Home Search Collections Journals About Contact us My IOPscience

Scanning probe microscope dimensional metrology at NIST

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

2011 Meas. Sci. Technol. 22 024001

(http://iopscience.iop.org/0957-0233/22/2/024001)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 78.170.81.15
The article was downloaded on 19/01/2011 at 21:03

Please note that terms and conditions apply.

IOP PUBLISHING MEASUREMENT SCIENCE AND TECHNOLOGY
Meas. Sci. Technol. 22 (2011) 024001 (11pp) doi:10.1088/0957-0233/22/2/024001

Scanning probe microscope dimensional

metrology at NIST
John A Kramar, Ronald Dixson and Ndubuisi G Orji
National Institute of Standards and Technology, Gaithersburg, MD 20899, USA
E-mail: [email protected]

Received 1 April 2010, in final form 9 June 2010

Published 21 December 2010
Online at stacks.iop.org/MST/22/024001

Abstract
Scanning probe microscope (SPM) dimensional metrology efforts at the US National Institute
of Standards and Technology (NIST) are reviewed in this paper. The main SPM instruments
for realizing the International System of Units (SI) are the Molecular Measuring Machine, the
calibrated atomic force microscope and the critical dimension atomic force microscope. These
are optimized for long-distance measurements, three-dimensional measurements over
conventional SPM distances and critical dimension or linewidth measurements, respectively.
10 mm distances have been measured with the relative standard uncertainty, uc , of 1.5 × 10−5 ;
step heights at the 100 nm scale have been measured with the relative uc of 2.5 × 10−3 and
sub-micrometer linewidths have been measured with uc = 0.8 nm.
Keywords: SPM, AFM, CD-AFM, traceable dimensional metrology, nanometrology,
linewidth, pitch, step height, uncertainty budget
(Some figures in this article are in colour only in the electronic version)

1. Introduction In this paper, we describe the SPM dimensional

metrology work at the US National Institute of Standards
Since the invention of scanning probe microscopy (SPM), and Technology (NIST). We will focus on three efforts: the
its great potential for dimensional metrology has been Molecular Measuring Machine (M3 ), the calibrated atomic
recognized, but the formidable challenges for achieving force microscope (C-AFM) and the critical dimension AFM
accurate measurements with SPM have also been apparent (CD-AFM). These are the SPM instruments at NIST that
[1–3]. The ability to generate real-space, atomic-resolution directly realize the unit of length according to an International
images of surfaces opened the door to the tantalizing prospect System of Units (SI) best practice, utilizing interferometers
of reaching the quantum limit in dimensional metrology— as the metric; in the case of the CD-AFM, it is directly
a single atom. Conversely, the possibility of using the traceable to the realized SI unit with a rigorous uncertainty
atomic crystal lattice as a measurement artifact was also analysis. Other dimensional measurements have also been
recognized. However, early scanning tunneling microscopes performed at NIST using SPMs that have been calibrated using
(STM) did not incorporate position or displacement sensors in transfer standards. Notable among these efforts is the work
their scan systems, and they relied on notoriously nonlinear on dimensional metrology of nanoparticles using a fluid AFM.
and hysteretic piezoceramics for motion generation. General This work is the subject of another paper in this special feature.
practice was to roughly calibrate the scan system, often based
on the known characteristics of the sample. Schemes were also 2. Molecular Measuring Machine
put in place to linearize the response of piezoceramics. Soon,
national measurement institutes (NMIs) around the world M3 [4, 11] was historically the first effort at NIST to develop
began to develop metrological SPM instruments to realize the a metrological SPM. It was conceived as an ambitious project
metrology potential [4–10]. Many commercial ventures also to realize the potential of atomic-resolution imaging for
began to offer SPMs with displacement sensors of various metrological purposes, both for making measurements with
types on the motion axes to bring more metrological rigor to atomic precision and accuracy, and for mutual validation of
the SPM imaging process. the metrology instrument through comparison with the ordered

0957-0233/11/024001+11$33.00 1 © 2011 IOP Publishing Ltd Printed in the UK & the USA
Meas. Sci. Technol. 22 (2011) 024001 J A Kramar et al

Figure 1. Cut-away view drawing of M3 . For scale, the core structure is about 0.35 m in diameter.

