RESEARCH ARTICLE | DECEMBER 18 2023

Tabletop extreme ultraviolet reflectometer for quantitative

nanoscale reflectometry, scatterometry, and imaging 
Yuka Esashi  ; Nicholas W. Jenkins ; Yunzhe Shao ; Justin M. Shaw ; Seungbeom Park ;
Margaret M. Murnane ; Henry C. Kapteyn ; Michael Tanksalvala

Rev. Sci. Instrum. 94, 123705 (2023)

https://doi.org/10.1063/5.0175860

CrossMark

 
View Export
Online Citation

14 January 2024 13:29:39

 Review of ARTICLE pubs.aip.org/aip/rsi
Scientific Instruments

Tabletop extreme ultraviolet reflectometer

for quantitative nanoscale reflectometry,
scatterometry, and imaging
Cite as: Rev. Sci. Instrum. 94, 123705 (2023); doi: 10.1063/5.0175860
Submitted: 9 September 2023 • Accepted: 18 November 2023 •
Published Online: 18 December 2023

Yuka Esashi,1,a) Nicholas W. Jenkins,1 Yunzhe Shao,1 Justin M. Shaw,2 Seungbeom Park,3
1 1,4
Margaret M. Murnane, Henry C. Kapteyn, and Michael Tanksalvala1

AFFILIATIONS
1
Department of Physics, JILA, and STROBE NSF Science and Technology Center, University of Colorado Boulder and NIST,
Boulder, Colorado 80309, USA
2
Quantum Electromagnetics Division, National Institute of Standards and Technology, Boulder, Colorado 80305, USA
3
Core Technology R&D Team, Mechatronics Research, Samsung Electronics Co., Ltd., Hwasung 18848, Republic of Korea
4
KMLabs Inc., Boulder, Colorado 80301, USA

a)
Author to whom correspondence should be addressed: [email protected]

14 January 2024 13:29:39

ABSTRACT
Imaging using coherent extreme-ultraviolet (EUV) light provides exceptional capabilities for the characterization of the composition and
geometry of nanostructures by probing with high spatial resolution and elemental specificity. We present a multi-modal tabletop EUV imaging
reflectometer for high-fidelity metrology of nanostructures. The reflectometer is capable of measurements in three distinct modes: intensity
reflectometry, scatterometry, and imaging reflectometry, where each mode addresses different nanostructure characterization challenges. We
demonstrate the system’s unique ability to quantitatively and non-destructively measure the geometry and composition of nanostructures with
tens of square microns field of view and sub-nanometer precision. Parameters such as surface and line edge roughness, density, nanostructure
linewidth, and profile, as well as depth-resolved composition, can be quantitatively determined. The results highlight the applicability of EUV
metrology to address a wide range of semiconductor and materials science challenges.
Published under an exclusive license by AIP Publishing. https://doi.org/10.1063/5.0175860

I. INTRODUCTION microscopy (SEM), optical and x-ray reflectometry and scatterom-

etry, ellipsometry, atomic force microscopy (AFM), and scanning
Advances in next-generation semiconductor and quan- transmission electron microscopy (STEM). These techniques have
tum devices can benefit from new nanometrology capabilities sensitivity to different quantities, and it is often the case that no sin-
that can accurately and non-destructively probe with high gle metrology technique is comprehensive enough to capture all the
spatial resolution.1,2 Modern devices often rely on intricately relevant parameters of a sample. Moreover, the diverse requirements
designed 3D multilayer structures with layer thicknesses as posed by different samples and applications, such as field of view,
small as a few nanometers. As device dimensions shrink to the spatial resolution, elemental and chemical contrast, throughput, ease
nanoscale, their functional properties can no longer be accurately of sample preparation, and non-contact or non-destructive modes
described by macroscopic models due to the increasing impact of of operation, cannot be fulfilled by a single technique. In pursuing
imperfections—such as interface roughness and oxide layers—on holistic approaches to metrology where combinations of multiple
their properties and performance.3–5 While precise control over the techniques are used to characterize a sample, it is desirable to have a
geometry and composition of these structures is crucial for device wide range of techniques with different capabilities.
function, non-destructive characterization poses a great challenge. Here, we present a multi-modal extreme ultraviolet (EUV)
Common techniques for measuring the geometry and com- reflectometer with unique capabilities. This system enables quan-
position of fabricated nanostructures include scanning electron titative, non-destructive, and spatially resolved extraction of nano-

Rev. Sci. Instrum. 94, 123705 (2023); doi: 10.1063/5.0175860 94, 123705-1
Published under an exclusive license by AIP Publishing
 Review of ARTICLE pubs.aip.org/aip/rsi
Scientific Instruments

structure geometry and composition with minimal or no sample (CDI). This mode is suitable for evaluating complex, non-periodic
preparation. It can probe large areas, achieve angstrom-level sen- structures.
sitivity to topography, and provide good elemental contrast. This EUV light offers exceptional capabilities for compositional
EUV reflectometer can be used in three distinct modes, with each characterization, owing to its large interaction cross section over a
addressing distinct length scales and periodicity of nanostructures. wide range of materials as well as its high elemental specificity due to
The first mode is intensity reflectometry, where the specular reflec- the presence of element-specific absorption edges in this wavelength
tivity of a transversely uniform sample is measured as a function of range. Moreover, except near the characteristic core-level absorption
the incidence angle. This mode is useful for assessing layer thick- edges in materials, EUV photons primarily coherently scatter off
nesses, surface/interface roughness, and the composition of bulk core electrons and are not influenced by electrons that participate in
or multilayered samples.6–15 The second mode is scatterometry, chemical bonding. This allows for the ab initio calculation of optical
where the diffraction efficiency of periodic structures is measured constants for arbitrary materials using tabulated elemental scattering
as a function of the incidence angle. This mode allows for pre- factors and the number density of each elemental species. In con-
cise determination of the geometry and material parameters of trast, optical constants in the visible, vacuum ultraviolet (VUV), and
periodic nanostructures.16–24 The third mode, which is the most ver- infrared (IR) spectral regions are sensitive to chemical bonds and
satile, complex, and unique mode of this system, is imaging reflec- usually require experimental measurements at specific wavelengths
tometry.25 In this mode, diffraction data are used to reconstruct for each material, even for a minor change in chemical composi-
high-resolution phase-and-amplitude reflectivity images of a sam- tion. Regarding dimensional measurements, the short wavelength
ple at multiple incidence angles using coherent diffractive imaging of EUV light (∼10–100 nm) enables high-resolution imaging at

14 January 2024 13:29:39

FIG. 1. Multi-modal extreme ultraviolet tabletop reflectometer setup. (a) Schematic of the entire beam line. (b) Schematic of the reflectometer. Degrees of freedom that are
controllable under vacuum are indicated by pink arrows; stages indicated by solid-line arrows are primarily used during data collection, while ones indicated by dotted-line
arrows are primarily only used for alignment. (c) Zoom-in from a different angle, showing that the reflected beam from the beam splitter is directed straight to the camera to
act as a reference beam; the transmitted beam reflects from the sample.

Rev. Sci. Instrum. 94, 123705 (2023); doi: 10.1063/5.0175860 94, 123705-2
Published under an exclusive license by AIP Publishing
 Review of ARTICLE pubs.aip.org/aip/rsi
Scientific Instruments

