Sie sind auf Seite 1von 11

Assessing surface imperfections of

freeforms using a robotic light


scattering sensor

Tobias Herffurth
Marcus Trost
Matthias Beier
Ralf Steinkopf
Nils Heidler
Tom Pertermann
Sven Schröder

Tobias Herffurth, Marcus Trost, Matthias Beier, Ralf Steinkopf, Nils Heidler, Tom Pertermann,
Sven Schröder, “Assessing surface imperfections of freeforms using a robotic light scattering sensor,” Opt.
Eng. 58(9), 092609 (2019), doi: 10.1117/1.OE.58.9.092609.

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 09 May 2019


Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Optical Engineering 58(9), 092609 (September 2019)

Assessing surface imperfections of freeforms


using a robotic light scattering sensor
Tobias Herffurth,* Marcus Trost, Matthias Beier, Ralf Steinkopf, Nils Heidler, Tom Pertermann, and Sven Schröder
Fraunhofer Institute for Applied Optics and Precision Engineering, Jena, Germany

Abstract. Light scattering-based characterization techniques are fast, sensitive, and robust, which can be
exploited for roughness measurements or homogeneity and defect assessment. However, only compact
scattering instruments enable even large freeform surfaces to be inspected close to fabrication or even within
fabrication processes. The combination of a light scattering sensor with an industrial robot, as well as its appli-
cation for full surface evaluation of aspheriscal and freeform surfaces, is addressed. In addition to the technical
considerations regarding the sensor, the robotic handling system, or the sample curvature, examples of appli-
cation to different optical components fabricated at the Fraunhofer Institute for Applied Optics and Precision
Engineering using state-of-the-art techniques are presented. The results include the discussion of anisotropic
surface structures induced by diamond turning or polishing as well as the analysis of the influence of a protection
layer. © 2019 Society of Photo-Optical Instrumentation Engineers (SPIE) [DOI: 10.1117/1.OE.58.9.092609]
Keywords: optical inspection; roughness; defects; scattering; assessment; freeforms and aspheres.
Paper 190156SS received Feb. 1, 2019; accepted for publication Apr. 5, 2019; published online Apr. 30, 2019.

1 Introduction instrumentation and results presented in this paper contribute


The increasing capabilities to fabricate extended aspherical to bridging the gap between the application of light scattering
or freeform optical components1–6 place challenging de- techniques in the laboratory and its use in the production
mands on the corresponding characterization techniques. environment.
In addition to the characterization of the form, this is particu- This paper is organized as follows: in Sec. 2 relevant def-
larly the case for surface quality assessment regarding rough- initions and the theoretical background are briefly summa-
ness, homogeneity, and defects. rized. The robotic light scattering sensor is described in
Light scattering-based techniques have demonstrated detail in Sec. 3, which is followed by three examples of its
their potential for this challenging task since they are highly applications in Sec. 4.
sensitive, noncontact, and robust regarding vibrations.7–9
In addition, they provide a direct measure of the scattering 2 Definitions and Theoretical Background
that is caused by surface imperfections and, therefore, pro- Light scattering distributions can be described by the angle-
vide valuable information on the disturbing influence of the resolved scattering (ARS), which is defined as the power,
imperfections,7,10–18 which could also be exploited in stray- ΔPs , scattered into the solid angle, ΔΩs , normalized to the
light modeling. Conversely, there is still a lack of noncom- incident light power, Pi . In addition to a cosine factor, ARS
plex and easy-to-use scattering instruments and measurement is identical to the bidirectional reflectance distribution func-
approaches for the fast characterization of extended aspheri- tion (BRDF) for surfaces.
cal or freeform surfaces with complex shapes. For optically smooth surfaces with a root mean square
In addition to the sophisticated and flexible but complex (rms) roughness much smaller than the illumination wave-
laboratory instruments,8–10 compact CMOS-matrix-based length, σ ≪ λ, the ARS is directly connected to the power
sensor approaches7,19 have been developed at Fraunhofer spectral density (PSD) function of the surface roughness
Institute for Applied Optics and Precision Engineering by the first-order perturbation scattering theory8,26–29
(Fraunhofer IOF). These sensors are used for the characteri-
zation of optical components or even biological objects.7,20–22 ΔPs ðθi ; θs ; φs Þ
ARSðθi ; θs ; φs Þ ¼
ΔΩPi
EQ-TARGET;temp:intralink-;e001;326;250

In addition, the compactness and robustness of these sensors


enable their combination with other instruments, for example 16π 2
a white-light interferometer (WLI) in Ref. 7 or with handling ¼ Q cos θi cos2 θs PSDðf x ; f y Þ: (1)
λ4
systems such as the scanning stage in Ref. 23.
In this paper, further developments of the sensor, its Here, Q is an optical factor, which includes the conditions of
measurement modes, and its combination with an industrial illumination and detection, such as polarization, incidence
robot that enables the characterization of extended optical angle θi , polar and azimuthal scattering angles (θs ; φs ), as
freeform surfaces are addressed. The characterized samples well as the dielectric function of the ideal interface.26 The
in this paper were fabricated and designed at Fraunhofer definitions of the scattering angles and the scattering relevant
IOF using approaches such as additive manufacturing, light- quantities are summarized in Fig. 1.
weight design, or subaperture polishing.3,24,25 Therefore, the For metallic surfaces and scattering angles close to the
specular direction, it is a good approximation to replace
*Address all correspondence to Tobias Herffurth, E-mail: tobias.herffurth@
iof.fraunhofer.de 0091-3286/2019/$28.00 © 2019 SPIE

Optical Engineering 092609-1 September 2019 • Vol. 58(9)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 09 May 2019


Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Herffurth et al.: Assessing surface imperfections of freeforms using a robotic light scattering sensor

Incident (a)
beam
ΔΩs Specular reflected beam
+ scaered light
θs
θi
ϕs

Fig. 1 Basic geometry for the definitions of specular and scattering


quantities.