atomic structure and validation of the atomic structure as an The metric is the 633 nm wavelength of a He–Ne
intrinsic calibration artifact. The aim was to achieve atomic- laser, traced to an I2 -stabilized He–Ne laser (a recommended
scale resolution and accuracy over technologically relevant radiation for the realization of the meter in the visible). It
areas; specifically to achieve 1 nm or better uncertainty for is applied to measure the stage motions using a Michelson
point-to-point measurements over a 50 mm × 50 mm area, a interferometer of a dual-pass, inside, differential optical path
relative measurement uncertainty of about 1 × 10−8 . configuration with polarization-based routing (figure 2). The
interferometer optics are carried on the same motion stage
2.1. Design as the SPM probe, and the motion of the optics is measured
differentially relative to the opposite inside mirrored surfaces
The design of M3 is in itself an attempt to apply the principles of a metrology reference frame box that contains the sample.
of precision instrument design to a challenging specific test Fringe interpolation is done using the heterodyne method.
case. Fundamental principles of kinematic mounting, motion Only the X and Y axes are measured with interferometers. The
repeatability, structural loop, symmetry, metrology reference vertical Z-axis has an integrated capacitance gage sensor that
frame, the Abbe principle, etc are all explicitly considered in measures the fine motion of the SPM tip holder relative to the
the design. Compromises and trade-offs between the different vertical coarse motion stage. The relative configuration of the
principles are of course necessary since the implementation metrology axis and motion generation systems is illustrated in
demands of the different principles often conflict, not to block diagram form in figure 3.
mention cost constraints that are always an engineering reality. The range and resolution requirements for the motion
The design can be discussed in terms of the probe, metric, generation system are met by carrying fine motion stages
motion system and the metrology reference frame, which on coarse motion stages in all three axes (figure 3). The
includes environmental isolation and control. An overview X and Y stages are not stacked on each other, but move
drawing of the instrument is presented in figure 1. independently on V slideways built into the machine core,
The probe is an SPM. In initial design and implementation, sliding on kinematically located pads. The fine motion
it was furthermore a STM, although an exchangeable AFM stages in all three axes are flexure-guided and piezoceramic-
probe of the tuning-fork sensor design has now been added. driven through cross-axis-motion decoupling links, with an
An STM is in some ways the simplest SPM to implement and is approximately 10 μm range. The Z-axis was a particular
perhaps the most convenient for achieving atomic resolution, challenge due to the required compactness, and required a
but it does have the disadvantage of being limited to electrically motion amplifying flexure in the drive train to achieve the
conductive samples. desired range [12]. The coarse motion stages are driven

2
Meas. Sci. Technol. 22 (2011) 024001 J A Kramar et al

Figure 2. The X-axis interferometer configuration for M3 . The metrology mirrors are opposite sides of a box that contains the sample. The
interior optical components are a single assembly (shown exploded) that moves with the tip along the Y-axis while the sample moves with
the mirrors in the X-axis. The fine-beam-steering wedges below the half-wave plate are not used.

made from optically contacted ultra-low expansion coefficient

M3 Machine Core Structure material. The metrology box is carried on kinematic supports.
Key components linking the interferometer optics and Z stages
are made of Invar. In addition, a temperature control system
Y Coarse Motion
maintains stability at a few-millidegree level. Vibration
isolation includes two pneumatic isolator stages and an active
Y Fine Motion
vibration isolation system [13, 14]. Also, the machine core
and interferometry paths are housed in a high vacuum chamber
Capacitance Gage
for stability of the probe sample interaction and stability of the
X and Y
Interferometers
laser wavelength. The combination of these environmental
SPM isolation and control elements is critical for providing the
reference frame and sample dimension stability necessary
X Fine Motion
Flexures for meaningful measurements at this level of resolution and
X Coarse Motion
uncertainty.
Slideway
The instrument control system is assembled from com-
mercial components with in-house developed software. The
components include time interval analyzers for reading
Figure 3. Block diagram representation of the M3 motion guides
the interferometers, digital signal processors for interpreting
and basic metrology system, illustrating the stage stack up, the
mechanical structural loop, and the metrology loop. In this the interferometer data and for servo controllers, and custom
schematic view, only the X-axis interferometer is illustrated. interface hardware. Maintaining and updating the hardware
and software over the span of the project, for compatibility
using piezoceramic stepper motors. Extensive decoupling with rapidly advancing information technology, has proved to
of parasitic cross motions was implemented for the X and be a significant challenge.
Y coarse-motion drives. For the Z coarse motion, cross-axis
decoupling was not implemented, but is less necessary since
2.2. Uncertainty considerations and methodology
the Z coarse-motion stage is stationary when measurements
are being made. In general, the standard approach [15] to uncertainty budgets
A stable metrology reference frame is critical for accurate adopted by NIST [16] is to develop an estimated contribution
metrology. The basis of the reference frame for M3 is for every known source of uncertainty in a given measurement
the mirrored metrology box, previously mentioned, which is and to include terms pertaining to both the instrument used