the diffraction limit, as well as enhanced sensitivity to interfaces consisting of an 80 nm Ni coating on an Al substrate, designed for
and small topographic deviations. Furthermore, EUV light exhibits a 5-degree angle-of-incidence. The ellipsoid focuses the beam onto
a favorable combination of penetration depth and reflectivity: it the sample, with a demagnification of 22×, to a spot of ∼3 μm dia-
can penetrate several tens to hundreds of nanometers into mate- meter. (The EUV focus size can be measured by moving a sharp edge
rials that are opaque to visible light, making it possible to access of a sample across the beam to perform a knife-edge test and con-
buried interfaces. EUV light also has a high reflectivity even at inci- firmed by computational imaging reconstructions of the beam; see
dence angles of tens of degrees from grazing, allowing the incident Sec. V C and Fig. S1 in the supplementary material.) Since the beam
beam to remain small without suffering from the projection effects is oriented near grazing incidence with respect to the sample, the
encountered by x-ray reflectometers that operate at extreme graz- beam elongates, and its width in the horizontal direction increases
ing angles.26–29 Finally, the ability to generate bright and coherent by up to 5.8× at 10○ from grazing. The ellipsoidal optic is mounted
EUV light using high harmonic generation (HHG) tabletop sources on a hexapod (Smaract SmarPod 70.42-HV) that allows rotation and
presents an exciting opportunity for the development of new and translation around three axes for precise alignment. The incident
compact nano-characterization tools. beam on the sample is S-polarized. The beam divergence half-angle
at the sample is roughly 0.5○ , depending on how tightly the iris is
closed. Note that the first EUV multilayer mirror can be translated
II. EXPERIMENTAL SETUP
out of the EUV beam, which then bypasses the reflectometer and
Figure 1(a) shows the experiment layout for the tabletop EUV instead is incident onto a diagnostic camera for viewing the direct,
reflectometer beamline. A modelocked Ti:sapphire laser of center unfocused beam.
wavelength λ = 795 nm, 80 MHz repetition rate, and 5 nJ pulse Immediately following the ellipsoid is a removable 50 nm thick
energy seeds a Ti:Sapphire regenerative amplifier (KMLabs Wyvern Si membrane (Norcada NSUF1065A) coated with 3 nm of Pt on
HP). The amplifier provides 38 fs pulses at a 3 kHz repetition rate, a rotation stage (Smaract SR-2013). This is used as an EUV beam
with pulse energy up to 2.6 mJ. To reduce the detrimental nonlin- splitter to generate a reference beam that is directed to the edge of
ear B-integral associated with a high peak laser intensity, we use a the camera for reflectometry measurements that greatly benefit from
fraction (<1 mJ) of the laser energy and propagate it through the EUV beam intensity normalization. The transmitted EUV beam is
beamline as a chirped pulse. The pulse is then compressed using a focused on the sample, and the specular reflected beam from the
sequence of chirped mirrors immediately prior to the high harmonic sample is directed to the center of the camera sensor. This geometry
EUV generation step. This approach works well for high harmonic is shown in Fig. 1(c). The Pt coating on the beam splitter membrane

14 January 2024 13:29:39

wavelengths of ∼30 nm that do not require high laser pulse intensity. serves a dual purpose: first, it helps to balance the reflected and trans-
The pointing of the driving laser is controlled using a pair of cam- mitted power from the beam splitter (∼5% and ∼13%, respectively;
eras and piezo-actuated mirrors that stabilize the beam centroid at the transmitted arm is set higher to account for the further reduc-
two points along its propagation. tion from the reflection on the sample). Second, it minimizes the
The beam is focused into either a hollow-core waveguide or a sensitivity of the reflectivity to small variations in wavelength and
semi-infinite gas cell (modified prototype KMLabs XUUS4) to gen- incidence angle, which would otherwise introduce noise into the ref-
erate coherent extreme ultraviolet light via HHG around photon erence measurement. The power ratio can further be optimized by
energies of ∼41.5 eV. The rest of the beamline is kept at a vacuum adjusting the rotation of the beam splitter. For measurements that
of <10−5 Torr to prevent absorption of the EUV light in air. The do not benefit from a reference, the beam splitter can be rotated out
EUV source is detailed further in Sec. III. of the beam path.
Several optics are used prior to the reflectometer to separate the The sample is mounted on a customized stack of stages that
EUV beam from the driving IR beam. Two rejector mirrors (Gooch allows for 3D linear translation (three Smaract SLC-1730-HV) and
& Housego) used near the Brewster angle of the IR light reflect the sample-plane rotation (SR-2013-S-HV) for navigation, orientation,
EUV while absorbing much of the driving laser. A plasma cleaner and scanning, as well as rotation around the vertical axis (SR-2812-
(PIE Scientific EM-KLEEN) positioned above the first rejector mit- S-HV) to change the incidence angle of the EUV beam. A total of
igates hydrocarbon contamination. The plasma cleaner is typically five modified SEM stubs can be loaded for sample mounting, and
run with a 20%–80% mixture of O2 and Ar for optimal cleaning. The samples of thickness up to ∼2 mm are accommodated. The sample-
rejectors are followed by up to two metal foil filters (Luxel) that fur- mounting stage can be positioned with the sample face parallel to the
ther attenuate the IR intensity to significantly below the noise floor. illumination and can be retracted away from the beam path, enabling
The filters can be swapped or removed without breaking the vac- the beam to propagate freely to the detector. This is useful for deter-
uum. An adjustable iris sets the on-camera mode shape of the EUV. mining the intensity of the incident beam on the sample as well as
A shutter (Uniblitz VS14S2T0-EC2) is synchronized to the camera for measuring the out-of-focus beam mode.
exposures to prevent streaking during camera readout. The in-vacuum CCD EUV camera (Princeton Instruments
Figure 1(b) shows the reflectometer setup and the mechani- MTE2 used for these measurements) is mounted on a rotation stage
cal degrees of freedom that are controllable under vacuum. First, (Newmark RM-3-101V–C1) and is designed to rotate with the sam-
the EUV beam is incident onto two 45○ angle-of-incidence multi- ple in a θ–2θ configuration to maintain the EUV light on the camera
layer mirrors (NTT Advanced Technology Corporation) that allow as the sample is rotated. The camera uses a flexible readout cable and
for alignment of the beam into the reflectometer setup and the final cooling water line bellows to allow for smooth movement inside the
selection of the EUV wavelength from the comb of harmonic fre- chamber. The vertical rotation axes of the camera and the sample are
quencies generated by HHG. The EUV beam is then focused by a aligned to each other using a positive stop. These axes of rotation are
custom ellipsoidal optic (machined by Welch Mechanical Design), mounted together on two linear stages (OptoSigma TSDS-652CUU),

Rev. Sci. Instrum. 94, 123705 (2023); doi: 10.1063/5.0175860 94, 123705-3
Published under an exclusive license by AIP Publishing
 Review of ARTICLE pubs.aip.org/aip/rsi
Scientific Instruments

which are used to align the rotation axes with the beam focus. The gas handling systems were not optimized for the high gas pressures
surface of the sample is brought to the rotation axes by inserting (∼600 Torr of He gas) and high laser pulse energies (∼2.7 mJ pulses)
one of the 3D linear translation stages, which is necessary to main- required.
tain the beam’s position on the sample as the incidence angle is The total EUV flux integrated over the harmonic comb is on
changed. the order of 1012 photons/sec at the source using Ar gas, with the
The camera is configured such that 200 light-insensitive over- flux limited by the 3 kHz repetition rate of the driving laser (using a
scan columns are read out with each frame. The pixels in these higher repetition rate laser and a more optimal geometry, the EUV
columns serve to characterize several independent sources of back- flux can be further enhanced by ∼100×; however, the throughput
ground or noise on the detector, such as the standard deviation of of the system would still be limited by the CCD camera readout
the readout noise and the analog-to-digital converter (ADC) offset. rate). From this beam, the incident flux on the sample is roughly
These statistics can be used to minimize the effect of noise between 109 photons/sec in a single harmonic order at 30 nm due to optic loss
frames or even within a single frame. and narrowed bandwidth. For example, assuming minimal surface
The stage stack can perform reflectometry over a range of 0–55○ roughness and no surface contamination, the reflectivity for 30 nm
from grazing while keeping the diffraction on the camera sensor. light from a single rejector or a multilayer mirror is ∼0.5, that of the
If moving this full range, the sensor can be as close to the sam- ellipsoid is ∼0.8, and the transmission of a single 0.2 μm Al filter
ple as 30 mm, corresponding to 0.42 NA and a diffraction limited is ∼0.4.
resolution in the vertical direction of ∼36 nm when using 30 nm We tested the stability of the source over 30 minutes by first
illumination (or ∼16 nm when using 13 nm illumination). Due to using a photodiode to monitor the pulse-to-pulse IR laser inten-
conical diffraction, the resolution in the horizontal direction suffers sity as well as the second harmonic (generated by passing the IR
by a factor of ∼1/sin(θ), where θ is the incidence angle measured laser beam through a beta barium borate crystal). The EUV inten-
from grazing. As discussed in the later sections, the axial precision sity and pointing stability were then measured after propagating the
is on the order of an angstrom, depending on the sample and the HHG beam 2.5 m from the source onto a CCD camera with a 10 ms
amount of data collected. exposure time. The EUV spectrum was also measured using a diag-
The material of the rejector optics, metal filters, and multilayer nostic beamline consisting of a toroidal focusing optic and a grating.
mirrors employed in the beamline depends on the target EUV wave- The IR, the second harmonic, and the EUV intensity rms stabili-
length. For 30 nm, Si rejectors, Al filters, and SiC/Mg multilayer ties were 0.85%, 1.7%, and 1.36%, respectively. The unfocused EUV
mirrors are used. For 13 nm, Ru-coated Si rejectors, Zr filters, and beam had an rms pointing stability of 15 μrad, corresponding to <1%