Q by the reflectance of the sample.26 This also holds for (b)


some specific cases of multilayer surfaces with highly corre-
lated surface roughness10 for which otherwise multilayer scat-
tering models30,31 would have to be applied (see Sec. 4.2).
In Eq. (1), the roughness information is included as PSD
function, which represents the relative strength of roughness
component as a function of the spatial frequencies f x and
f y . The PSD is defined as the squared modulus of the
Fourier transform of a three-dimensional surface profile.
Integrating the PSD over a certain spatial frequency range
yields the bandwidth limited rms roughness:26,32
Z 2π Z f
EQ-TARGET;temp:intralink-;e002;63;481

max
σ ¼
2
PSDðfÞfdfdφ
0 f min
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Fig. 2 (a) Robotic light scattering sensor with (b) a scheme of the
with φ ¼ arctanðf y ∕f x Þ and f¼ f 2x þ f 2y : (2) optical setup: sample (1)/(7), sensor head (2), robot (3), laser source
(λ ¼ 650 nm) and optical fiber (4), spatial filter (5), polarizer (6),
sample (7), CMOS matrix detector (8).
Thereby also an azimuthally averaged PSD2Diso can be
defined, which enables comparison of PSD results in two-
dimensional diagrams. a clean beam on the sample; and a mirror, which images
In the context of scattering, spatial frequencies and scat- the pinhole of the spatial frequency filter onto the CMOS-
tering angles are linked by the grating equations: detector matrix. Thereby the angle of incidence onto the
sample is 18 deg. Calibrated ARS results are obtained with
sin θs cos φs − sin θi sin θs sin φs this setup using a white reflectance standard (Spectralon®) to
fx ¼ and fy ¼ : (3)
λ λ determine a signal proportional to the incident laser power.
EQ-TARGET;temp:intralink-;e003;63;358

The instrument signatures, represented by measurements


Hence, the spatial frequency limits of a scattering measure- of a superpolished silicon wafer, are shown for different
ment are defined by the illumination wavelength and the designs of the beam preparation system in Fig. 3, to empha-
accessible scattering angles. On the other hand, a PSD deter- size the challenges in designing (compact) scattering instru-
mined from scattering data provides averaged roughness ments. These measurements, which have to be analyzed over
information over the entire illumination spot, which is typ- several decades below the level of the specular reflected
ically 0.5 to 3 mm in diameter. beam, are the primary benchmark to assess the performance
of a scatterometer. The setups (a) and (b) are based on a plane-
3 Instrumentation convex lens, a low-cost laser diode module, and a 150-μm
pinhole, whereas setup (c) features a mirror as final imaging
3.1 Light Scattering Sensor element in the beam preparation system. In addition, the bet-
The main principle of the scattering sensor approach is based ter beam quality of the fiber-coupled laser diode module in
on a combination of a laser source, an illumination system, (c) enables a smaller pinhole of only 80 μm to be applied.
and a CMOS-detector matrix. Consequently, wavelengths Zemax was used to optimize the spot diagram on the
from about 370 to 950 nm could be used, depending on the detector and in combination with light scattering models
sensitivity of CMOS pixels. to identify contributions to the instrument signature caused
Figure 2(a) shows the current implementation of the light by component scattering or system internal reflections.
scattering sensor concept, presented in Refs. 7 and 19, as a These contributions are indicated in the figure. For configu-
compact sensor head in combination with an industrial ration (b), the plano-convex lens was tilted to reduce the con-
robot during the characterization of the hyperbolic mirror, tributions of back reflections between lens and pinhole to the
which will be discussed in Sec. 4.1. In part (b) of the figure, signature. Conversely, the increasing aberrations induced by
a scheme is given to introduce the primary sensor compo- this measure drastically increased the near-angular limit of
nents, which are a fiber-coupled laser diode module with the signature and, therefore, the capability for the roughness
a wavelength of 660 nm; a spatial frequency filter to obtain assessment at low spatial frequencies. However, setups (a)

Optical Engineering 092609-2 September 2019 • Vol. 58(9)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 09 May 2019


Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Herffurth et al.: Assessing surface imperfections of freeforms using a robotic light scattering sensor

Fig. 3 Instrument signatures obtained by measuring a silicon wafer for (a) lens-based, (b) a tilted lens-
based, and (c) a mirror-based setup of the beam preparation module. Zemax and light scattering models
of the imaging elements were applied to generate the spot diagrams and to identify the different
contributions to the instrument signature: pinhole re-reflex of the plane (1) or convex (2) lens side, bulk
scattering and internal reflexes of the lens (3), aberrations (4), scattering of the silicon waver (5), and
scattering from the mirror as final imaging element (6).

