Extreme ultraviolet spectrometer based on a transmission electron microscopy grid

Extreme ultraviolet (EUV) light based on high harmonic generation (HHG) [1–3] is finding an increasing array of applications from attosecond spectroscopy [4–7], high harmonic molecular spectroscopy [8–13], and lensless imaging [14–17] to ultrafast element sensitive spectroscopy at the M-edges of 3d transition metals [18–20]. The radiation created by strong field HHG provides a broadband EUV spectrum. The spectrum is structured in the form of odd harmonics of the fundamental laser frequency. The optimization of the HHG process often requires a quick estimate of the spectral content. However measuring the spectrum is much more complex than in the optical domain. In the EUV, reflection gratings only work under grazing incidence geometry, often reducing the collected light and making alignment (necessarily done in vacuum) difficult.

The work of Kornilov et al [21] addresses this issue by showing a compact and efficient EUV spectrometer design based on a special transmissive amplitude grating instead of a reflective grating. This lithographically produced element [22, 23] exhibits line densities in the 10 000 mm⁻¹ range, is rather fragile and expensive, but features a high dispersion and thus a very compact spectrometer design.

Here we propose the use of a commercial transmission electron microscope (TEM) grid as a grating for EUV light. The TEM grid is inserted between the focusing beamline optics (instead of in a separate setup [21]) and provides sufficient resolution to distinguish different harmonics in the EUV spectrum down to a wavelength of 27 nm (photon energy of 45 eV). We use a TEM grid of 80 lines/mm for this purpose. These devices are mass produced and feature densities up to 660 mm⁻¹ are commercially available for a low price compared to commercial EUV gratings. Since all EUV beamlines utilize some sort of focusing elements, the TEM grid spectrometer can be often implemented with minimal modifications.

Figure 1(a) shows our beamline layout. The EUV light is produced by HHG from a 30 fs, several mJ Ti:sapph laser, which is centered around a wavelength 800 nm and operated at a repetition rate of 1 kHz. For symmetry reasons, we only generate odd harmonics of the fundamental laser frequency. The details of our EUV beamline can be found in [24]. The infrared fundamental and EUV radiation are co-propagating and the much stronger infrared component is filtered using reflection off a silicon wafer [25] and a thin aluminum filter passage. The 100 nm thick aluminum filter is transparent in the photon energy range from 20 to 70 eV, passing harmonics 13–45. The Si wafer–Al filter combination serves to filter out the IR light, as required by all practical implementations of HHG. Without this filter, the CCD detector would saturate from the zero order of the IR light, which is orders of magnitude stronger than the EUV light. As shown recently, a microchannel plate can be used alternatively to filter the infrared light from the EUV [26]. The generation region of the EUV light is imaged without magnification on a detector by a grazing incidence toroidal mirror in 4f (in our case 4f = 2.4 m) geometry. A Si folding mirror at the place of the interaction region for an EUV transient grating setup [27] is hit under grazing angle 22^°. This mirror just serves to steer the EUV light onto the detector and provides some alignment opportunity and allows for measuring the spectrum without major changes in the setup. The detector is a back-thinned, EUV sensitive CCD camera with (27.6 × 6.9) mm² active area and a pixel size of (26 × 26) μm².

Zoom In Zoom Out Reset image size

The TEM grid (GS2000HS purchased from SPI) is inserted at R = 790 mm in front of the detector and attached to a gate valve with mounting hole for optical windows or filters. The mesh structure has a bar width of 5 μm, a hole width of w = 7.5 μm and a pitch of 12.5 μm. The pitch, corresponding to 80 lines/mm was independently checked using the diffraction image of a He–Ne laser.

Figure 1(b) shows the CCD detector image, with the EUV light hitting the TEM grid, on a logarithmic grayscale. At the point (0, 0) mm the image shows the bright non-diffracted light, which we refer to as zero-order light. The diffraction of the EUV light is clearly visible in terms of four stripes, two of them, respectively, on the x and y axes, the other two in an angle of 45° to the axes. Due to constraints in the vacuum tubes, we could only transport half of the diffraction features to the detector. The complete eightfold structure, expected from the diffraction of a coherent source on a mesh can be reconstructed by inversion at the (0, 0) point. We first analyze the data of the stripes parallel to the X and Y axes. Even in the raw image one clearly discerns different spots along these lines. We will show that they correspond to different high order harmonics of the 800 nm drive laser.

Figure 2(a) shows an integration along the abscissa from point (0, 0) to point (6.6, 0) with an integration width along the ordinate of ¼ mm. We chose a logarithmic representation of the signal as in figure 1(b) in order to show the details of higher order diffraction. We analyze the peak position in terms of the dispersion relation sinθ = nλ/d, where the diffraction angle is θ, d is the pitch of 12.5 μm and n is the diffraction order (not to be confused with the harmonic order). The distance of the diffracted light from the zero order on the detector D = tanθ R, where R = 790 mm is the distance from TEM grid to detector. The calculated positions of harmonics 13–29 in the first order are shown as black lines, fitting the peaks closest to the zero order very well. The first order diffraction defines the wavelength scale on the upper figure axis. The second order (n = 2) diffraction of harmonics 17–29 are shown as gray dashed lines. The prediction of second order diffraction of harmonics 17–23 clearly fit peaks with much lower intensity compared to the first order diffraction. Higher harmonics in second order are visible as shoulders on first order peaks. We even discern some third order diffraction of the grating. The calculated positions for harmonics 23–29 are given as dashed–dotted light gray lines and fit some smaller structures in the spectrum well corresponding to third order diffraction of harmonics 25 and 27.