14 January 2024 13:29:39

Mo/Si multilayer mirrors are used. of the beam radius, while the spectrum had an rms energy stability
The experimental software for this system allows automatic of 22 meV for the 27th harmonic at around 30 nm.
and programmed movement of the sample and the camera for data We also verified the harmonic selectivity of the system by
collection. Importantly, when the sample is rotated to change the inserting a grating as the sample and collecting the diffraction orders
incidence angle of the beam, a slight misalignment between the beam on the CCD camera. Not accounting for the slight wavelength-
focus, the sample rotation axis, and the sample surface can cause dependence of the diffraction efficiency, we found the 25th, 23rd,
the beam spot to translate horizontally on the sample. The soft- and 21st orders to have 4%, 0.6%, and 0.07% of the intensity of
ware allows for a calibrated correction of this offset such that it is the 27th harmonic.
possible to stay on the same field of view as data are collected across
a range of incidence angles. The software also allows the user to set IV. SAMPLE MODELING AND DATA
different exposure times for data frames collected at different inci- FITTING PROCEDURES
dence angles to maximize the use of the camera’s dynamic range
despite the change in reflectivity as a function of the incidence angle. All three modes of reflectometry that the system offers involve
At a single beam position on the sample, duplicate frames with var- the inverse problem of finding a sample model with a theoreti-
ied exposure times can also be collected and subsequently combined cally calculated signal that matches the experimentally measured
for higher dynamic range data. Finally, the order in which incidence signal.
angles are addressed is randomized to decouple any systematic error To calculate the expected reflectivity or diffraction efficiency
from the resultant curve. for a given sample, it is first necessary to determine the index of
refraction of the constituent materials assumed by the model at the
illumination wavelength. The light–matter interaction of EUV pho-
III. HIGH-HARMONIC GENERATION tons is dominated by core electrons; thus, the index of refraction n
EUV LIGHT SOURCE depends only on the number density of each elemental species
and can be calculated using pre-characterized elemental scattering
The high-harmonic generation setup consists of either a 150 μm
factors in the following equation:30
inner diameter hollow-core fiber or a semi-infinite gas cell (sealed
with a blank metal Swagelok VCR gasket) that is optimized to pro-
re 2
duce a comb of discrete frequencies around 30 nm. To generate n = 1 − δ − iβ = 1 − λ ∑ nj ( f 1,j + i f 2,j ). (1)
30 nm of light, the fiber/gas cell is filled with ∼30 Torr of Ar gas, and 2π j
the pulse energy is set to roughly 0.83 mJ at the input of the fiber/gas
cell. In this case, harmonics between 27 and 38 nm are generated. In the x-ray and EUV regimes, the index is very close to 1, and so it
Although it is possible to use this same setup to generate 13 nm light is often expressed using δ, β ≪ 1. Note that the sign on β can change
(harmonics between 11 and 16 nm), the laser delivery optics and depending on the sign convention used to describe electromagnetic

Rev. Sci. Instrum. 94, 123705 (2023); doi: 10.1063/5.0175860 94, 123705-4
Published under an exclusive license by AIP Publishing
 Review of ARTICLE pubs.aip.org/aip/rsi
Scientific Instruments

plane waves. re is the classical electron radius, λ is the wavelength of The error metric that is minimized in the genetic algorithm is
the light in vacuum, nj is the number of atoms of type j per unit vol- the chi-square, χ 2 ,41
ume, and f 1,j and f 2,j are the real and imaginary parts of the unitless
N
yi − y(θi ∣a) 2
χ2 = ∑ ( ),
atomic scattering factor for that atom type. Scattering factors are tab-
(3)
ulated in multiple databases;31,32 in this paper, we use the database i=1 σi
from the Center for X-ray Optics (CXRO).31
Second, the index of refraction is used in Parratt formalism33 where the summation is over N data points. yi is the measured data,
calculations to predict the complex reflectivity from a transversely y(θi ∣a) is the theoretically calculated datapoint for incidence angle
uniform region or a large structure (i.e., with a width many times θi and the vector of solved-for parameters a, and σ i is the standard
the wavelength). Parratt formalism uses Fresnel coefficients at each error of the mean for that data point.
interface in a recursive computation to rigorously calculate both Finally, once a good fit to the data have been found, the error
the amplitude and phase of the reflected (or transmitted) wave as bars of the solved parameters can be calculated using the covariance
it hits a multilayered stack for a given incidence angle and wave- matrix;41 this procedure is outlined in the supplementary material.
length. This formalism also allows the use of Névot–Croce factors While the error bars reported from the covariance matrix take
to approximate the effect of surface and interface roughness on account of the correlation between the parameters (hence we refer to
reflectivity.34,35 this as the “multi-parameter” confidence interval), it is also possible
Reflectivity calculated by Parratt formalism can be combined to calculate a “single-parameter” confidence interval by individually
with Fourier optics to obtain the diffraction efficiency of periodic varying parameters in the solution until the χ 2 error metric increases
structures that are wide and thin relative to the wavelength. In doing by Δχ 2 . This is a rough estimate of how low the confidence intervals
so, it is important to account for the geometric phase shift that arises could be if more parameters were fixed, meaning less parameters
from surface topography. This can either be incorporated into the are jointly solved for. It is intended to give an idea of the order-
Parratt calculation by placing a layer on top of the recessed parts of-magnitude sensitivity of 30 nm light to each parameter. For the
of the sample with the vacuum index or by adding the following demonstrations in this paper, both of these confidence intervals are
geometric phase shift φ: reported for 1σ (i.e., Δχ 2 = 1).

φ = 360○ [2h sin (θ)/λ], (2) V. THREE MODES OF REFLECTOMETRY

14 January 2024 13:29:39

A. Intensity reflectometry
where h is the recessed height, θ is the incidence angle measured In intensity reflectometry mode, we measure the specular
from grazing, and λ is the wavelength of the light in the external reflectivity of the EUV beam on the sample as a function of incidence
medium. angle to solve for layer thicknesses, surface and interface rough-
While this was not used for the demonstrations in this publica- ness, and other material parameters on transversely uniform sample
tion, for sub-wavelength and/or higher aspect-ratio structures where surfaces. EUV intensity reflectometers have been demonstrated for
3D effects of diffraction become more prominent, electromagnetic the characterization of EUV optical constants and multilayer optics,
solvers such as Rigorous Coupled-Wave Analysis (RCWA)36,37 both at synchrotrons6–10 and on table-top systems.11–15
should be used at the expense of increased computation In this mode, four types of data are collected. The first type
time. are the “sample” frames that measure the beam reflecting from the
Third, to optimize the sample model to fit the experimental sample at each incidence angle. In these frames, the reference beam
data, the sample model must be parameterized in terms of material, that is split off from the beam splitter is also captured simultane-
layer thicknesses, and surface/interface roughness. Out of all the ously. The second type are the “beam” frames, where the sample is
parameters, some are chosen as parameters of interest to be solved, translated out of the way of the beam to allow it to go directly onto
while others are fixed at their nominal or measured values. In addi- the camera at normal incidence. These frames serve to measure the
tion to the sample parameters, we self-calibrate our system by fitting absolute flux of the EUV beam. The third and fourth types are the
for small offsets in the wavelength and sample incidence angle to background frames for both the sample and the beam frames, which
account for slight misalignments. In addition, we sometimes also are collected with the shutter closed so that no EUV photons reach
solve for the deposition rate of carbon contamination due to EUV- the detector. The collected data are processed (procedure detailed in
induced dissociation of hydrocarbons. This is solved as a scaling of the supplementary material) to obtain the absolute reflectivity as a
the cumulative fluence on the sample, which is the dominant factor function of angle.
that determines the carbon deposition rate.38 To demonstrate the intensity reflectometry mode, the reflec-
Fourth, we use a multi-variable optimization scheme to fit the tivity of a 100 nm thick Au film deposited on a Si substrate was
model to the data. While there are many available, we use the genetic measured. Intensity reflectometry mode can utilize the optional
algorithm, which is an optimization method that is based on natu- EUV beam splitter for normalization of the incident intensity, so
ral selection, with Matlab’s ga() function. The genetic algorithm has first, to characterize the performance of the beam splitter, 400 sam-
been shown to work well with X-ray Reflectivity (XRR) measure- ple frames were collected at 30○ from grazing. All frames (for the
ments due to their robustness and ability to find the global minimum beam splitter characterization and the actual reflectometry) were
when many local minima are present.39,40 The algorithm can also be collected at the 2 MHz readout rate of the CCD cooled to 10 ○ C.
run multiple times with different sets of initial populations to verify While the camera can be cooled down to a lower temperature, we
that the found solution is the global minimum. only cool it down to a temperature that is necessary to sufficiently