and (b) could be implemented with commercial optomechan- Δθs ¼ 0.008 deg and Δφ ¼ 0.025 deg. However, these val-
ical components. A compact implementation of setup (c) ues are slightly changing over the area of the CMOS-matrix.
with a low incidence angle of 3 deg onto the mirror for dif- In a future version of the scattering sensor, a piezo-driven
fraction-limited imaging of the pinhole onto the detector stage that moves the pinhole and the optics in front of the
required the development of a customized x − y-stage for pinhole will be implemented into the beam preparation sys-
pinhole alignment. tem. This is required to compensate the widening of the
The final instrument signature of the scattering sensor (c) specular reflex by the sample curvature, which would other-
reveals a sensitivity level of below 10−4 sr−1 and an angular wise result in misleading roughness results or a loss of
range that covers a cone angle of ±8 deg around the specular accessible spatial frequencies at the low bandwidth edge.
reflected beam in the center. In combination with the near Exemplary measurement results with a test setup using a
angular limit of 0.3 deg, the grating equations yield a spatial piezo in the beam preparation system are shown in Fig. 4.
frequency range from 0.008 to 0.3 μm−1 for roughness and The sample was a diamond-turned aspherical aluminum sur-
PSD determination, which is similar to a WLI with a 10× face with a curvature radius of 100 mm at the vertex.
objective. The corresponding rms roughness of the signature Even if the defocus induced by the sample is not compen-
is about 0.3 nm. sated, the PSD function obtained with the scattering sensor
Because of the large number and high density of demonstrates a good agreement to that from topography
CMOS-pixels (2080 × 2080), the angular resolution is about data, which also demonstrate the robustness of the scattering

Optical Engineering 092609-3 September 2019 • Vol. 58(9)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 09 May 2019


Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Herffurth et al.: Assessing surface imperfections of freeforms using a robotic light scattering sensor

move the sensor to the measurement positions on the sample.


(a) With refocus (b) Without refocus Thus far, RoboDK is also used for referencing the TCP as
well as the samples with respect to the base coordinate sys-
tem of the robot, which is the base to link the stl-model of
the sample to real-world coordinates.
One of the major concerns before the combination of
robot and sensor was related to the positioning uncertainty
of the robot, in particular, regarding sensor sample orienta-
tion. Whereas the TCP position of industrial robots was
highly reproducible with uncertainties of only ±0.1 mm, the
information given by robot suppliers regarding absolute
(c) positioning accuracy as well as orientation accuracy was
very limited. However, in experiments as those discussed
in Sec. 4, no reasonable deviations of the generated sensor
paths were observed. Possible deviations of sensor and
sample orientation remained below 0.5 deg.

3.3 Measurement Procedure, Accuracy, Speed, and


Spatial Resolution
At each measurement position in an area-covering surface
mapping, an image of the ARS is recorded from which the
local surface roughness is calculated. Areas influenced by
the specular reflex and parts of the instrument signature close
to specular reflex are excluded from the roughness calcula-
Fig. 4 Demonstration of the defocus and its compensation using a tion. While moving the sensor, the position of the specular
piezo stage for scattering and PSD measurement results on a curved reflected beam on the CMOS matrix is continuously tracked
aluminum sample. (a) and (b) Scattering distributions with and without to adapt the position of the excluded angular range and to
defocus compensation. The angular scaling of the scattering distribu- compensate for minor deviations of the measurement dis-
tions is 1 deg for θs and 5 deg for φs . (c) Comparison of 2Diso-PSD
results.
tance or the angular orientation of the robot. The width of
the excluded angular range stays constant to guarantee a con-
stant low spatial frequency limit for the comparability of the
approach. However, the near-angular limit and, therefore, the roughness results.
lowest spatial frequency with reasonable data, is reduced Modeling of the entire measurement process using Monte
to about 0.05 μm−1 . Compensating this defocus extends the Carlo methods according to the “guide to the expression of
low spatial frequency limit to 0.008 μm−1 —as it is for sam- uncertainty in measurement”34,35—in particular regarding
ples without curvature. After refocus, the diffraction peaks, this compensation and geometrical uncertainties—are sum-
caused by the diamond turning structure, become more marized in Fig. 5. The figure shows the different contribu-
defined and exhibit an excellent agreement to PSD from tions to the ARS, which are shown as function of the position
topography data as well. on the CMOS matrix for a measurement (high accuracy
mode) of an ideal mirror with a sensor-relevant roughness
of 1 nm. For the sample orientation deviation in this estima-
3.2 Robotic Handling Systems and Path Generation tion, Gaussian distributions with mean deviations of 2 mm
For the results presented in this paper, the scattering sensor is and of 0.5 deg for the working distance and for the polar
moved using a collaborative six-axis UR10 robot (Fig. 2), orientation, respectively, in combination with an equal dis-
produced by “Universal Robots A/S,” with a working radius tribution for the azimuthal orientation are assumed for the
of 1300 mm and maximum handling mass of 10 kg. Owing sample as well as for the calibration sample orientation.
to collaborative properties, the required robot-specific safety In addition to the CMOS noise, the highest contributions to
measures are reduced and it is possible to guide the scatter- the relative ARS uncertainty are caused by deviations of the
ing sensor by hand to representative positons on samples. sensor-sample distance and the angular orientation of the
The referencing effort for samples and for the sensor’s illu- sensor. However, the contributions are significantly reduced
mination point on the sample—which corresponds to the by tracking the specular reflex, recalculating the respective
tool center position (TCP)—is also reduced by the hand scattering coordinate system, and excluding the correct
guidance. section of the specular reflected beam.
The key feature for the generation of automated measure- Regarding the uncertainty of the roughness results
ment paths is the parallel orientation of the light scattering obtained from this modeling [Fig. 5(b)]: for an ideal sen-
sensor to the local sample normal in the correct working dis- sor-sample orientation, the uncertainty induced by laser and
tance of 13 mm. This path generation is achieved using func- detector noise is approximately 1.7% (k ¼ 2), whereas, it
tions provided by the software RobotDK33 and its python only increases to about 4.8% by taking into account the geo-
interface. Therefore, regular grids are projected onto a highly metrical contributions to the overall uncertainty. These con-
resolved stl-file of the sample and the intersection points and tributions are deviations of the working distance as well as
the local surface normals are calculated. Thereafter, a list of of the angular orientation of the sensor to the local surface
robot poses is generated that is used by the sensor software to normal.

