collection angle.
5. Residual spectral aberrations of a grazing incidence concave grating
5.1 Cancelation of the bowtie aberration (Δλ21)
The dominant aberration of the above plane grating may be eliminated by use of a meridionally-concave surface (e.g. a cylinder or sphere). The required curvature is obtained by setting Eq. (41) to zero:
which is verified by numerical raytracings to cancel Δλ21 (within the 64-bit calculation accuracy of ~10−13 radians). Because Δλ20, Δλ30 and Δλ40 were already eliminated (Sec. 3.4), the only spectral aberrations remaining to a power-sum ≤ 3 are Δλ12 and Δλ31.In the case of unit magnification (η = ρ), the small-angle approximation reveals κ ≃ η/2 at ρ = 2 (inside order) and κ ≃ η at ρ = 1/2 (outside order). Thus, the outside order results in twice the radius of curvature, of practical benefit in forming the grating groove profile. The condition κ = η = ρ corresponds to horizontal focusing on the Rowland circle [14], for which Eqs. (37) and (38) confirm the VLS coefficients N20 and N30 are nearly zero. Their residuals (∝ μ) arise from the nonzero values of the vertical aberrations 11z′ and 21z′ in Eqs. (21) and (25). From Eq. (32), these project in the direction of the tilted image width, thus contributing to 20x″ and 30x″, albeit an increasingly small amount at grazing angles. This divergent groove grating may replace the classical one in a Rowland circle spectrometer, resulting in wavelength scanning by sagittal translation of the grating. In this configuration, the exit slit would be fixed (though tilted relative to the classical case) and the object would also be fixed (though not be an extended slit, per the discussion in Sec. 7.2).
5.2 Astigmatic curvature aberration (Δλ12) and exit slit curvature
Equations (27), (32) and (33) determine this 2nd power extremum wavelength aberration:
which is in exact agreement (~10−13 radians) with numerical raytrace extractions given by Eqs. (80) and (86) of a recent paper [1]. The small angle approximation yields:The solid collection angle (Ω), the graze angle (γ) and the angle ratio (ρ) affect the throughput, while the grating dimensions S and gS impact the manufacturing cost. The form factor factor f12 converts the extremum aberration to the marginal resolution of a scanned line doublet. As for the f21 aberration (also of power-sum = 2) discussed in Sec. 4, raytrace scans determined f12 ~ 1/3 when using an exit slit width which transmits 50% of the diffracted energy at the line peak. The dashed curves in Fig. 3 plot Eq. (49) using g = 1/4 and the remaining parameters as previously given. This aberration is seen to be considerably smaller (factors of 10) than the dominant term (Δλ21/λ)sep for the plane grating.
Use of a curved exit slit to match the shape of this aberration provides additional improvement. As derived in Fig. 5(a), the nominal radius of this slit is fixed by the astigmatism and astigmatic curvature terms:
where the last expression omits the μ term in the numerator and is accurate at grazing incidence. The image curvature radius shown in the spot diagram of Fig. 5(b) is rRex ≃ 1615 mm, with other parameters as follows: sin γ = 0.052, M = 1, η = 2, ρ = 2, β = 4°, L/S = 12.5, c = 8.664, m = −1, μ = .0018283, ψ′ = −8.607°, r = 3m, r′ = 6m.Equations (19)–(22) and (34) determine . If this ratio is less than , the slit width (Δx″) required to include the extremum aberration [Eq. (48)] is less than , as derived from Fig. 5(a):
In Fig. 5(b), a grating aperture aspectg = 0.25, scan range of q = 3 octaves and solid angle of Ω = 8 × 10−6 provide , resulting in Δx″ = 0.316 . For marginal line separation, this is multiplied by the form factor of f12 to yield a slit width of 0.0015 mm, thereby narrowing the transmitted wavelength profile from (Δλ12/λ)sep = 2.85 × 10−5 (straight slit) to 0.9 × 10−5. In the line profile (1D) scan of Fig. 5(c), use of a wider slit (0.003 mm) reveals a net resolution of ~ 2.5 × 10−5, limited by the higher-power term Δλ31 (Sec. 5.3). It is noted that a low value of g not only allows the curved slit to compensate for the increased value of Δλ12 in Eq. (49), but also reduces the Δλ31 aberration in Eq. (53) and provides for a significantly smaller grating. The optimum value for g is set by requiring the ruled width (gS) encompass a sufficient number of grooves to achieve the desired diffraction-limited resolution, as determined by Eq. (66).