Rev. Sci. Instrum. 94, 123705 (2023); doi: 10.1063/5.0175860 94, 123705-5
Published under an exclusive license by AIP Publishing
 Review of ARTICLE pubs.aip.org/aip/rsi
Scientific Instruments

FIG. 3. Intensity reflectometry on a 100 nm Au thin film on a Si substrate to

determine the surface roughness and density. The measured reflectivity (black
datapoints with error bars) and the theoretical solution fit (solid light green) with a
surface roughness value of 1.5 nm. Additional theoretical curves calculated from
the found solution with varied surface roughness are also shown in dotted lines.
Inset: cross-sectional sample schematic and solved parameters.
FIG. 2. Intensity normalization using the EUV beam splitter. Normalized intensity
of the EUV beam reflecting off the Au thin film sample as a function of camera
frames, with and without normalization using the EUV beam splitter. Fluctuations
are greatly reduced to a level close to the shot-noise limit. Each frame collects
∼1.5 × 106 photons in the beam reflecting off the sample. TABLE I. Fitted surface roughness and density with confidence intervals of the Au
thin film sample measured by intensity reflectometry.

Confidence interval
Fitted
mitigate noise while minimizing condensation of moisture on the parameters Multi-parameter Single-parameter
camera sensor from the imperfect vacuum.
Figure 2 shows the intensity of the beam reflecting from the Surface 1.5 0.9 <0.1

14 January 2024 13:29:39

sample, measured in the 400 frames collected at 30○ , with and with- roughness (nm)
out beam splitter intensity normalization. With a 0.18 s exposure Au density 18.5 3.0 0.01
time, each frame collects ∼1.5 × 106 photons. The rms of the per- (g/cm3 )
frame intensity is reduced from 1.30% to 0.16% via normalization,
and this stability is very close to the shot-noise limit of 0.11% for the
photon flux in this experiment.
For the actual intensity reflectometry on the gold film sam-
ple, 50 frames of “sample” data were collected at 19 angles between solution but with different surface roughness. The fitted parameters
26 and 44○ from grazing in 1○ increments, at an exposure time of are consistent with the surface roughness that was measured on
0.22 s. At each angle, over the course of collecting the frames, the an AFM (1–1.5 nm), optical profilometer (0.8–1.5 nm), and XRR
sample was translated vertically by 735 μm (15 μm per frame) to (∼1.7 nm), the density measured by XRR (18.5–19.1 g/cm3 ), and the
ensure that the reflectivity measured at one angle is the average expected density of Au in a thin-film (18.4–19.2 g/cm42,43 ). Slight
across an area and is not negatively influenced by small defects that discrepancies are deemed reasonable, as each technique measured a
might be present on the sample; this translation can be decreased slightly different location and field of view on the sample at different
or eliminated to measure an area as small as a few μm2 if desired. spatial resolutions.
Ten frames of “beam” data were collected in between each group
of “sample” frames at an exposure time of 0.13 s. For each group of B. Scatterometry
“sample” or “beam” frames, three background frames were collected. EUV scatterometers for the characterization of nanoscale grat-
Figure 3 shows the measured reflectivity from the gold thin ings have been developed in several different configurations;16–24
film sample as a function of the incidence angle. The use of the diffraction efficiency can be measured at one or many wavelengths,
beam splitter normalizes out the incident beam power fluctuation, at only some or many diffracted orders. In our system, we measure
resulting in small error bars of ∼0.1%. The model of the sample the diffraction efficiency in multiple diffracted orders using a single
was parameterized by the surface roughness and the density of the wavelength as a function of the incidence angle.
gold film (the model of this and other samples that were used in The collected data and the analysis procedure are similar to the
this publication is detailed in the supplementary material). The fit- intensity reflectometry mode. We collect “sample” data that captures
ted model parameters found using the genetic algorithm are shown the diffraction pattern as well as the associated background frames.
in Table I. While the multi-parameter confidence interval on the It is possible to interleave direct beam image frames with sample
Au density is relatively large, the surface roughness is solved to a scattering frames to monitor the beam power; however, this is not
sub-nanometer confidence interval. This sensitivity is also shown by critical for this mode since the diffraction efficiency is calculated
the <1 Å single-parameter confidence interval, as well as the well- as a ratio between the different orders of the diffraction, making it
separated theoretical curves in Fig. 3 that are calculated using the insensitive to power fluctuations.

Rev. Sci. Instrum. 94, 123705 (2023); doi: 10.1063/5.0175860 94, 123705-6
Published under an exclusive license by AIP Publishing
 Review of ARTICLE pubs.aip.org/aip/rsi
Scientific Instruments

The data are processed as outlined in the supplementary was able to improve the fit significantly, but the genetic algorithm
material to obtain the frame-by-frame counts in each of the diffrac- still did not find a set of parameters that fit all three diffraction effi-
tion orders at each incidence angle. Then, diffraction efficiency for ciency curves. As shown in Fig. 4(f), only when both the LER and the
±mth order Dm is calculated using the following equation: pedestal were introduced was it possible to fit all three curves.
LER was modeled using a Debye–Waller-like factor that
C+m + C−m
Dm = mean( ), (4) approximates the effect of LER as a damping of the scattered
C+m + C−m + C0 intensity,44,45
where Cm is the counts recorded in the mth order in a single frame,
and the mean is taken across multiple frames collected at a given IDWF (qy ) = I0 (qy ) exp (−ξ 2 qy 2 ), (5)
incidence angle.
To demonstrate this mode, we measured Ni line gratings with a where I DWF and I 0 are the intensity of the diffraction orders of grat-
nominal 100 nm linewidth, 400 nm period, and 12.9 nm height pat- ings with and without LER, respectively, qy = (2πm/grating period),
terned using electron lithography on a polished sapphire substrate. and m is the order of diffraction. ξ is the standard deviation of the
50 frames of grating diffraction were measured at 30 angles between grating edge displacement. The applicability of this factor for EUV
11 and 40○ from grazing in a 1○ increment at varying exposure times scatterometry has been demonstrated, as long as the grating edge
between 0.23 and 8.5 s, depending on the incidence angle. At each displacement follows a normal distribution and when qy 2 ξ 2 ≪1.22
angle, over the course of collecting the frames, the sample was trans- For the grating under consideration, which is thin relative to the
lated vertically by 98 μm (2 μm per frame). Three background frames wavelength, this formulation of LER can be incorporated into the
were also collected at each angle. The grating was oriented with grat- Fourier optics calculation of diffraction efficiency.
ing lines parallel to the plane of incidence of the EUV beam (conical In the final fit shown in Fig. 4(f), the solved parameters were
mounting). All frames were collected at the 2 MHz readout rate of the width of the tall main part of the grating, the main grating
the CCD cooled to 10 ○ C. LER, and the width and height of the pedestal. These parameters
While this sample was designed for a rectangular cross- are illustrated in Fig. 4(a). The solved parameter values and their
sectional profile, the AFM image as shown in Fig. 4(b) revealed that confidence intervals are reported in Table II. The AFM measured a
the fabricated structures deviated in two important ways: first, there main grating width of 56–80 nm, an LER of 5.6–6.5 nm, a pedestal
is a significant line edge roughness (LER) to the lines of the grating, width of 100–140 nm, and a pedestal height of 4–6 nm. In general,

14 January 2024 13:29:39

and second, there is a pedestal of several nanometers at the base of some discrepancy between EUV scatterometry and AFM measure-
the grating. ments is to be expected since the AFM measures over a much
The AFM imaging of the sample was motivated by the fact that smaller area of the grating. The discrepancy between EUV scat-
it was not possible to fit the diffraction efficiency of all first, second, terometry and AFM in LER is likely due to the finite tip size of
and third orders with a simple grating model of a rectangular cross the AFM (1–2 nm radius of curvature), making it insensitive to
section, as shown in Fig. 4(c). As shown in Figs. 4(d) and 4(e), the high frequency components of the LER, and because AFM mea-
introduction of only the LER or the pedestal in the sample model sures a top-down picture of the gratings while EUV scatterometry

FIG. 4. Scatterometry of Ni line gratings on a sapphire substrate to measure the grating profile. (a) Cross-sectional sample schematic and solved parameters. (b) Atomic
force microscopy image of the sample. (c)–(f) Measured first, second, and third order EUV diffraction efficiency as a function of the incidence angle and best fits obtained for
(c) simple rectangular grating, (d) grating with line edge roughness (LER), (e) grating with a pedestal, and (f) grating with LER and a pedestal. The best agreement with the
data are for a model that includes both LER and a pedestal.