Optical Engineering 092609-4 September 2019 • Vol. 58(9)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 09 May 2019


Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Herffurth et al.: Assessing surface imperfections of freeforms using a robotic light scattering sensor

measurement positions as well as for the instrument signa-


ture measurement.
The high accuracy mode consists of an image series with
multiple integration times, which is combined to form a sin-
gle calibrated scattering image for further evaluation. In this
mode, a dynamic range of about 5.5 decades can be covered.
However, recording and analyzing the data can take up to 2 s.
In the fast measurement mode, only a single image with
a fixed time and a dynamic range of only about 2.3 decades
(8-bit image including dark signal) is recorded and analyzed,
which provides enough information to assess roughness,
homogeneity, and defects. Only a single a-priori informa-
tion—a measurement at a clean position—is required to
identify the optimal exposure time for the measurement.
The fast measurement mode, only limited by the speed
of the camera, also enables an advanced regime with mea-
surements while the robot is continuously moving over the
surface. In this configuration, the local resolution in one
dimension is given by the density of the adjacent paths,
whereas the resolution in the other dimension is determined
by the ratio of current measurement rate and path velocity.
For example, for a speed of 10 mm/s and a frame rate of
20 Hz, the resolution or sampling distance would be 0.5 mm.
This mode enables measurement rates which are not limited
by the necessity to accelerate or decelerate the robot with the
frame rate of the CMOS sensor which is up to 24 Hz in full
resolution and even more than 60 Hz using a 2 × 2 hardware
binning. However, to exploit these fast acquisition rates,
parallelization of data record, transfer, and analysis and
smart measurement and analysis routines are currently under
development.
Results obtained from both movement approaches are
compared in Fig. 6. The resolution of both mappings is
about 0.5 × 0.5 mm with 165,000 single measurement points.
In the start–stop regime, the measurement did take 11 h,
whereas it was only 5¾ h for the continuous movements.
(The measurement rate was limited to about 8 Hz, as data rec-
ord and analysis is not yet parallelized.) To enable the continu-
ous mode, a ring-buffer software interface was implemented;
otherwise, the robot would not have been able to process the
number of target points (>2000) of a path without start–stop
movement.

3.4 Summary of Systems Parameters


A summary of the primary sensor and system parameters is
given in Table 1.

4 Application to Metallic Mirrors


Fig. 5 (a) Summary of different contributions to the relative uncertain- 4.1 Large Freeform Mirror
ties of the ARS as a function of the position on the CMOS matrix
obtained from Monte Carlo modelling. In the upper rows the contribu- The characterization of a hyperbolic primary mirror with a
tions of sensor-sample misalignment are shown for application with diameter of 600 mm for a Ritchey–Chrétien–Cassegrain
and without reflex tracking (please note the different color scaling). telescope is discussed in this section. The mirror form was
(b) Corresponding histogram for rms roughness obtained from the
modeled ARS distributions.
fabricated by diamond turning in aluminum.36 Thereafter, an
electroless nickel (NiP) coating was added and the mirror
was polished to remove the turning structure. The mapping
comprised about 94,000 single measurement positions and
To provide fast overview or detailed results of the sample was recorded in about 4 h in the continuous movement
scattering or its roughness, the sensor can be operated in mode with a resolution of ∼2 × 2 mm2 . In the mapping, the
two modes: a fast measurement mode and a high accuracy two concentric rings of increased roughness were the most
mode. The first is applied for the majority of raster scans prominent features, beside local defects and a slight increase
of surfaces, whereas the latter is applied for the single in the lower left.

Optical Engineering 092609-5 September 2019 • Vol. 58(9)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 09 May 2019


Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Herffurth et al.: Assessing surface imperfections of freeforms using a robotic light scattering sensor

Table 1 Summary of primary sensor parameters.