3
Meas. Sci. Technol. 22 (2011) 024001 J A Kramar et al

and the particular specimen measured. Terms evaluated

exclusively by statistical methods are known as type A
evaluations. Other terms, known as type B evaluations, are
evaluated using some combination of measured data, physical
models or assumptions about the probability distribution.
All of these terms are then added in quadrature to obtain a
combined standard uncertainty, uc , for the measurement. For
reporting measurement results, this is usually multiplied by a
coverage factor k to obtain a combined expanded uncertainty,
Uc [16]. The most common coverage factor used is k = 2,
which would correspond to approximately 95% confidence for
a normal (Gaussian) distribution. The same basic approach
is used for all traceable measurements at NIST, and the
implementation of this approach for the three instruments
described in this paper is very similar. All measurements Figure 4. M3 measurement of the (TEET)[Ni(dmit)2 ]2 cleavage
plane, a = (1.09 ± 0.06) nm, b = (0.70 ± 0.01) nm, γ = 106◦ ± 3◦
reported here are expanded uncertainties with k = 2 unless
(statistical standard uncertainty only). For comparison,
otherwise noted. crystallography measurements have yielded a = 1.020 nm,
b = 0.751 nm, γ = 105.5◦ .
2.3. Measurements and uncertainties using M3
Table 1. Uncertainty budget for M3 measurement of LFAD Cr
3
The capability of M for long-range, high-precision and grating, 1 mm line spacing (k = 1).
atomic-resolution measurements has been demonstrated. For Component Standard uncertainty
validation of long-range measurements, a comparison was
made with the NIST line scale interferometer (LSI) through the Type A (repeatability, reproducibility, Observed SD
common measurement of a NIST standard reference material sample non-uniformity, etc)
Type B
SRM 2800 pitch calibration artifact. At the largest measured Metric (λHeNe in vacuum) 1 × 10−8 La
line spacing of 10 mm, the LSI measurement was 39 nm over Interferometer cosine error 2 × 10−5 L
nominal with Uc = 4 nm, and the M3 measurement was 60 nm Interferometer periodic nonlinearity 2 nm
below nominal with Uc = 300 nm (relative Uc = 3 × 10−5 ), Abbe errors (stage tilt is roughly 1 × 10−5 L
agreement comfortably within the uncertainty bounds. To proportional to L)
carry out the comparison, it was necessary to coat the artifact Z-to-X coupling uncertainty 1.5 × 10−5 H b
Temperature fluctuation uncertainty 5 nm
with a palladium–gold alloy after the LSI measurements, since Sample alignment cosine error 1 × 10−8 L
M3 was using an STM probe. Line center determination 4 nm
A significant body of work has been done in the Combined standard uncertainty, uc = [(SD)2 + (6.7 nm)2 +
measurement of grating artifacts. M3 has provided, within (1.5 × 10−5 H)2 + (2.2 × 10−5 L)2 ]1/2 = 34 nm
NIST, a unique instrument for measuring the long-distance SD = standard deviation.
grating pitch when the line spacing is below the Rayleigh a
The uncertainty is proportional to the measured value of
criterion for optical microscopes. Reference artifacts of length, L.
b
200 nm and 400 nm nominal pitches were measured as The uncertainty is proportional to the height difference
between the two end points, H, which is setup dependent.
traceability standards for the manufacture of gratings for the
x-ray transmission grating spectrometers for the space-based
Chandra x-ray Observatory. Both were measured over a and the average pitch was measured to be 212.69 nm ±
distance of 10 mm, tracking the probe tip over the entire 0.01 nm. The discrepancy between the measurements and
distance. Measurements along the grating lines were also the uncertainty estimates has not yet been resolved. The
necessary in order to accurately characterize the orientation of uncertainty budget for this measurement, presented in table 1,
the grating within the M3 coordinate system so as to ensure is based on the 1 mm maximum line spacing measured, which
that the pitch measurements were aligned with the grating is then divided by the number of lines to yield the average pitch
direction. The average line spacings were determined to be uncertainty.
200.01 nm ± 0.01 nm and 400.80 nm ± 0.02 nm, a relative Atomic-resolution metrology has also been achieved with
Uc of 5 × 10−5 for both measurements. M3 . As an example, we present in figure 4 a measurement of
Another significant grating artifact measurement was that the surface crystal lattice parameters for an organic conductor,
of a laser-focused atomically deposited chromium grating. (TEET) [Ni(dmit)2 ]2 . As for all M3 measurements, these data
The artifact was made by depositing chromium atoms onto were acquired while scanning the sample in X and Y stages
a surface through a standing wave of laser light that is tuned under a closed-loop servo control using the interferometers as
near an atomic transition where the dipole interaction is strong the displacement sensors. The measured lattice parameters
[17]. The expected pitch is 212.78 nm ± 0.02 nm, based are as much as 8% different than the expected values based on
on the laser frequency, geometric and thermal considerations, bulk crystal x-ray diffraction measurements. The measured
etc. With M3 , a 1 mm length along the grating was imaged, periodic nonlinearity of the M3 interferometers of up to 2 nm