Rev. Sci. Instrum. 94, 123705 (2023); doi: 10.1063/5.0175860 94, 123705-7
Published under an exclusive license by AIP Publishing
 Review of ARTICLE pubs.aip.org/aip/rsi
Scientific Instruments

TABLE II. Fitted parameters and confidence intervals for the linewidth, LER, and In particular, the phase upon reflection exhibits high sensitivity to
pedestal height width and height of the Ni gratings on the sapphire substrate as elemental composition and topography.25,54 Therefore, by collecting
measured by EUV diffraction scatterometry.
multiple ptychographic images of a sample at different incidence
Confidence interval angles, it is possible to perform reflectometry in a spatially resolved
Fitted manner. We call this technique EUV coherent diffractive imaging
parameters Multi-parameter Single-parameter reflectometry.25
For imaging reflectometry, a ptychographic scan is collected at
Main grating 64.7 0.8 0.1 multiple incidence angles. In addition, beam frames are also col-
width (nm) lected in between the ptychography scans to monitor the absolute
LER (nm) 10.0 0.5 0.1 flux of the EUV beam as well as to characterize the propagated
Pedestal 151 8 1.6 beam profile, which can be used in the ptychography algorithm as
width (nm) a constraint.55 As before, background frames are collected for each
Pedestal 1.2 0.3 <0.1 ptychography scan and each set of beam frames. The pre-processing
height (nm) steps of the raw data are detailed in the supplementary material.
To reconstruct the images, here we used the mPIE algorithm,56
although the technique does not depend on the use of any specific
implementation of ptychography. It is often helpful to reconstruct
measures over the height of the grating .45 The discrepancy in the images using two mutually-incoherent modes57,58 to reduce the
pedestal height could also partially be due to any remaining sub- effects of model mismatch and noise on the main reconstruction.54
tle differences between the physical sample and the model assumed The processed beam data can be used for modulus enforced probe
in EUV scatterometry (rectangular cross section, consisting of Ni, (MEP) constraints55 that can assist in the convergence of the recon-
with the same NiO thickness and surface roughness as the main grat- structed beam probe. In addition, in the case of a piecewise flat
ing). Obtaining a good estimate of absolute parameters using EUV sample (as is the case for most lithographically fabricated samples),
scatterometry relies on accurate modeling of the sample (as with denoising schemes such as total variation regularization25,59 can be
any model-based technique); the single-parameter confidence inter- incorporated into the ptychography loop to improve the image fur-
vals show a good sensitivity to parameter variations. In comparison ther. The detailed reconstruction procedure for this demonstration
with AFM or STEM and similarly to critical-dimension small-angle is included in the supplementary material.

14 January 2024 13:29:39

x-ray scattering (CD-SAXS), the strength of this technique is in the The reconstructed images are registered to each other, and the
ability to detect spatial variations in sample critical dimensions over phase step between two separate regions of the sample is calculated
large regions in a non-contact, non-destructive mode, with potential as a function of the incidence angle (the image analysis procedure is
applications for in-line inspection or metrology. detailed in the supplementary material). While imaging reflectome-
The small error bars on the diffraction efficiency are achieved try theoretically allows the analysis of both the amplitude and phase
by measuring the ratio of powers in the diffraction orders frame- of the ptychographic images, typically we only look at the phase as it
by-frame, which removes any effect of fluctuation in the incident is often reconstructed with higher fidelity. Note that since the phase
beam power. The residual variance at each angle comes from sample is only meaningful as a relative measurement, two distinct regions
inhomogeneity; in particular, LER is known not only to decrease the from the images must be chosen to measure the phase step.
diffraction efficiency but also to increase its variance.44 We performed imaging reflectometry on a calibrated AFM
We speculate that the unexpected LER and pedestal structure test sample from BudgetSensors (CS-20NG-UM) that consists of
are due to insufficient optimization of the electron-beam lithogra- SiO2 structures on a Si substrate, as shown in Fig. 5(a). A total of
phy on an insulating substrate, i.e., sapphire. We also note that a 11 ptychographic images were collected between 19○ and 29○ from
precise measurement of LER was challenging on the SEM, again due grazing in a 1○ increment. The distance between the beam focus
to the insulating substrate. This necessitated the deposition of a thin on the sample and the camera sensor was set to 45 mm, giving
discharge layer on top, which decreased the signal level from the an image pixel size of 61 nm (height) by 192 nm (width). At each
Ni grating. incidence angle, a ptychographic scan was collected in a 15 × 10
rectilinear grid of 0.35 and 0.6 μm step sizes, at exposure times rang-
C. Coherent diffractive imaging reflectometry ing between 0.4 and 1.0 s. In between the ptychographic scans, ten
Ptychography is a coherent diffractive imaging technique that beam frames were collected. For each ptychographic scan and beam
enables precise, quantitative imaging of a sample’s complex reflec- data, three background frames were collected. All frames were col-
tivity or transmissivity.46–49 In ptychography, the far-field intensity lected at the 1 MHz readout rate of the camera, which was cooled to
diffraction patterns are collected as a coherent beam is scanned over −10 ○ C. The ptychographic reconstruction procedure is detailed in
overlapping positions on the sample. An iterative phase retrieval the supplementary material.
algorithm is then used to generate a real-space phase-and-amplitude Figure 5(b) shows a full, complex ptychographic reconstruction
image of the sample. of the sample. The wide field of view compared to the scan area
Ptychography has found extensive application in the EUV and covered by the 1/e2 extent of the probe (roughly indicated by the
x-ray wavelength ranges due to its ability to achieve diffraction- white dotted square) is an advantage of grazing-incidence reflection-
limited imaging with high photon-efficiency and without an mode ptychography.25,60 The high-fidelity center regions from all
image-forming lens.50–53 Ptychography extracts the spatially 11 reconstructions are shown in Fig. 5(c), and the corresponding
resolved complex (i.e., amplitude and phase) reflectivity of an object. reconstructed probes are shown in the supplementary material. The

Rev. Sci. Instrum. 94, 123705 (2023); doi: 10.1063/5.0175860 94, 123705-8
Published under an exclusive license by AIP Publishing
 Review of ARTICLE pubs.aip.org/aip/rsi
Scientific Instruments

14 January 2024 13:29:39

FIG. 5. Imaging reflectometry on SiO2 structures on a Si substrate to measure substrate SiO2 thickness and structure height. (a) Sample schematic and the measured
parameters. (b) Ptychographic reconstruction at 24○ from grazing; brightness corresponds to the amplitude and the hue of the phase, as indicated by the color wheel, which
is shared with (c). The white dotted rectangle indicates the cropped region for (c) and corresponds roughly to the area covered by the 1/e2 extent of the beam. (c) Center of
the ptychographic reconstructions taken at 11 incidence angles. The black and white rectangles shown in the 19○ reconstruction indicate the pixels used in regions on the
structure and the substrate, respectively, to calculate the phase-step curve. (d)–(e) Measured phase step between the structure and the substrate (black datapoints) and the
theoretical solution fit (solid light green). In addition, shown are the curves calculated from the found solution with varied substrate SiO2 thickness between 1.5 and 4.5 nm in
a 0.25 nm increment in (d) and varied structure height between 18.0 and 22.5 nm in a 0.3 nm increment in (e).