Comment

Angle of incidence 18 deg

Illumination wavelength 660 nm

Sensitivity <1 × 10−4 sr−1

Dynamic range 5.5 decades High accuracy


mode (image series)

Solid angle per pixel 2.1 × 10−8 sr Average

Angular range 8 deg Around reflex

Near-angular limit 0.3 deg

Signature corresponding 0.3 nm


roughness

Spatial frequency range 0.008 to 0.3 μm−1

ARS acquisition rate 0.5 per second High accuracy


mode (image series)

Up to 24 per Single shot


second measurement

the mapping. The corresponding scattering images (e.g., X4)


reveal a significantly different structure to those of the posi-
tions dominated by roughness. In addition to the increased
scattering, the fine speckle pattern of pure roughness is
superimposed by coarse scattering structures that exhibit
several minima and maxima.
These minima are also present for a fine scratch in the
image X3. Analyzing the angular distance of the minima
and the periodicity7,15 reveals an angular distance of the min-
ima of about 1 deg and, therefore, a width of the scratch of
about 106 μm.

Fig. 6 Comparison of mapping results obtained in (a) start–stop 4.2 Printed and Polished Metallic Surfaces with
movement and (b) continuous movement. Protective Coating
To achieve highly resolved roughness information over the
entire surface of a lightweight convex mirror,25 a raster scan
The raw scattering images from different positions, shown with a local resolution of 0.5 mm was performed (Fig. 9). At
in Fig. 7 as X1 to X4, are extracted from a screen capture each measurement position, an image of the ARS was
of the CMOS-sensor image while the robot is moving recorded (start–stop mode, single exposure time) from which
(Video 1). They reveal different levels of residual turning the local surface roughness was calculated. Thereby, the spa-
marks that change their orientation in correspondence to the tial frequency range, which was reduced by the widening of
local orientation of the turning structure (X1, X2). The angu- the specular reflex due to the sample curvature, was about
lar distance of the diffraction peaks to the specular reflex is f ¼ 0.03 to 0.3 μm−1 .
about 4 deg and corresponds to a turning frequency of f ¼ The mirror base body is designed using optimized cell-
0.1 μm−1 or a period of 10 μm. The increased scattering like internal support structures to realize a lightweight and
between the turning marks and perpendicular to the turning stiff design as well as a mass reduction of ∼65%. The
direction is induced by the jitter of the turning process. fabrication of this design is solely possible using additive
In addition to the scattering from turning marks, two com- manufacturing.24 The base material is a powder of aluminum
ponents of increased scattering with constant orientations silicon material with 40 weight % silicon (AlSi40) that is
from the top right to the bottom left in the scattering images processed with selective laser melting.3 The successive
as well as perpendicular to this direction can be observed. processing steps include diamond turning, plating with elec-
The corresponding roughness structure on the surface is a troless nickel, polishing processes, and a final coating using
result of the polishing process. protected silver.25
A chip that is carried over the surface during turning Although the surface is coated with a 110 nm Al2 O3 pro-
caused the two concentric rings of increased roughness in tection layer, the roughness calculations are performed using

Optical Engineering 092609-6 September 2019 • Vol. 58(9)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 09 May 2019


Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Herffurth et al.: Assessing surface imperfections of freeforms using a robotic light scattering sensor

1.010
F prot. Ag Q Ag
1.005
R prot. Ag R Ag

optical factor / reflectance


1.000 AOI scattering sensor angular limits
0.995

0.990

0.985

0.980

0.975

0.970
0 5 10 15 20 25 30 35
scattering angle / angle of incidence in °

Fig. 8 Comparison of the optical factors Q and F for single surface


and multilayer scattering theory, respectively, as a function of the
scattering angle. In addition, the corresponding reflectivity curves of
silver and protected silver are given as function of the incidence
angle together with the incidence angle and the angular limits of the
scattering sensor (s-polarization for all curves).