4
Meas. Sci. Technol. 22 (2011) 024001 J A Kramar et al

over the 80 nm optical fringe period is sufficient explanation

for a short-range scale error of this magnitude. We believe that
this source of error can be mapped and significantly reduced,
especially in the context of vacuum interferometry.
The ambitious measurement performance goal for M3 of
1 nm uncertainty for 70 mm point-to-point measurements
using an SPM probe has not yet been achieved. Nevertheless,
a significant measurement instrument has been developed that
has demonstrated 3 × 10−5 relative expanded uncertainty
for pitch measurements using an SPM probe, and is capable
of atomic-resolution measurements and scanning under the
interferometer-based servo control.

3. Calibrated atomic force microscope

The C-AFM was conceived as a complementary metrology

instrument to M3 . Its performance goals were somewhat
Figure 5. Mechanical design of a C-AFM scanner and metrology
more modest: a scan range commensurate with typical frame. (The AFM head is not shown.)
AFMs and operation in an ambient environment instead of
vacuum. Measurement and validation relative to crystal lattice
measurements was not an explicit goal. From its inception, of thermal expansion (CTE) materials: Invar and Zerodur. The
however, the C-AFM incorporated traceable Z-axis (height) sensitivity coefficient of the cap gauge is calibrated as needed
metrology. in an offline fashion using an interferometer head that mounts
kinematically on the metrology frame in place of the AFM
3.1. Overview and design head.
In the current generation of the C-AFM, the lateral stage is
The NIST C-AFM is a custom AFM constructed by NIST a six-axis scanner with integrated capacitance sensors allowing
to perform traceable dimensional metrology. It is intended real-time control of six degrees of freedom (three translational
primarily to calibrate physical standards for other AFMs. The and three angular). The Z-axis resonant frequency of the
design, performance and uncertainties of the system have been stage, however, is too low for use in scanning, so we are still
discussed elsewhere [5, 18–20]. The C-AFM has metrology employing an independent Z-axis stage as was done in the
traceability via the 633 nm wavelength of an I2 -stabilized He– original design.
Ne laser for all three axes, accomplished using heterodyne The stage was supplied with a stand-alone programmable
laser interferometers. The lateral axes are controlled closed digital controller that allows for closed-loop operation
loop using interferometry. The Z-axis uses a capacitance gauge using the integrated capacitance sensors. Since we use
for real-time displacement metrology, and this cap gauge is interferometers and our AFM scan controller to independently
calibrated offline using interferometry. The C-AFM is capable close the loop, we operate the stage itself open-loop with
of performing both pitch and step height measurements, and it respect to the cap gauges for the X and Y axes. However,
has been used by NIST to perform both types of measurements by using the closed-loop control for the other four degrees of
for external customers. freedom, we are able to reduce the undesired angular motion
The basic mechanical design of the scanner and metrology of the stage and the resulting lateral axis Abbe errors by
frame is shown in figure 5. A scanning sample design three orders of magnitude relative to prior generations of the
was chosen, largely for the ease of interferometry integration instrument [18,19]. We were also able to reduce the out-of-
with low Abbe offsets—a few millimeters in this case. The plane (i.e. the Z-straightness of the X and Y axes) motion error
composite scanner consists of an X–Y flexure stage and an to less than 1 nm over the almost 100 μm lateral scan range.
independent Z-stage with an integrated capacitance sensor A commercially available heterodyne displacement
to which the specimen platform is attached. This scanner measuring interfometer system in a double pass configuration
displaces the specimen platform relative to the metrology is used in the C-AFM. For system integration, the lateral
frame. Both the lateral interferometer optics and the AFM interferometer optical assemblies were removed from the
head are mounted kinematically to the metrology frame. This original housing and mounted on custom platforms that were
design permits the lateral axis interferometry to be done real machined out of Invar for drift mitigation. These platforms
time, allowing the option of closed-loop position control in the also have the kinematic mounts for placing the interferometer
lateral axes. assemblies on the metrology frame.
The Z-axis represents a design compromise because the The AFM controller is a modified version of a system that
capacitance gauge displacement metrology is not directly was available commercially in the mid-1990s. The major
referenced to the metrology frame. The resulting longer modifications were (1) inclusion of custom circuit boards
metrology loop carries the risk of greater drift susceptibility, to read directly from the interferometer electronics, which
but this is partially mitigated through the use of low coefficient allows the AFM controller to run a closed loop using the