Zoom In Zoom Out Reset image size

Figure 2(b) shows the integration from point (0, 0) mm in the −45° direction towards point (4, −4) mm. The diffraction image of a mesh with equal periodicity in both directions has equal periodicity along X and Y axes of the detector [28]. Thus in the direction 45° to the axis, the distance to the zero order of a diffraction spot increases by a factor of √2. The expected position of first order diffraction is shown as black lines and fits the structure very well. Once more, the first order diffraction defines the wavelength scale in the upper figure axis.

The highest harmonics in first order diffraction are certainly better resolved in the diagonal integration shown in figure 2(b) compared to the integration along the x-axis shown the figure 2(a). In addition, the second order diffraction in figure 2(a) shows better resolution compared to first order. The observation hints at the fact that the resolution is limited by the EUV spot size on the detector, which is 150 μm full width half maximum. For higher dispersion (as along the diagonal or for second order) the resolution increases consequently. Harmonic 19 has a full width half maximum of 2.5 nm in figure 2(a); in figure 2(b) the width is decreased to around 1.7 nm. For an infinitely small spot size on the detector, the resolution will be determined by the number of illuminated structures on the mesh. Our experimental results in the diagonal direction are a factor of 3 wider than this limit.

We now discuss the intensities observed in the diffraction pattern. The overall transmission of the grid is 36% resulting from the large bar width. Due to the convolution theorem, the diffraction intensity is given by the intensity of the Fourier transform of a single square hole multiplied by the Fourier transform of the grid periodicity [28]. The intensity of the Fourier transform of the single square opening in two dimensions as a function of the detector coordinates (X, Y) is given as

$$\begin{eqnarray}&&I\left( X,Y \right)={{I}_{0}}{{\left( \frac{{\rm sin} \left( \frac{\pi wX}{\lambda R} \right)}{\frac{\pi wX}{\lambda R}} \right)}^{2}}{{\left( \frac{{\rm sin} \left( \frac{\pi wY}{\lambda R} \right)}{\frac{\pi wY}{\lambda R}} \right)}^{2}}.\end{eqnarray} \tag{ 1 }$$

Here, w is the opening width (7.5 μm), λ is the wavelength and R the distance from the grid to the detector. For any particular wavelength, the diffraction intensity of the first order compared to the zero order is predicted to be 25% in the X or Y direction and 6.25% in the diagonal direction. Both the Fourier transform of a single square opening as well as the diffraction from the pitch scale identically with λ. Thus, the diffraction is independent of the wavelength. We now compare the experimentally determined intensities to this relation. Since our source has an enormous bandwidth, we need to estimate the spectral composition of the zero order radiation. The five strongest harmonics, which exceed their neighbors by at least a factor of three are the five odd harmonics from 15 to 23. We thus assume that the zero order radiation is composed of five equal parts of these harmonics. The strength of the zero order, which is normalized to one in figure 2 is thus 1/5 = 0.2. In the x-direction, the individual harmonics 15–23 have a signal of 0.05, which is 25% of the spectrally pure zero order, as predicted. In the diagonal direction, the signal strength is about 0.008 which corresponds to 4%. This is lower than the expected 6.25%, but in the light of the coarse estimate still reasonable.

The high diffraction efficiency and its wavelength independence are advantages compared to grazing incidence gratings. We anticipate applications of the TEM grid spectrometer going far beyond a simple characterization of HHG sources. The EUV spectral range covers M-edges of important 3d transition metals. Time resolved spectroscopy with probes in the M-edge spectral region therefore provides a path to understand the nuclear and electronic dynamics in complex materials via locally sensitive probes as shown recently [18–20]. The materials studied are often only available in the form of thin films that need to be supported by a grid. Thus including the dispersive TEM grid in the sample will provide a very efficient way to perform this type of spectroscopy.

To summarize, we show that TEM grids can be used to disperse an EUV spectrum produced by HHG of an intense infrared laser. Dispersion and diffracted intensity are in agreement with the expected diffraction of the TEM grid. The setup uses a toroidal mirror to image the HHG region on a detector. Due to its flexibility using already implemented beamline optics (and also its negligible investment) we anticipate that these grids are a superb tool for quickly estimating the spectral shape of the high harmonic spectrum. In addition, the diffraction efficiency of the transmission grating is independent of the wavelength making calibration of the grating response obsolete. Obviously, the relatively low line density limits the dispersion and therefore the resolution to shorter wavelengths. TEM grids with higher feature density around 660 lines/mm are commercially available. For our setup, they would allow resolution of harmonic orders around 65, corresponding to a wavelength of 12 nm and a photon energy of 100 eV.

We acknowledge discussions and/or experimental support with/from J Grilj, TJA Wolf, M Koch and A Belkacem. MG acknowledges funding via the Office of Science Early Career Research Program through the Office of Basic Energy Sciences, US Department of Energy. This work was supported by the AMOS program within the Chemical Sciences, Geosciences, and Biosciences Division of the Office of Basic Energy Sciences, Office of Science, US Department of Energy.