deviation is below 0.5%. Even if the reflectivity of a pure


silver surface had been applied, the deviation would be
below 1%.
The mapping result in Fig. 9 reveals an increasing rough-
ness from the lower left, with an average roughness of about
0.5 nm, to the top right, with an average of about 1 nm. (The
large number of singular increased roughness values corre-
sponds to particles, since the measurements are not per-
formed in a clean room environment.)
Exemplary scattering distributions are shown in the lower
part of Fig. 9 as well. The measurement positions X1 to X3
reveal that the increase of the scattering level is not caused by
contamination or defects, since the changing vertical and
nonisotropic scattering component shows the typical fine
granular speckle structure of the roughness. This is typically
not the case for localized particles or defects like that at posi-
tion X4.
Fig. 7 Roughness mapping (a) of a hyperbolic mirror with scattering The roughness-induced anisotropic scattering, which is
images (X1 to X4) that are obtained from a video with continuous sen- increasing from X1 to X3, corresponds to a stochastic but
sor movement (Video 1). Figure 2 shows a photograph of this sample
taken during the characterization. The angular scaling in the pictures
partially parallel structure on the surface. This structure
(X1 to X4) is 2 deg for θs and 10 deg for φs . (Owing to the limitations might be induced by the MRF polishing process, as, for
of the computer memory and the early state of software development example, discussed in Ref. 20. In addition, the ARS meas-
in this measurement, a direct correlation of sample position and scat- urement at position X3 reveals a very weak residual turning
tering image is not yet possible for this sample. Therefore, the position
labeled as X1 to X4 in the mapping are rough estimations. On the
structure on the surface, indicated as minor diffraction peaks
other hand, the correlation of roughness value and position in the in red circles, that is not removed by the polishing process.
mapping is valid.) (Video 1, .mp4, 10.5 MB [URL: https://doi.org/10 Moreover, a scattering component which should correspond
.1117/1.OE.58.9.092609.1]). to a parallel structure at low spatial frequencies is visible
close to the specular beam (labeled by green ellipses).
These effects additionally demonstrate that the roughness
a single surface approximation of the protective coating.10
in this spatial frequency region is dominated by roughness
Figure 8 shows a comparison of the optical factor Q follow-
components that are replicated from the substrate and, there-
ing Rayleigh–Rice for single surfaces [Eq. (1)] and the
fore, additionally justify the high level of roughness corre-
reflectivity of pure silver37 as well as the optical factor30,31
lation required for the application of the single surface
and the reflectivity31 for the protection layer. For the latter,
approximation.
a perfect roughness replication and, therefore, full roughness To identify the relevant surface structure and for a cross-
correlation are required for the relevant spatial frequency check, WLI topography data are recorded within the regions
range, which is typically fulfilled for thin coatings with low close to the positions labeled in Fig. 9. The results of X1 and
intrinsic roughness.30,38 X3 are shown in Fig. 10 together with 2Diso -PSD functions
The curves reveal that, for the roughness calculation, the determined from topography and scattering results.
optical factor can be approximated by the reflectivity of the First, the curves reveal an excellent agreement in the over-
protected surface while the additionally induced relative lapping regions. In addition, the topography data verify the

Optical Engineering 092609-7 September 2019 • Vol. 58(9)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 09 May 2019


Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Herffurth et al.: Assessing surface imperfections of freeforms using a robotic light scattering sensor

(a)

(b) 1E+9
WLI Sensor
X1
X1
1E+8
X2
X2
X3

4
X3

PSD in nm
1E+7

1E+6

1E+5

1E+4
0.01 0.1 1
spatial frequency f in µm-1

Fig. 10 (a) Topography data of position X1 and X3 from WLI with


objectives 10× and 50× with measurement areas of 698 × 523 μm2
and 140 × 105 μm2 , respectively. Comparison of 2Diso-PSD func-
tions obtained by WLI with 50× objective and scattering sensor for
the measurement positions X1 to X3.

The scatter map results of both mirrors, which are


recorded in a single run (start–stop regime, single exposure),
are summarized in Fig. 11. The different polishing processes
applied to the optical mirror surfaces result in an rms rough-
ness of only 0.4 nm (WLI 50X), which is below the sensi-
Fig. 9 Light scattering-based roughness analysis of a lightweight con- tivity limit of the scattering sensor and leads to a significant
vex aspherical mirror. (a) Roughness mapping obtained from light influence of the instrument signature to the single ARS dis-
scattering, together with a photograph of the sample. (X1) to (X4): light tributions. Nevertheless, the results demonstrate that the
scattering distribution recorded at four different positions on the roughness, within the spatial frequency range of the sensor,
sample.
is below 0.35 nm on both surfaces. On the other hand, the
ARS distributions at some positions, for example position
change of roughness over the surface, in particular, at higher M4 X3, reveal a residual roughness structure close to the
spatial frequencies (50× objective) as well as the parallel specular direction which corresponds to the incomplete
grove structure visible in the scattering distributions close removal of turning marks in the polishing process. Therefore,
to the specular reflex (10× objective). On the other hand, the even for surfaces with lowest roughness, valuable informa-
structure with a period of about 5 μm, corresponding to the tion on defects, disturbing roughness components or fulfil-
scattering feature at position X3 (red label) is, however, not ment of roughness limits can be assessed.
visible in the WLI images.
5 Summary and Outlook
4.3 Mirrors with Lowest Roughness The results presented in this paper demonstrate the enormous
The M2M4 assembly is fabricated as a component with two potential of the surface assessment of freeform optics using
mirrors on a single lightweight carrier structure to be used the robotic light scattering sensor. This is demonstrated for
within an afocal anamorphic imaging telescope for VIS appli- single opaque surfaces as well as for a protected silver coat-
cations. M2 and M4 are biconic aspheres with higher-order ing using a robot-guided sensor setup. Regarding roughness,
correction terms and deviations from a rotational symmetry of the comparability of the results to those of a WLI is demon-
about a few millimeters each. Details on the fabrication proc- strated and the influence of a protective layer is discussed.
ess and the optical design can be found in Ref. 2. In addition to the pure scattering information and single

Optical Engineering 092609-8 September 2019 • Vol. 58(9)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 09 May 2019


Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Herffurth et al.: Assessing surface imperfections of freeforms using a robotic light scattering sensor

Fig. 11 Roughness analysis of two mirrors on the same carrier structure. (a) Photography of the
sensor and the sample, (b) roughness mapping, and (X1 to X3) ARS distributions of three different
positions.