5
Meas. Sci. Technol. 22 (2011) 024001 J A Kramar et al

History of C-AFM Pitch Measurement Uncertainties

0.02 400 nm and 600 nm pitch
(Customer A)
0.018 1 x 10-3

0.016 100 nm pitch

Relative Expanded Uncertainty

for Au nano-particle
200 nm pitch SEM calibration
0.014
(Customer B) 2000 nm pitch
0.012 (Customer C)

0.01
International
0.008 Comparison
International
Nano4
0.006 Comparison Ongoing
(290 nm and 700nm)
Nano5 70 nm pitch
0.004 (1000 nm and 290 nm) Comparison

0.002

0
Measurement

Figure 6. History of C-AFM pitch measurement expanded uncertainties, expressed as a fraction of the measured value. Recent performance
has surpassed the level of 1 × 10−3 .

interferometers and (2) modification of output voltage stages refined. It is also important to recognize that measurement
to match the input voltage ranges of the custom scanner. uncertainties are always a function of both the instrument and
There were many advantages to the modular design of the sample being measured. The roughness and uniformity
the C-AFM system and the use of commercially available of the sample features will often limit the achievable
subsystems with minor modifications. However, the rapid measurement uncertainties. In figures 6 and 7, a sampling
evolution of commercial SPM technology since the project of the history of C-AFM pitch and height measurement
inception and the resulting obsolescence of many of our uncertainties is shown. This history includes NIST internal
component systems have presented us with challenges. measurements, measurements for external organizations and
Consequently, one of our goals for the future is to reduce measurements performed as part of several international
reliance on the commercial subsystems by exploiting the comparisons.
programmable capabilities of our X–Y stage more fully. By Since the late 1990s, there has been an international
migrating the lateral scan motion generation from the AFM effort underway to establish the ‘metrological equivalency’
controller to the stage controller, we will be able to replace the of some types of measurements performed by the NMIs
original AFM controller. This would mean using the lateral cap of various nations. On the technical side, a series of
gauges for closed-loop control rather than the interferometers. ‘preliminary key comparisons’, coordinated by the CIPM
But since the interferometers would still be available for real- (Comité International des Poids et Mesures) are being
time monitoring, metrological accuracy would still be assured. conducted among NMIs including NIST. On the legal side,
a mutual recognition arrangement (MRA) has been negotiated
3.2. Measurements and uncertainty between the participating NMIs. NIST has now participated
in three such comparisons involving AFM dimensional
The general approach to uncertainty described in section 2.2 metrology.
for M3 is also used for the C-AFM. In the current generation These comparisons involved measurements of one-
of the C-AFM system, the performance level on the relative dimensional pitch, step height and two-dimensional pitch
standard uncertainty uc of pitch and step height measurements and were referred to as NANO4, NANO2 and NANO5,
are generally on the order of 1.0 × 10−3 for pitches in the respectively. The results of these comparisons have now been
100 nm to 1 μm scale and step heights at the 100 nm scale. published [21–23]. A C-AFM image of one of the NANO5
More specifically, in the case of a 100 nm pitch measurement, specimens—which has a nominal 1 μm pitch—is shown in
we have previously attained a relative standard uncertainty uc figure 8. The measurement uncertainties achieved for these
of 3.0 × 10−4 . Most of our major step height measurements comparisons are included in figures 6 and 7. In the case of the
have had a relative standard uncertainty uc of about 2.0 × 10−3 . two-dimensional pitch measurements for NANO5, the relative
However, we believe that 1.0 × 10−3 is attainable and have expanded uncertainty of the C-AFM results was typically
performed NIST internal step height measurements having a 1.5 × 10−3 but approached 1.0 × 10−3 for one specimen.
relative standard uncertainty uc only slightly above this level. This performance level is of importance in semiconductor
In the future, we expect further improvements in manufacturing metrology and is used in figures 6 and 7 as
performance as the uncertainty budgets are currently being a benchmark for evaluating our performance.