The above choice of unit magnification has the advantages of both minimizing the system length (given a minimum practical slit width) and maintaining the horizontal angular aperture upon diffraction. The resolving power of 40000 shown in Fig. 5(c) compares well with the value of 25000 calculated for a recently-proposed single-element plane grating design “SEPGM” [1] otherwise having the same nominal performance (2γ = 6°, Ω = 8 × 10−6 sr, ρ = 2 and q = 3). Advantages of the present design are: 1) The mechanically simple and robust scanning motion of rectilinear translation of the grating, contrasting with the multiple correlated rotations of the grating and the slits required with the previous design [1]; 2) At unit magnification (M = 1), the system length L = (1 + Mρ) r is only ~ 0.64 times that required for the previous design (where 〈M〉 = η/〈ρ〉 = 3.7 / 2 = 1.85). This results in a 9 m length, compared to 14.1 m for the SEPGM, given the same object distance (r = 3 m) and exit slit width (0.005 mm would yield a resolving power of ~ 30,000 for the present design and a point object). 3) The potential to maintain peak blaze efficiency at all scanned wavelengths (Sec. 2.2). Disadvantages of the present design are: 1) The need for a meridionally-curved (non-planar) surface; 2) A larger grating (180 mm × 720 mm vs. 214 mm diameter), albeit the long dimension is along the groove lengths and may thus be constructed from an (incoherent) linear array of smaller gratings (e.g. 4 gratings, each 180 mm square); 3) A curved groove pattern; and 4) Restriction to use with a point object, rather than an entrance slit extending out of the meridional plane.
5.3 High-power mixed aberration (Δλ31) and a “sweet magnification”
Equations (30), (31), (32) and (33) determine this 3rd power extremum wavelength aberration. Employing Eq. (42) and neglecting the higher-power terms in Eqs. (30) and (31):
Further simplification uses the previous substitutions for c, Q, and , and the following: 21T = (12S) (sinβ) tan ψ′ ≃ c3μ/(1 + 1/η)− c, Eq. (37) for N20, Eq. (38) for N30, and Eq. (47) for 1/κ, where small-angle approximations are also employed in all terms:
and where raytracings find f31 ~ 1/5. Though Eqs. (52) and (53) are only approximations (~ 1% accuracy), the full and exact Fermat expansion term resulting from Eqs. (30)–(31) agrees precisely (~10−13 radians) with the raytraced numerical extraction Eqs. (45)–(46) of Sec. 4.2 The dominant bracketed [ ] term in Eq. (53) vanishes if the magnification is chosen to becorresponding to η = 1/ρ. Figure 3 plots two examples: Msweet = 1/4, showing the cancelation of Δλ31 for an inside spectral order (near ρ = 2); and Msweet = 4, showing the cancelation for an outside spectral order (near ρ = 1/2). Figure 5(d) shows the numerical raytrace employing the following parameters: sinγ = 0.052, M = 4, η = 2, ρ = 1/2, β = 2°, L/S = 12.5, c = 2.888, m = +1, μ = .0018283, ψ′ = +5.762°, r = 8.33m and r’ =16.67 m. As Eq. (49) reveals that Δλ12 scales with q2, reducing the coverage to 1 octave and use of a curved exit slit of width 0.002 mm encloses the extremum aberration at Δλ12/λ ~ 2.2 × 10−6. Such a narrow exit slit and long focal length corresponds to an angular uncertainty of only ~ 10−7 radians. While the rigidity of a preloaded rectilinear-only grating stage maximizes the inherent mechanical stability, residual environmentally-induced changes to the angles of incidence and diffraction would likely require optical monitoring and mechanical corrections in real time.6. Aberration correction of a plane grating in non-grazing mounts
6.1 Normal incidence reflection or transmission
Figure 6 illustrates a reflection or transmission grating illuminated at normal incidence. Setting α = π/2, β = π/2 − β*, Γ = 0, τ = sinβ*, μ = sinβ* and ρ = cosβ*, where β* is the angle of diffraction relative to the grating normal, Eq. (41) becomes:
in which tan ψ′ = μc/(1 + 1/η)/cosβ* = qη/(η + 1)2(L/S)(ln2) tanβ*. Furthermore, η = Mcosβ*, where M is the horizontal magnification, thus the term in brackets | | becomes a quartic in cosβ* whose root yields the condition whereby Δλ21/λ vanishes for any given value of M. For example, consider a small grating (S = 32 mm), a short path length from object to image (L = 90 mm) and a scan over the visible spectrum (400–700 nm, thus q = log2(7/4) ≈ 0.807). Numerical iteration of the quartic at a magnification of unity yields the aberration-removal solution of cosβ* = 0.08367 (β* = 85.2° and η = 0.08367); at a magnification of 1.42 it yields cosβ* = 0.256 (β* = 75.2° and η = 0.369); and at a magnification of 2 yields cosβ* = 0.344 (β* = 69.9° and η = 0.688).The other mixed term of power-sum = 2 is sagittal coma, obtained from Eq. (48):
yielding Δλ12/λ ≈ 0.0026 for a square aperture of ϕm = ϕs = 0.05. However, the aberration is reduced by a factor of 16 (at the same solid aperture) by increasing ϕm to 0.20 and reducing ϕs to 0.0125. The resulting Δλ12/λ = 0.000163 is verified by extracted from the (independent) raytrace spot diagram inset to Fig. 6, and use of Eq. (33). Also extracted is , in agreement with Eqs. (30)–(32). Multiplying these extrema values by f12 = 1/3 and f31 = 1/5 yields the marginal aberrations. The net resolution then includes the contributions from Eq. (65) for dispersion (assuming 0.004 mm widths for both the object and the exit slit) and Eq. (66) for physical diffraction from 14400 grooves (at the low-density end of 1341 g/mm), yielding (Δλ/λ)sep ≈ 1/6000.6.2 Littrow (retroreflection)
A self-focusing plane reflection grating mounted in or near Littrow (Fig. 7) and capable of wavelength scanning on-blaze provides an efficient and compact monochromator for the UV, visible or near IR regions. Setting α = π − β, Γ = −1, μ = 2 sinβ* and ρ = 1 in Eq. (41):
in which tan ψ′ = qη/(η + 1)2(L/S) (2 ln2) tanβ*. The term in brackets | | is thereby a quadratic in β* whose root yields the condition whereby this aberration vanishes. To horizontally retro-focus the object (at a magnification of −1), one chooses η = 1. To facilitate construction of a monochromator, the object (or entrance slit) and image (or exit slit) may be displaced in x (a “near-Littrow” configuration). In a very compact configuration, L = 90 mm (r′ = r = 45 mm), Ω = ϕm ϕs = 0.0025 and q = 0.807, the above solution for δλ21/λ = 0 is β* = 49.696°.The remaining aberration is dominated by Eq. (27):
Given ϕs = 0.0125 and ϕm = 0.200, Eq. (58) provides Δλ12/λ = 0.000065, verified by extracted from the raytrace inset to Fig. 7. Also extracted is , in agreement with Eqs. (30)–(32). Multiplying these extrema values by f12 = 1/3 and f31 = 1/5 yields the marginal aberrations. The net resolution then includes the contributions from Eq. (65) for dispersion (assuming 0.004 mm widths for both the object and the exit slit) and Eq. (66) for physical diffraction from 31000 grooves (at the low-density end of 2179 g/mm), yielding (Δλ/λ)sep ≈ 1/14000. It is noteworthy that this ultra-high resolution single-element scanning system is only 45 mm in length. Given the object distance is r = L/(1 + η) and the image distance is r′ = L/(1 + 1/η), the monochromator length along its y-axis in Figs. 6 and 7 is between L and L/2, depending upon η and β.
Equations (55)–(58) have been verified by extractions of these power terms from numerical raytracings. Similar resolution may be obtained for other high angular deviation plane grating mountings (e.g. for “near-normal” monochromators). This is due to the ability to balance the component terms of Δλ21 in different powers of μ [e.g. see Eq. (55)], where μ = sinβ*. However, such balancing is not possible at a low angular deviation 2γ (grazing incidence) where μ ≪ 1 per Eq. (9). Due to the very high meridional aperture (0.20 radians), the spot diagrams in Figs. 6 and 7 contain a small amount (0.002 mm) of . This high-power S-shaped aberration may be easily corrected by a nonzero value of N60, though not accomodated in the current raytrace routine. However, a nonzero value of N50 was used to remove the high-power term, which would have otherwise added a U-shaped aberration of magnitude 0.002 mm in Fig. 6 and 0.006 mm in Fig. 7.