change in contrast between the structure and the substrate among TABLE III. Fitted parameters and confidence intervals of SiO2 structures on a Si
the images comes from the incidence-angle dependent change in substrate measured by imaging reflectometry.
the phase-upon-reflection of EUV light from the sample; the cir- Confidence interval
cular contrast change at the center is due to the slight buildup of
Fitted
EUV-induced carbon contamination.
parameters Multi-parameter Single-parameter
The phase step between the SiO2 structure and the substrate
was calculated as the difference of the mean of the phase values of the Substrate SiO2 3.0 0.6 <0.1
pixels in the black and white rectangles shown in the 19○ reconstruc- thickness (nm)
tion of Fig. 5(b). This is shown as the black datapoints in Figs. 5(d) Structure 20.1 0.8 <0.1
and 5(e). height (nm)
The genetic algorithm solved for the thickness of the passive
SiO2 on the Si substrate and the topographic height of the SiO2 struc-
tures. The solution found and the confidence intervals are reported
in Table III. Both parameters are solved with sub-nanometer con- EUV imaging reflectometry is an extremely unique tech-
fidence intervals and are consistent with the estimate of the SiO2 nique in that it can measure the composition and layer thick-
thickness (∼2.5 nm) and the calibrated height (20.8 ± 0.4 nm) ness on and around non-periodic nanostructures in a spatially
provided by the manufacturer. The sensitivities to the two solved resolved manner without destructive sample preparation or milling.
parameters are also shown by the well-separated theoretical curves This set of capabilities is not provided by many other commonly
in Figs. 5(d) and 5(e) that are calculated from the solution but with used metrology techniques. Furthermore, the ability to identify
individually varied parameter values. defects or impurities in the reconstruction and avoid them in the

Rev. Sci. Instrum. 94, 123705 (2023); doi: 10.1063/5.0175860 94, 123705-9
Published under an exclusive license by AIP Publishing
 Review of ARTICLE pubs.aip.org/aip/rsi
Scientific Instruments

composition reconstruction is an advantage over traditional model- Science Foundation Science and Technology Center, Grant No.
based techniques. DMR-1548924. We also acknowledge support from Grant No.
AWD-22-06-0106 from Samsung Telecommunications America,
VI. CONCLUSION AND OUTLOOK LLC, for the imaging reflectometry demonstration in this work and
for suggesting the AFM calibration sample. The authors would like
We have demonstrated the performance of a tabletop EUV to thank Christina Porter as one of the initial co-designers of the
reflectometer for imaging and characterization of nanostructures. reflectometer in 2016–2018 (JILA, University of Colorado Boulder);
We present measurements from three modes of reflectometry that Weilun Chao (Lawrence Berkeley National Laboratory) for fabri-
are possible on this system, namely, intensity reflectometry, scat- cation of the gold thin film sample; Samuel Marks (University of
terometry, and coherent diffractive imaging reflectometry, each Colorado Boulder) for assisting the collection of x-ray reflectivity
demonstrating the high sensitivity of EUV light to topography and measurements on the gold thin film sample; J. Curtis Beimborn
composition. II and Amy Ekiriwang (JILA, University of Colorado Boulder) for
The three modes of reflectometry each cater to different length assisting with other components fabrication and metrology; and
scales and periodicities of nanostructures: intensity reflectometry Brendan McBennett and Joshua L. Knobloch (JILA, University of
for transversely uniform bulk and thin film geometry, scatterometry Colorado Boulder) for helpful discussions. J.M.S. fabricated the
for periodic structures, and imaging reflectometry for non-periodic grating sample used in the scatterometry demonstration and per-
structures. While the demonstration of the three modes was con- formed correlative metrology. He is a member of the Quantum
ducted on different samples in this study, we note that it is possible Electromagnetics Division of the National Institute of Standards and
to characterize a single sample with multiple modes for enhanced Technology (NIST). Certain commercial equipment, instruments,
precision of measurement. For example, a sample with periodic or materials are identified in this paper in order to adequately specify
structures may have a large portion of the substrate that is uncovered the experimental procedure. Such identification is not intended to
by structures, which can first be measured by intensity reflectom- imply recommendation or endorsement by NIST, nor is it intended
etry. The solution can then assist in the modeling of the sample to imply that the materials or equipment identified are necessarily
in scatterometry for a more precise characterization of the struc- the best available for the purpose.
tures. In addition, while not described in this publication, the system
can also be used to qualitatively characterize non-periodic nanoscale
AUTHOR DECLARATIONS
structures on the surface of a sample (such as nanoparticle size and

14 January 2024 13:29:39

distribution) from the shape and intensity of the EUV scatter pattern Conflict of Interest
or simply operate as a ptychographic microscope by taking an image
at a chosen incidence angle. H.C.K. and M.M.M. have a financial interest in KMLabs, which
The ability to non-destructively measure the geometry and produced the laser source used in this work. H.C.K. is partially
composition of nanostructures on a few to tens of μm2 field of employed by KMLabs. M.M.M., H.C.K., and M.T. are inventors on
view with sub-nanometer precision is a combination of strengths a U.S. patent application (2021/0325301 A1, published October 21,
unique to this EUV reflectometer and is largely unmatched by the 2021) and a patent in the European Patent Office (EP3458837B1)
existing array of commonly used metrology techniques. By realiz- related to this work filed by the University of Colorado.
ing this special capability on a tabletop setup, the EUV reflectometer
demonstrates the applicability of EUV metrology to a wide range of Author Contributions
semiconductor and material science challenges. Finally, we note that Yuka Esashi: Conceptualization (supporting); Data curation
in the future, imaging reflectometry could be used to locally char- (lead); Formal analysis (lead); Investigation (lead); Methodology
acterize highly periodic samples by using structured EUV beams (lead); Project administration (supporting); Software (supporting);
with orbital angular momentum that enhances the diversity in the Validation (lead); Visualization (lead); Writing – original draft
collected scatter patterns.61 In addition, by increasing the EUV (lead); Writing – review & editing (lead). Nicholas W. Jenkins:
flux (∼100×) and also using faster readout CMOS cameras, the Conceptualization (supporting); Data curation (supporting); For-
throughput of the instrument can be further improved. mal analysis (supporting); Investigation (supporting); Methodology
(supporting); Software (supporting); Validation (supporting);
SUPPLEMENTARY MATERIAL
Writing – review & editing (supporting). Yunzhe Shao: Data
See the supplementary material for (1) procedure for the calcu- curation (supporting); Formal analysis (supporting); Investigation
lation of confidence intervals of solved parameters; (2) data process- (supporting); Validation (supporting); Writing – review & editing
ing procedure for intensity reflectometry and scatterometry; (3) data (supporting). Justin M. Shaw: Formal analysis (supporting);
pre-processing, ptychographic reconstruction, and image analysis Resources (supporting). Seungbeom Park: Resources (support-
procedure for imaging reflectometry; (4) ptychographically recon- ing). Margaret M. Murnane: Conceptualization (lead); Funding
structed probes in the imaging reflectometry demonstration; and (5) acquisition (lead); Investigation (lead); Methodology (lead);
detailed sample model descriptions. Project administration (lead); Resources (lead); Supervision
(lead); Validation (supporting); Writing – original draft (lead);
Writing – review & editing (supporting). Henry C. Kapteyn:
ACKNOWLEDGMENTS Conceptualization (lead); Funding acquisition (lead); Investiga-
This research was performed at JILA, University of Colorado. tion (lead); Methodology (lead); Project administration (lead);
The authors acknowledge support from the STROBE National Resources (lead); Supervision (lead); Writing – original draft (lead);

Rev. Sci. Instrum. 94, 123705 (2023); doi: 10.1063/5.0175860 94, 123705-10
Published under an exclusive license by AIP Publishing
 Review of ARTICLE pubs.aip.org/aip/rsi
Scientific Instruments