6
Meas. Sci. Technol. 22 (2011) 024001 J A Kramar et al

Examples of C-AFM Step Height Measurement Uncertainties

0.045
69 nm height
0.04
calibration for Au
Relative Expanded Uncertainty nano particle AFM
0.035 calibration
NIST Internal
(17 nm, 90 nm, 187 nm, 20 nm)
0.03

0.025 International
Comparison
0.02 Nano2 Goal for
(21 nm, 67 nm, 290 nm) 2nd round
21 nm and 98 nm
(for Company A) Au NP
0.015

0.01
1 x 10-3
0.005

0
1 2 3 4 5 6 7 8 Measurement
9 10 11 12 13 14 15 16 17 18 19 20

Figure 7. History of C-AFM step height measurement expanded uncertainties, expressed as a fraction of the measured value.

4. Critical dimension atomic force microscope

Both M3 and C-AFM are primarily optimized for measuring

feature placement or pitch in both one dimension and two
dimensions, and in the case of the C-AFM, also step height. In
principle, either can also measure feature sizes (i.e. linewidths
or critical dimensions); however, neither is optimized to take
advantage of the intrinsic nanometer level sensitivity possible
with the SPM techniques, especially for features with even
modest aspect ratios. This is because the approximately
conical tips employed are not able to access the nearly vertical
sidewalls that are typically of interest.
In the previous section, we discussed a C-AFM linewidth
measurement of a 500 nm feature with an 11 nm expanded
uncertainty. At the time that work was done, the
Figure 8. C-AFM image of the two-dimensional pitch grating from semiconductor industry was at the 180 nm technology node,
NANO5. as defined by the ITRS roadmap [26]. Even at that time,
however, 11 nm was larger than the uncertainties needed for
More recently, we have used the C-AFM to support a NIST semiconductor manufacturing reference metrology.
project to develop Au nanoparticle reference materials (RM) Critical dimension atomic force microscopy (CD-AFM),
[24]. A 100 nm grating pitch was measured with a relative Uc the topic of this section, is a technology developed within the
of 6 × 10−4 , and used to calibrate the SEM size measurements. semiconductor manufacturing industry to work around the
A 70 nm height standard was also measured as a transfer limitations of conventional AFM, and meet the need for
standard and used to calibrate the AFM measurements of linewidth reference measurements. During the last 10 years,
NIST has worked closely with industry in an effort to bring
particle height. At present, we are involved in an inter-
traceability to this application space as well.
laboratory comparison of 70 nm grating pitch measurements.
Another type of measurement we have performed with
the C-AFM is linewidth, discussed further in the next section. 4.1. Overview and design
Linewidth metrology is more challenging for the C-AFM In contrast to conventional AFM, CD-AFM uses flared
because it uses conical tips and conventional one-dimensional tips and more sophisticated feedback and scan control to
feedback. Despite such challenges, however, we have permit the imaging of structures with near-vertical sidewalls
successfully performed top width measurements using the [27]. From its inception to the present, the measurement
C-AFM as part of an inter-method comparison of linewidth of etched silicon and photoresist features in semiconductor
metrology techniques [25]. An expanded uncertainty of 11 nm manufacturing has been the primary application arena for CD-
was ultimately attained for the C-AFM linewidth measurement AFM [28]. More recently, the technology is also gaining use
of a 500 nm structure—limited by the characterization of the within the data storage industry [29]. The technology has
conical tip width. been available commercially since the mid-1990s. The Veeco

7
Meas. Sci. Technol. 22 (2011) 024001 J A Kramar et al

Figure 9. Functional schematic of first generation CD-AFM operation.