7. Inclusion of an entrance slit
7.1 Tilt angle and curvature: principal ray terms
In Eqs. (35) and (36), the tilt angle ψ′ of the exit slit was determined by the paraxial focusing of sagittal and meridional rays originating from a single object point in the horizontal (meridional) plane. In Fig. 2, this “in-plane” point is designated Oo and defines the origin (0,0). Now consider an infinitesimal width entrance slit centered on this point, but extending a finite distance (z) out of the horizontal plane. To avoid image broadening due solely to the off-plane position (x, z) of a self-luminous object point (O) along this slit, the (principal) ray from this point to the grating pole must be reflected to an image point pI which also lies along the in-plane image orientation given above. Figure 2 illustrates a general object point (open circle) and its principal ray (dashed). The algebraic solution constrains not only the entrance slit tilt angle (ψ), but also its curvature radius (Ren) and higher derivatives of a polynomial function specifying the horizontal coordinate x =∑ zk (1/k!) ∂x/∂kzk, where ∂x/∂z = tan ψ and ∂2x/∂z2 = 1/Ren. Using the general equation of a recent paper [1, Eq. (11)] and substituting θ = 0 for the present grating mount:
The coefficient of the next higher degree (∂3x/∂z3) was obtained by numerical optimization. As the independent variables (ψ′, ρ, c and γ) are fixed, the entrance slit tilt angle and its shape coefficients are also fixed. For example, given the parameters used for the Fig. 4 raytrace (γ = 3°, ρ= 2, η = 2, and r = 166.67 mm), the above equations determine tan ψ = −2 tan ψ′ = 0.3027 and Ren = 78.9 mm, confirmed by the numerical raytracing shown in Fig. 8. This curvature corresponds to a deviation of 0.16 mm at the ends of a 10 mm long slit. Such a correction is of practical advantage only if it is larger than the grating aberrations, including those resulting from object points along this tilted slit. It is also noted that, instead of curving the entrance slit by Ren, the exit slit could be curved to a radius of Rex = ηρRen/cos ψ, equal to 329.7 mm using the above parameters. However, Fig. 8 shows this induces significantly larger off-axis spectral aberrations.
7.2 Off-axis grating spectral aberrations
The object points along a finite length entrance slit deviate from the on-axis (0,0) point in both the vertical and horizontal directions. This induces additional wavelength aberrations:
These terms arise mainly from the slit tilt of Eq. (59), which results in a nonzero horizontal coordinate x = z tan ψ. This changes the meridional angle incident to the grating and thus the image tilt (i = 1, j = 1), the focal length (i = 2 and j = 0) and higher-power terms (not derived here by expansion of the light-path, but quantified by extraction from numerical raytracings). For a plane grating, the dominant term in Eq. (61) is image tilt, having a total power-sum ≡ i + j − 1 + k = 2. Expansion of the light-path (Sec. 3.1) yields:
which is valid for both the plane and meridional-only curved (cylindrical) gratings.The other lateral aberration of power-sum = 2 is the defocus term, whose expansion coefficient for a plane grating (e.g. as employed in Sec. 4) is:
If (giving equal horizontal and vertical angular apertures in the diffracted beam), Δλ201/Δλ111~ tan ψ′ for a plane grating. However, for the concave grating (Sec. 5), Fermat expansion and raytracing reveals a substantial increase in the defocusing term:
This is due to the tilt of the entrance slit, which changes the meridional graze angle and thus the focal length of the grating. From Eqs. (61) and (64), the maximum length entrance slit which may be used while preserving the resolution of 1/40000 for the unit magnification concave grating monochromator (Sec. 5) is only 2žr = 0.072 mm. This restricts the ultra-high resolution, concave divergent groove monochromator to use with a pointlike object.