Writing – review & editing (supporting). Michael Tanksalvala: Technol., B: Microelectron. Nanometer Struct.–Process., Meas., Phenom. 22,
Conceptualization (lead); Methodology (lead); Software (lead); 3059–3062 (2004).
17
Supervision (supporting); Validation (supporting); Visualization F. Scholze, C. Laubis, U. Dersch, J. Pomplun, S. Burger, and F. Schmidt,
(supporting); Writing – original draft (lead); Writing – review & “Influence of line edge roughness and CD uniformity on EUV scatterometry for
editing (supporting). CD characterization of EUV masks,” in Modeling Aspects in Optical Metrology
(SPIE, 2007), Vol. 6617, pp. 400–409.
18
H. Gross, A. Rathsfeld, F. Scholze, and M. Bär, “Profile reconstruction in
DATA AVAILABILITY extreme ultraviolet (EUV) scatterometry: Modeling and uncertainty estimates,”
The data that support the findings of this study are available Meas. Sci. Technol. 20, 105102 (2009).
19
from the corresponding author upon reasonable request. M.-A. Henn, S. Heidenreich, H. Gross, A. Rathsfeld, F. Scholze, and M. Bär,
“Improved grating reconstruction by determination of line roughness in extreme
ultraviolet scatterometry,” Opt. Lett. 37, 5229–5231 (2012).
REFERENCES 20
Y.-S. Ku, C.-L. Yeh, Y.-C. Chen, C.-W. Lo, W.-T. Wang, and M.-C. Chen, “EUV
1 scatterometer with a high-harmonic-generation EUV source,” Opt. Express 24,
N. G. Orji, M. Badaroglu, B. M. Barnes, C. Beitia, B. D. Bunday, U. Celano, R. J.
28014 (2016).
Kline, M. Neisser, Y. Obeng, and A. E. Vladar, “Metrology for the next generation 21
of semiconductor devices,” Nat. Electr. 1, 532–547 (2018). L. Bahrenberg, S. Danylyuk, S. Glabisch, M. Ghafoori, S. Schröder, S. Brose,
2
G. Orji, B. Bunday, and Y. Obeng, International Roadmap for Devices and J. Stollenwerk, and P. Loosen, “Characterization of nanoscale gratings by spec-
Systems: Metrology (IEEE, 2022). troscopic reflectometry in the extreme ultraviolet with a stand-alone setup,”
3
K. M. Hoogeboom-Pot, J. N. Hernandez-Charpak, X. Gu, T. D. Frazer, E. H. Opt. Express 28, 20489 (2020).
22
Anderson, W. Chao, R. W. Falcone, R. Yang, M. M. Murnane, H. C. Kapteyn, A. Fernández Herrero, F. Scholze, G. Dai, and V. Soltwisch, “Analysis of
and D. Nardi, “A new regime of nanoscale thermal transport: Collective diffusion line-edge roughness using EUV scatterometry,” Nanomanuf. Metrol. 5, 149–158
increases dissipation efficiency,” Proc. Natl. Acad. Sci. U. S. A. 112, 4846–4851 (2022).
(2015). 23
L. M. Lohr, R. Ciesielski, S. Glabisch, S. Schröder, S. Brose, and V. Soltwisch,
4
T. D. Frazer, J. L. Knobloch, J. N. Hernández-Charpak, K. M. Hoogeboom-Pot, “Nanoscale grating characterization using EUV scatterometry and soft x-ray
D. Nardi, S. Yazdi, W. Chao, E. H. Anderson, M. K. Tripp, S. W. King, H. C. scattering with plasma and synchrotron radiation,” Appl. Opt. 62, 117–132 (2023).
Kapteyn, M. M. Murnane, and B. Abad, “Full characterization of ultrathin 5-nm 24
C. Porter, T. Coenen, N. Geypen, S. Scholz, L. van Rijswijk, H.-K. Nienhuys,
low-k dielectric bilayers: Influence of dopants and surfaces on the mechanical J. Ploegmakers, J. Reinink, H. Cramer, R. van Laarhoven, D. O’Dwyer, P. Smoren-
properties,” Phys. Rev. Mater. 4, 073603 (2020).
5
burg, A. Invernizzi, R. Wohrwag, H. Jonquiere, J. Reinhardt, O. el Gawhary,
J. Fontcuberta, H. B. Vasili, J. Gàzquez, and F. Casanova, “On the role of inter-

14 January 2024 13:29:39

S. Mathijssen, P. Engblom, H. Chin, W. T. Blanton, S. Ganesan, B. Krist, F. Gstrein,
faces on spin transport in magnetic insulator/normal metal heterostructures,”
and M. Phillips, “Soft x-ray: Novel metrology for 3D profilometry and device pitch
Adv. Mater. Interfaces 6, 1900475 (2019).
6 overlay,” in Metrology, Inspection, and Process Control XXXVII (SPIE, 2023), Vol.
E. Gullikson, S. Mrowka, and B. Kaufmann, “Recent developments in EUV
12496, pp. 412–420.
reflectometry at the advanced light source,” in Emerging Lithographic Technologies 25
V, (SPIE, 2001), Vol. 4343. M. Tanksalvala, C. L. Porter, Y. Esashi, B. Wang, N. W. Jenkins, Z. Zhang,
7 G. P. Miley, J. L. Knobloch, B. McBennett, N. Horiguchi, S. Yazdi, J. Zhou,
F. Scholze, J. Tümmler, and G. Ulm, “High-accuracy radiometry in the EUV
range at the PTB soft x-ray beamline,” Metrologia 40, S224–S228 (2003). M. N. Jacobs, C. S. Bevis, R. M. Karl Jr., P. Johnsen, D. Ren, L. Waller, D. E. Adams,
8
C. Tarrio, S. Grantham, M. B. Squires, R. E. Vest, and T. B. Lucatorto, “Towards S. L. Cousin, C.-T. Liao, J. Miao, M. Gerrity, H. C. Kapteyn, and M. M. Murnane,
high accuracy reflectometry for extreme-ultraviolet lithography,” J. Res. Natl. Inst. “Nondestructive, high-resolution, chemically specific 3D nanostructure character-
Stand. Technol. 108, 267–273 (2003). ization using phase-sensitive EUV imaging reflectometry,” Sci. Adv. 7, eabd9667
9
S. Nannarone, F. Borgatti, A. DeLuisa, B. P. Doyle, G. C. Gazzadi, A. Giglia, P. (2021).
26
Finetti, N. Mahne, L. Pasquali, M. Pedio, G. Selvaggi, G. Naletto, M. G. Pelizzo, E. Chason and T. M. Mayer, “Thin film and surface characterization by specular
and G. Tondello, “The BEAR beamline at elettra,” AIP Conf. Proc. 705, 450–453 X-ray reflectivity,” Crit. Rev. Solid State Mater. Sci. 22, 1–67 (1997).
27
(2004). K. N. Stoev and K. Sakurai, “Review on grazing incidence X-ray spectrometry
10
H. Iguchi, H. Hashimoto, M. Kuki, T. Harada, H. Kinoshita, T. Watanabe, and reflectometry,” Spectrochim. Acta, Part B 54, 41–82 (1999).
Y. Y. Platonov, M. D. Kriese, and J. R. Rodriguez, “Extreme-ultraviolet collec- 28
G. Freychet, D. Kumar, R. Pandolfi, D. Staacks, P. Naulleau, R. J. Kline,
tor mirror measurement using large reflectometer at NewSUBARU synchrotron D. Sunday, M. Fukuto, J. Strzalka, and A. Hexemer, “Critical-dimension grazing
facility,” Jpn. J. Appl. Phys. 55, 06GC01 (2016). incidence small angle x-ray scattering,” Metrology, Inspection, and Process Control
11
E. M. Gullikson, J. H. Underwood, P. C. Batson, and V. Nikitin, “A soft for Microlithography XXXII (SPIE, 2018), Vol. 10585, pp. 217–223.
X-ray/EUV reflectometer based on a laser produced plasma source,” J. X-Ray Sci. 29
W. Wu, R. Joseph Kline, R. L. Jones, H.-J. Lee, E. K. Lin, D. F. Sunday, C. Wang,
Technol. 3, 283–299 (1992).
12 T. Hu, and C. L. Soles, “Review of the key milestones in the development of crit-
D. L. Windt and W. K. Waskiewicz, “Soft X-ray reflectometry of multilayer
ical dimension small angle x-ray scattering at National Institute of Standards and
coatings using a laser-plasma source,” in Multilayer Optics for Advanced X-Ray
Technology,” J. Micro/Nanopatterning, Mater., Metrol. 22, 031206 (2023).
Applications (SPIE, 1991), Vol. 1547. 30
13 D. Attwood and A. Sakdinawat, Scattering, diffraction, and refraction of electro-
M. Banyay and L. Juschkin, “Table-top reflectometer in the extreme ultraviolet
for surface sensitive analysis,” Appl. Phys. Lett. 94, 063507 (2009). magnetic radiation, X-Rays and Extreme Ultraviolet Radiation (Springer, 2016),
14 pp. 20–24.
S. Döring, F. Hertlein, A. Bayer, and K. Mann, “EUV reflectometry for thick- 31
ness and density determination of thin film coatings,” Appl. Phys. A 107, 795–800 B. L. Henke et al., “X-ray interactions: Photoabsorption, scattering, transmis-
(2012). sion, and reflection at E = 50-30,000 eV, Z = 1-92,” At. Data Nucl. Data Tables 54,
15
J. J. Abel, F. Wiesner, J. Nathanael, J. Reinhard, M. Wünsche, G. Schmidl, 181–342 (1993).
32
A. Gawlik, U. Hübner, J. Plentz, C. Rödel, G. G. Paulus, and S. Fuchs, “Absolute C. T. Chantler, “Detailed tabulation of atomic form factors, photoelectric
EUV reflectivity measurements using a broadband high-harmonic source and an absorption and scattering cross section, and mass attenuation coefficients in the
in situ single exposure reference scheme,” Opt. Express 30, 35671–35683 (2022). vicinity of absorption edges in the soft X-ray (Z = 30–36, Z = 60–89, E = 0.1
16
J. Perlich, F.-M. Kamm, J. Rau, F. Scholze, and G. Ulm, “Characterization keV–10 keV), addressing convergence issues of earlier work,” J. Phys. Chem. Ref.
of extreme ultraviolet masks by extreme ultraviolet scatterometry,” J. Vac. Sci. Data 29, 597–1056 (2000).