SXM3201 is a first generation CD-AFM. We have experience height measurements. Other measurements performed using
in implementing this type of instrument as a reference a newer generation instrument show that it is possible to
measurement system (RMS) and have had an instrument of achieve step height uncertainties about a factor of 2 lower
this type at NIST since 2003 [30, 31]. [32].
A simplified functional diagram of our CD-AFM is shown Our CD-AFM uncertainty budgets were developed in
in figure 9. Although the sample stage (which is large the manner described for the C-AFM. The current SXM320
enough to accommodate 300 mm wafers) does move for coarse uncertainty budget for pitch and linewidth is shown in
positioning purposes, it is not suitable for imaging due to table 2, and the uncertainty budget for height is shown in
both speed and motion error considerations. The instrument table 3. We have previously described the origin and estimate
employs a scanning head design for AFM imaging. of most of the uncertainty contributions [30, 31], so this
The scanner on the SXM320 is modular and consists of discussion will be limited to the most significant terms. For
three commercially available piezo actuators with integrated pitch measurements, the scale calibration and nonlinearity are
capacitance displacement sensors. The Z-axis assembly currently the most important contributions. The SXM lateral
consists of a coarse approach inchworm, which moves the scale calibration is checked using pitch artifacts on single
piezo stage up and down to engage the surface. This assembly crystal critical dimension reference material (SCCDRM)
sits in the carriage of an X–Y flexure stage which is driven samples [33]. Initially, the traceability for these pitch artifacts
by the X–Y piezo actuators. Newer generations of CD- was derived from earlier comparison measurements with the
AFM instruments use an integrated head assembly instead new NIST photomask standard reference material (SRM)
of a modular scanner, but the basic principles of operation SRM 2059 [34]. Subsequently, we have now measured
remain the same, and our calibration methods are applicable the SCCDRM pitch features using the C-AFM so that the
to instruments of any generation. traceability of the SXM scale calibration is directly anchored
Although our CD-AFM does not have built-in traceability, by the C-AFM. For height measurements, the scale calibration
its displacement sensors can be calibrated using standards is the most important contribution. The SXM vertical scale
measured on other instruments—such as the C-AFM—and calibration is routinely monitored using an SCCDRM. Initially,
full uncertainty budgets for pitch, height and linewidth this height traceability was derived from measurements on a
measurements have been developed. We use this CD- height specimen that was previously calibrated at NIST using
AFM primarily for linewidth measurements of near-vertical stylus profiler measurements. Subsequently, we have now
structures and to provide reference metrology for both internal also measured the height of features on an SCCDRM using
NIST projects and external dissemination. the C-AFM.
For linewidth measurements, the tip-related terms are the
most important sources of uncertainty. These are due to the
4.2. Measurements and uncertainty
uncertainty in the zeroth order tip width correction and to
Uncertainty budgets were developed for pitch, height higher order tip effects. Since we have discussed the nature
and linewidth measurements using this CD-AFM of these contributions in detail elsewhere [35, 37], we only
[30, 31]. Currently, the relative standard uncertainties give an overview here. Although the interaction of an AFM
are approximately 2 × 10−3 for pitch measurements up to tip with the imaged surface is complex, for many purposes a
several micrometers and 4 × 10−3 for sub-micrometer step highly simplified and two-dimensional model is useful. In this
1 Certain commercial equipment, instruments or materials are identified
basic model, the effect of the tip is represented as a simple
in this paper. Such identification does not imply recommendation or
additive offset which must be subtracted from the apparent
endorsement by NIST, nor does it imply that the products identified are width to obtain an accurate measurement. This offset is
necessarily the best available for the purpose. referred to as the zeroth order tip correction. Hence, the

8
Meas. Sci. Technol. 22 (2011) 024001 J A Kramar et al

Table 2. Uncertainty budget for SXM320 pitch and width measurements (k = 1).
Component Standard uncertainty

Type A (repeatability, reproducibility, sample non-uniformity, etc.) Observed SD or SDOM