8. Physical considerations
8.1 Linear dispersion
Differentiation of Eq. (9) yields the contribution to the fractional FWHM spectral resolution from an exit slit of effective width :
where D = (1 + 1/η)/tanβ* at normal incidence (α = π/2), D = (1 + 1/η)/(2 tanβ*) at Littrow (α = π − β) and D ≃ | (Mρ + 1)/(ρ − 1) |/sinγ at grazing incidence (γ ≪ 1). At grazing incidence, the 1/|ρ − 1| dependence encourages a larger value of ρ (or 1/ρ in the case of an outside spectral order), apart from any additional advantage this provides in decreasing the contribution from geometrical aberrations (see Fig. 3). The unit magnification design of Sec. 5.2 (ρ = 2 and γ = 3°) requires (Δx″ = 0.003 mm, Δx = 0) to provide (Δλ/λ)dispersion ~ 2.0 × 10−5 for a monochromator length of L = 9 m. For the most compact (L = 0.5 m) plane grating design proposed in Sec. 4.2, entrance and exit slits each of width 0.004 mm yield (Δλ/λ)dispersion = 0.7 × 10−3.8.2 Diffraction limit
As the concave grating monochromator (Sec. 5) cancels nearly all spectral aberrations, the contribution from physical diffraction is non-negligible. The Rayleigh criterion specifies:
Comparison with Eq. (65) reveals that . Given M = 1 and ϕm ≃ 0.002, this ratio is unity at the longest scan wavelength (λ = 8 nm) for . At γ = 3° and ρ = 2, Eq. (66) yields (Δλ/λ)diffraction = 2.5 × 10−5 for the large (L = 9 m) concave grating monochromator of Sec. 5.2 (ruled width = 180 mm). For the most compact plane grating monochromator of Sec. 4.2, (Δλ/λ)diffraction = 4.5 × 10−4 (ruled width = 13.5 mm).
Finally, consider the very large (L = 25 m, ruled width = 500 mm) nearly aberration-free monochomator proposed in Sec. 5.3. With M = 4, ρ = 0.5 and ϕm = 0.004, Eq. (66) yields (Δλ/λ)diffraction ≃ 2.2 × 10−6 at 2 nm (q = 1), equal to the contribution from Eq. (65) due to a 0.002 mm wide exit slit. The sum of these two effects thereby yields the resolution of (Δλ/λ)sep ≃ 3.3 × 10−6 shown in Fig. 5(e).
9. Conclusions
A new geometry has been proposed, whereby simple translation of a curved groove pattern along its central straight groove provides wavelength scanning within a fixed optical geometry. As a first result, a single-element monochromator employing a self-focusing such grating has been introduced and its imaging performance analyzed in detail. The spectral aberrations in pure meridional powers of the grating aperture were shown to vanish at all scanned wavelengths, at all angles of incidence (normal to grazing) and in either transmission or reflection. Additionally, the following parameters were also constrained to be independent of the scanned wavelength: the residual aberration, the image and object tilt angles, the beam aperture and (with advanced etching of the groove profiles) the relative diffraction efficiency.
A table-top (0.5 – 1 m in length) grazing incidence plane grating configuration of this monochromator scans 3 octaves in the soft x-ray (e.g. λ =1 – 8 nm) at unit magnification, a resolving power of 1000–2000 and a collection angle of 8 × 10−6 sr. Without degrading the resolution, the throughput of spatially extended sources may be significantly increased by use of a long entrance slit (fixed curvature) while maintaining a straight exit slit. A large (9 m) version employs a concave grating surface to eliminate the dominant geometrical aberration, providing λ/Δλ ~ 4 × 104, but is restricted to use with a pointlike object. At a “sweet magnification” of 4, the scanned spectrum is nearly free of geometrical aberrations, enabling diffraction-limited resolutions of λ/Δλ ~ 3 × 105 over one octave in wavelength. This ultimate design comprises > 105 spectral bins per grating, however the requirements on exit slit width (0.002 mm) and system length (25 m) raise issues of stability. At normal incidence (transmission or reflection) or in Littrow (retro-reflection), the dominant plane grating aberration may be removed by judicious choice of the diffracted angle. This enables high spectral resolution (λ/Δλ ~ 104) while maintaining the practical advantage of a plane surface for miniature-sized monochromators for applications in the UV to the infrared.