roughness numbers, valuable information on the different References


roughness components of anisotropic surfaces is assessable. 1. M. Beier et al., “Fabrication of high precision metallic freeform mirrors
In particular, the footprints of the different processes can be with magnetorheological finishing (MRF),” Proc. SPIE 8884, 88840S
identified and analyzed. (2013).
2. M. Beier et al., “Large aperture freeform VIS telescope with smart align-
With the sensor setup, measurement frequencies of up ment approach,” Proc. SPIE 9912, 99120Y (2016).
20 Hz and sensitivities down to 0.5 nm in a spatial frequency 3. N. Heidler et al., “Additive manufacturing of metal mirrors for TMA
range from 0.008 to 0.3 μm−1 can be achieved. Moreover, telescope,” Proc. SPIE 10692, 106920C (2018).
4. S. Risse, M. Beier, and J. Hartung, “Aspherical and freeform mirrors
the light scattering-based approach is robust and fast enough based on ultra-precise manufacturing for telescopes in the VIS spectral
that even measurements with a moving sensor system are range,” Proc. SPIE 10562, 105625T (2017).
5. F. Z. Fang et al., “Manufacturing and measurement of freeform optics,”
enabled. CIRP Ann. Manuf. Technol. 62(2), 823–846 (2013).
In future developments, the implementation of a piezo 6. P. Dumas et al., “Complete sub-aperture pre-polishing and finishing sol-
stage into the robotic light scattering sensor, as well as its ution to improve speed and determinism in asphere manufacture,” Proc.
SPIE 6671, 66711 (2007).
automation, is envisaged. The positive influence on the low 7. T. Herffurth et al., “Comprehensive nanostructure and defect analysis
spatial frequency limit of the sensor, which is discussed in using a simple 3D light-scatter sensor,” Appl. Opt. 52(14), 3279–3287
this paper, would thus be available for samples with higher (2013).
8. S. Schröder et al., “Angle-resolved scattering: an effective method for
variability of curvature compared to those presented in this characterizing thin-film coatings,” Appl. Opt. 50(9), C164–C171
work. A further speed-up of the measurement rates should be (2011).
9. A. V. Finck et al., “Characterization of optical coatings using a multi-
achievable by parallelization of movement, measurement, source table–top scatterometer,” Appl. Opt. 53(4), A259–A269
and result calculation within the software as well as by using (2014).
the binning mode of the camera to reduce the amount of 10. S. Schröder et al., “Spectral angle resolved scattering of thin film coat-
ings,” Appl. Opt. 53(4), A35–A41 (2014).
transferred scattering data. 11. M. Trost et al., “Structured Mo/Si multilayers for IR-suppression in
Another topic is the development of software modules laser-produced EUV light sources,” Opt. Express 21(23), 27852–27864
for the improved analysis of the scattering images regarding (2013).
12. S. Schröder et al., “Bulk scattering properties of synthetic fused silica
specific features of the different processes (diamond turning, at 193 nm,” Opt. Express 14(22), 10537–10549 (2006).
MRF polishing, conventional polishing). Moreover, the de- 13. M. Trost et al., “Evaluation of subsurface damage by light scattering
techniques,” Appl. Opt. 52(26), 6579–6588 (2013).
velopment of software modules to automatically detect and 14. T. A. Germer and C. C. Asmail, “Polarization of light scattered by
classify defects is investigated. microrough surfaces and subsurface defects,” J. Opt. Soc. Am. A 16(6),
1326–1332 (1999).
15. S. Maure, G. Albrand, and C. Amra, “Low-level scattering and localized
defects.” Appl. Opt. 35(28), 5573–5582 (1996).
Acknowledgments 16. Z.-H. Gu and J. M. Bennett, “Enhanced backscattering from very
smooth metal surfaces,” Proc. SPIE 2541, 45–53 (1995).
The authors would like to thank Luisa Coriand, Sebastian 17. J. E. Harvey, “Scattering effects in x-ray imaging systems,” Proc. SPIE
Korten, and Nadja Felde for their helpful discussions. 2515, 264–272 (1995).

Optical Engineering 092609-9 September 2019 • Vol. 58(9)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 09 May 2019


Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
Herffurth et al.: Assessing surface imperfections of freeforms using a robotic light scattering sensor