Rev. Sci. Instrum. 94, 123705 (2023); doi: 10.1063/5.0175860 94, 123705-11
Published under an exclusive license by AIP Publishing
 Review of ARTICLE pubs.aip.org/aip/rsi
Scientific Instruments

33 48
L. G. Parratt, “Surface studies of solids by total reflection of X-rays,” Phys. Rev. A. M. Maiden and J. M. Rodenburg, “An improved ptychographical phase
95, 359–369 (1954). retrieval algorithm for diffractive imaging,” Ultramicroscopy 109, 1256–1262
34 (2009).
L. Névot and P. Croce, “Caractérisation des surfaces par réflexion rasante de
49
rayons X. Application à l’étude du polissage de quelques verres silicates,” Rev. P. Thibault, M. Dierolf, O. Bunk, A. Menzel, and F. Pfeiffer, “Probe retrieval
Phys. Appl. 15, 761–779 (1980). in ptychographic coherent diffractive imaging,” Ultramicroscopy 109, 338–343
35
Y. Esashi, M. Tanksalvala, Z. Zhang, N. W. Jenkins, H. C. Kapteyn, and M. M. (2009).
50
Murnane, “Influence of surface and interface roughness on X-ray and extreme J. Miao, R. L. Sandberg, and C. Song, “Coherent X-ray diffraction imaging,”
ultraviolet reflectance: A comparative numerical study,” OSA Continuum 4, IEEE J. Sel. Top. Quantum Electron. 18, 399–410 (2012).
1497–1518 (2021). 51
J. Miao, T. Ishikawa, I. K. Robinson, and M. M. Murnane, “Beyond crystallogra-
36
M. G. Moharam and T. K. Gaylord, “Rigorous coupled-wave analysis of planar- phy: Diffractive imaging using coherent x-ray light sources,” Science 348, 530–535
grating diffraction,” J. Opt. Soc. Am. 71, 811–818 (1981). (2015).
37 52
M. G. Moharam, E. B. Grann, D. A. Pommet, and T. K. Gaylord, “Formulation F. Pfeiffer, “X-ray ptychography,” Nat. Photonics 12, 9–17 (2018).
for stable and efficient implementation of the rigorous coupled-wave analysis of 53
L. Loetgering, S. Witte, and J. Rothhardt, “Advances in laboratory-scale pty-
binary gratings,” J. Opt. Soc. A. A 12, 1068–1076 (1995). chography using high harmonic sources [Invited],” Opt. Express 30, 4133–4164
38
J. T. Hollenshead, L. E. Klebanoff, and G. Delgado, “Predicting radiation- (2022).
induced carbon contamination of EUV optics,” J. Vac. Sci. Technol., B: 54
E. R. Shanblatt, C. L. Porter, D. F. Gardner, G. F. Mancini, R. M. Karl,
Nanotechnol. Microelectron.: Mater., Process., Meas., Phenom. 37, 021602 (2019). M. D. Tanksalvala, C. S. Bevis, V. H. Vartanian, H. C. Kapteyn, D. E. Adams,
39
A. D. Dane, A. Veldhuis, D. Boer, A. J. G. Leenaers, and L. Buydens, and M. M. Murnane, “Quantitative chemically specific coherent diffractive imag-
“Application of genetic algorithms for characterization of thin layered mate- ing of reactions at buried interfaces with few nanometer precision,” Nano Lett. 16,
rials by glancing incidence X-ray reflectometry,” Physica B 253, 254–268 5444–5450 (2016).
(1998). 55
D. F. Gardner, M. Tanksalvala, E. R. Shanblatt, X. Zhang, B. R. Galloway, C. L.
40
A. Ulyanenkov, K. Omote, and J. Harada, “The genetic algorithm: Refinement Porter, R. Karl Jr, C. Bevis, D. E. Adams, H. C. Kapteyn, M. M. Murnane, and G. F.
of X-ray reflectivity data from multilayers and thin films,” Physica B 283, 237–241 Mancini, “Subwavelength coherent imaging of periodic samples using a 13.5 nm
(2000). tabletop high-harmonic light source,” Nat. Photonics 11, 259–263 (2017).
41 56
W. H. Press, S. A. Teukosky, W. T. Vetterling, and B. P. Flannery, Modeling A. Maiden, D. Johnson, and P. Li, “Further improvements to the ptychographi-
of data, Numerical Recipes: The Art of Scientific Computing, 3rd ed. (Cambridge cal iterative engine,” Optica 4, 736–745 (2017).
University Press, 2007), pp. 773–839. 57
P. Thibault and A. Menzel, “Reconstructing state mixtures from diffraction
42
W. Fischer, H. Geiger, P. Rudolf, and P. Wissmann, “Structure investigations on measurements,” Nature 494, 68–71 (2013).
single-crystal gold films,” Appl. Phys. 13, 245–253 (1977). 58
D. J. Batey, D. Claus, and J. M. Rodenburg, “Information multiplexing in
43
J. Siegel, O. Lyutakov, V. Rybka, Z. Kolská, and V. Švorčík, “Properties of gold ptychography,” Ultramicroscopy 138, 13–21 (2014).

14 January 2024 13:29:39

nanostructures sputtered on glass,” Nanoscale Res. Lett. 6, 96 (2011). 59
A. Beck and M. Teboulle, “Fast gradient-based algorithms for constrained total
44
A. Kato and F. Scholze, “Effect of line roughness on the diffraction intensities in variation image denoising and deblurring problems,” IEEE Trans. Image Process.
angular resolved scatterometry,” Appl. Opt. 49, 6102–6110 (2010). 18, 2419–2434 (2009).
45 60
A. Fernández Herrero, M. Pflüger, J. Probst, F. Scholze, and V. Soltwisch, C. L. Porter, M. Tanksalvala, M. Gerrity, G. Miley, X. Zhang, C. Bevis, E. Shan-
“Applicability of the Debye-Waller damping factor for the determination of blatt, R. Karl, M. M. Murnane, D. E. Adams, and H. C. Kapteyn, “General-purpose,
the line-edge roughness of lamellar gratings,” Opt. Express 27, 32490–32507 wide field-of-view reflection imaging with a tabletop 13 nm light source,” Optica
(2019). 4, 1552–1557 (2017).
46 61
D. Sayre, “Some implications of a theorem due to Shannon,” Acta Crystallogr. B. Wang, N. J. Brooks, P. Johnsen, N. W. Jenkins, Y. Esashi, I. Binnie,
5, 843 (1952). M. Tanksalvala, H. C. Kapteyn, and M. M. Murnane, “High-fidelity ptychographic
47
J. M. Rodenburg and H. M. L. Faulkner, “A phase retrieval algorithm for shifting imaging of highly periodic structures enabled by vortex high harmonic beams,”
illumination,” Appl. Phys. Lett. 85, 4795–4797 (2004). Optica 10, 1245–1252 (2023).

Rev. Sci. Instrum. 94, 123705 (2023); doi: 10.1063/5.0175860 94, 123705-12
Published under an exclusive license by AIP Publishing