The above gratings may be written by existing maskless lithography systems using visible or UV lasers at exposed feature sizes of ~ 500 nm to 200 nm (1/d ≲ 2500 g/mm), with a low integrated level (10−3) of stray light expected for 1/d ≲ 250 g/mm. For higher densities, the groove placement accuracy should be improved by upgrading the environmental and interferometer control sub-systems in these commercial pattern writers to those used for mechanical ruling engines. For plane gratings, commercial DUV projection lithography can reduce a transmissive mask made by the above process, providing a (smaller) grating having 4 times the groove density (1/d ≲ 10000 g/mm) at the same stray light level as the master.
The proposed new method of scanning wavelength is not restricted to a particular focusing condition or geometry. Though this introductory paper has presented in detail the example of a single-element astigmatic in-plane grating mount, several other applications of this method are under study. These include adding wavelength tunability to prior fixed-grating spectrograph geometries, multi-element (e.g. stigmatic and common path-length) systems and an extreme off-plane grating mount, where the central groove is oriented meridionally.
Acknowledgments
The author thanks Dr. Michael Lampton for upgrading his open-source numerical raytrace code (BEAM4) to include two-dimensional VLS density coefficients (Vij), as required for the curved groove grating introduced in this paper. For some years, BEAM4 has been an invaluable “gold standard” to confirm the accuracy of this author’s light-path formulations.
References and links
1. M. C. Hettrick, “A Single-Element Plane Grating Monochromator,” Photonics 3(1), 1–44 (2016). [CrossRef] [PubMed]
2. D. E. Aspnes, “High-efficiency concave-grating monochromator with wavelength-independent focusing characteristics,” J. Opt. Soc. Am. 72(8), 1056–1061 (1982). [CrossRef]
3. E. Ishiguro, M. Suzui, J. Yamazaki, E. Nakamura, K. Sakai, O. Matsudo, N. Mizutani, K. Fukui, and M. Watanabe, “Constant deviation monochromator for the range 100 A ≤ λ ≤ 1000 A,” Rev. Sci. Instrum. 60(7), 2105–2108 (1989). [CrossRef]
4. M. C. Hettrick, “In-focus monochromator: theory and experiment of a new grazing incidence mounting,” Appl. Opt. 29(31), 4531–4535 (1990). [CrossRef] [PubMed]
5. A. April and N. McCarthy, “Control of spectral aberrations in a monochromator using a plane holographic chirped grating,” Appl. Opt. 47(15), 2750–2759 (2008). [CrossRef] [PubMed]
6. A. N. Simonov, S. Grabarnik, and G. Vdovin, “Stretchable diffraction gratings for spectrometry,” Opt. Express 15(15), 9784–9792 (2007). [CrossRef] [PubMed]
7. T. Glaser, S. Schröter, S. Fehling, R. Pöhlmann, and M. Vlček, “Nanostructuring of organic and chalcogenide resists by direct DUV laser beam writing,” Electron. Lett. 40(3), 176–177 (2004). [CrossRef]
8. D. Radtke and U. D. Zeitner, “Laser-lithography on non-planar surfaces,” Opt. Express 15(3), 1167–1174 (2007). [CrossRef] [PubMed]
9. E. W. Palmer, M. C. Hutley, A. Franks, J. F. Verrill, and B. Gale, “Diffraction gratings,” Rep. Prog. Phys. 38(8), 975–1048 (1975). [CrossRef]
10. T. Harada, S. Moriyama, T. Kita, and Y. Kondo, “Development of numerical control ruling engine for stigmatic concave grating,” J. Jpn. Soc. Precision Eng. 42(501), 888–892 (1976). [CrossRef]
11. F. Polack, B. Lagarde, M. Idir, A. L. Cloup, and E. Jourdain, ”Variable groove depth gratings and their applications in soft x-ray monochromators,”Am. Inst. Phys., Conf. Proc. 879, 639–643 (2007).
12. M. C. Hettrick, “Corrections to ‘A Single-Element Plane Grating Monochromator’,” http://www.hettrickscientific.com (2016).
13. M. Lampton, “BEAM FOUR: Optical ray tracer, v. 194,” http://www.stellarsoftware.com (priv. comm., 2016).
14. H. A. Rowland, “On concave gratings for optical purposes,” Philos. Mag. 16(99), 197–210 (1883). [CrossRef]