18. J. E. Harvey et al., “Calculating BRDFs from surface PSDs for mod- Fraunhofer Institute for Applied Optics and Precision Engineering
erately rough optical surfaces,” Proc. SPIE 7426, 74260I (2009). (IOF) in Jena, Germany, working on methods and applications for light
19. S. Schröder et al., “Device and method for angularly resolved scattered scattering-based surface, defect, and thin film characterization.
light measurement,” Patent WO 002, 010, 127, 872 (2010).
20. M. Trost et al., “In situ and ex situ characterization of optical surfaces by
light scattering techniques,” Opt. Eng. 53(9), 092013 (2014). Marcus Trost graduated in physics in 2009 and received his PhD
21. A.-S. Munser, “Streulichtbasierte Detektion pathogener Keime,” in from the Friedrich–Schiller-University, Jena, in 2015. Since 2008,
DGaO Annual Meeting, Dresden (2017). he has been with the surface and thin film characterization group
22. A.-S. Munser et al., “Towards label-free characterization of biological at the Fraunhofer IOF Jena. His interests are directed to the study
samples using angle-resolved light scattering: from single cells to cell of optical surfaces, thin film coatings as well as optical systems with
aggregates” submitted to Biomedical Optics Express (2019) particular focus on nanostructures and microstructures, subsurface
23. S. Schröder et al., “Origins of light scattering from thin film coatings,” damage, and light scattering, as well as to the development of new
Thin Solid Films 592, 248–255 (2015).
24. E. Hilpert et al., “Precision manufacturing of a lightweight mirror body measurement and modeling techniques.
made by selective laser melting,” Precis. Eng. 53, 310–317 (2018).
25. E. Hilpert et al., “Design, additive manufacturing, processing, and char- Matthias Beier works as a scientist in the Precision Engineering
acterization of metal mirror made of aluminum silicon alloy for space Department at Fraunhofer IOF. He received his diploma (Dipl.-Ing.)
applications” Opt. Eng. 58(9) (2019). in precision engineering from the Technical University of Dresden
26. J. C. Stover, Optical Scattering: Measurement and Analysis, 3rd ed., in 2012 and his PhD in applied physics (Dr.-Ing.) from Friedrich-
SPIE, Bellingham, Washington (2012). Schiller-University of Jena in 2017. Currently, he is leading the
27. A. Duparré, “Scattering from surfaces and thin films,” in Encyclopedia
of Modern Optics, B. D. Guenther, D. G. Steel, and L. Bayvel, Eds., ultra-precision manufacturing research group at Fraunhofer IOF with
Elsevier, Amsterdam (2004). a focus on the fabrication and metrology of freeform optical compo-
28. E. L. Church, H. A. Jenkinson, and J. M. Zavada, “Relationship nents and systems.
between surface scattering and microtopographic features,” Opt. Eng.
18(2), 125–136 (1979). Ralf Steinkopf is a graduate engineer and works at the precision
29. J. M. Bennett and L. Mattsson, Introduction to Surface Roughness and engineering department of the Fraunhofer IOF, Jena. His field of work
Scattering, 2nd ed., Optical Society of America, Washington D.C and his current research is centered on developing technologies for
(1999).
30. P. Bousquet, F. Flory, and P. Roche, “Scattering from multilayer thin the manufacturing of metal optics. In particular, freeform optics and
films: theory and experiment,” J. Opt. Soc. Am. 71(9), 1115–1123 structured surfaces are his fields of work.
(1981).
31. P. Bussemer, K. Hehl, and S. Kassam, “Theory of light scattering from Nils Heidler graduated in precision engineering in 2008 and received
rough surfaces and interfaces and from volume inhomogeneities in an his PhD in mechanical engineering from the Technische Universität
optical layer stack,” Waves Random Media 1(4), 207–221 (1991). Ilmenau in 2015 under the supervision of Professor René Theska. His
32. A. Duparré et al., “Surface characterization techniques for determining current research interests include additive manufacturing for optical
the root-mean-square roughness and power spectral densities of optical
components,” Appl. Opt. 41(1), 154–171 (2002). applications and precision mechanical systems.
33. Software RoboDK, by RoboDK inc., https://robodk.com/index
(Accessed 14 Dec 2018) Tom Pertermann studied physics at the Friedrich Schiller University
34. Working Group 1, Joint Committee for Guides in Metrology (JCGM/ (FSU), Jena, and graduated in 2017. Currently, he is a PhD student
WG 1), Uncertainty of measurement - part 3: Guide to the expression of at the Precision Engineering Department at Fraunhofer IOF. His
uncertainty in measurement,” (GUM:1995, ISO/IEC Guide 98-3:2008) research focus lies in the evaluation of surface errors of diamond-
(2008). turned freeform mirrors, with the aim to reduce the surface waviness.
35. T. Herffurth, “Light scattering and roughness analysis of optical surfaces
and thin films,” PhD Thesis, Friedrich-Schiller-Universität, Jena (2008)
36. R. Steinkopf et al., “Metal mirrors with excellent figure and roughness,” Sven Schröder is the head of the surface and thin film characteriza-
Proc. SPIE 7102, 71020C (2008). tion group in the optical systems department of Fraunhofer IOF in
37. M. N. Polyanskiy, “Refractive index database,” https://refractiveindex Jena. He graduated in physics from FSU in Jena, in 2004 and
.info/ (Accessed 14 Dec 2018). received his PhD (Dr. rer. nat.) from the FSU in 2008. Since 2001,
38. M. Trost et al., “Scattering reduction through oblique multilayer dep- he has been with the surface and thin film characterization group at
osition,” Appl. Opt. 53(4), A197–204 (2014). the Fraunhofer IOF. In 2010, he spent one year at the University of
Central Florida/CREOL in Orlando, Florida, to work on surface rough-
Tobias Herffurth graduated in physics in 2008 and received his PhD ness and scattering models. His interests are directed to the study
from Friedrich–Schiller-University, Jena, in 2015. Since 2006, he has of roughness, light scattering, and functional properties of surfaces
been with the surface and thin film characterization group at the and thin film coatings.

Optical Engineering 092609-10 September 2019 • Vol. 58(9)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 09 May 2019


Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

Das könnte Ihnen auch gefallen