Type B
a
Algorithm (edge/peak detection)/measurand definition
Tip contributions applying to width only:
Tip width correction (zeroth order) using SCCDRM 0.8 nm
b
Tip-related (higher order, e.g., offset height, overhang, corner rounding, tip wear)
Scale calibration (linear term) 1.0 × 10−3 W/P c
Scale nonlinearity 2.0 × 10−3 W/P c
b
Differential Abbe errors (offset between calibration plane and wafer or mask plane)
b
Non-position-dependent motion errors (e.g., mechanical, tip holder piezo hysteresis)
Cosine errors (in-sample-plane) 0.15 × 10−5 W/P c
Cosine errors (out-of-sample-plane) 0.15 × 10−3 W/P c
Combined standard uncertainty
uc (width) = [(SD)2 + (0.8 nm)2 + (2.2 × 10−3 W )2 ]1/2
uc (pitch) = [(SD)2 + (2.2 × 10−3 P)2 ]1/2
SD = standard deviation; SDOM = standard deviation of the mean.
a
The major contribution is often included in the observed type A variations, and that the residual type B component
may be negligible. But the contribution should be evaluated for each case. Note also that algorithm, measurand
definition and tip-related terms are inter-related and not always easily partitioned.
b
Contribution is negligible in many circumstances, but has not been fully evaluated for all cases, or is sample
dependent and should be evaluated for every measurement.
c
The uncertainty is proportional to the measured value of linewidth W or pitch P.

Table 3. Uncertainty budget for SXM320 step height measurements (k = 1).

Component Standard uncertainty

Type A (repeatability, reproducibility, sample non-uniformity, etc) Observed SD or SDOM

Type B
a
Algorithm/measurand definition (indirect sample effects)
Scale calibration (linear term) 3.0 × 10−3 H b
Scale nonlinearity 2.0 × 10−3 H b
c
Differential Abbe errors (change in tip offset from cap gauge centroid)
c
Non-position-dependent motion errors (e.g., mechanical, tip holder piezo hysteresis)
c
Tip-related (edge-broadening effects, roughness bias)
Cosine errors (Z-axis and X–Y scan plan misalignment) 0.15 × 10−3 H b
c
Out-of-plane motion (X–Y scanner)
Combined standard uncertainty, uc (height) = [(SD)2 + (3.6 × 10−3 H)2 ]1/2
SD = standard deviation; SDOM = standard deviation of the mean.
a
The major contribution is often included in the observed type A variations, and that the residual type B
component may be negligible, but the contribution should be evaluated for each case. Note also that algorithm,
measurand definition and tip-related terms are inter-related and not always easily partitioned.
b
The uncertainty is proportional to the measured value of the step height H.
c
Contribution is negligible in many circumstances, but has not been fully evaluated for all cases or is sample
dependent and should be evaluated for every measurement.

zeroth order uncertainty component represents the uncertainty from the SCCDRM project, characterization and correction
in the value of this correction. The finer details of the tip– for these effects is now more important in CD-AFM width
sample interaction, pertaining to things like flare radius, offset measurements.
height, feature sidewall angle, feature corner radius and the An example of a CD-AFM image on a SCCDRM feature
three-dimensional nature of both the tip and sample (i.e. shape is shown in figure 10. The sidewall angle of these structures
in the axis perpendicular to the scan direction) are thought of is very close to 90◦ , as can be seen in the image. This feature
as being higher order tip effects [37]. Because these effects clearly illustrates the fundamental strong point of the CD-
have a strong dependence on the specific geometry of each tip AFM technique—the ability to image near-vertical sidewalls.
and feature, it is difficult to make general statements about A feature of this type is also the best way to monitor the CD-
the resulting uncertainties, and it is necessary to make a AFM tip width. If such a structure can be calibrated—as was
specific assessment for every measurement. Until recently, done in the SCCDRM project—then it becomes possible to
this was of somewhat lesser concern in many measurements traceably calibrate the CD-AFM tip width. As a result of the
because these components were typically smaller than the NIST SCCDRM project, it is possible to calibrate the CD-
uncertainty in the zeroth order correction. However, with AFM tip width with a standard uncertainty slightly better than
reduction in uncertainty of the zeroth order term that resulted 1 nm [33, 35, 36].

9
Meas. Sci. Technol. 22 (2011) 024001 J A Kramar et al

Figure 10. CD-AFM image of a NIST SCCDRM showing the near-vertical sidewalls of this preferentially etched structure.

More recently, we have been using the CD-AFM to [3] Griffith J E and Grigg D A 1993 Dimensional metrology with
provide reference metrology on photomask features—both for scanning probe microscopes J. Appl. Phys. 74 R83–109
the SRM 2059 project itself and for international comparisons [4] Teague E C 1989 The National Institute of Standards and
Technology molecular measuring machine project:
we are planning to undertake with other NMIs [38]. In the case metrology and precision engineering design J. Vac. Sci.
of SRM 2059, the CD-AFM measurements are used to anchor Technol. B 7 1898–902
a series of optical measurements that serve as a comparator [5] Schneir J, McWaid T H, Alexander J and Wilfley B P 1994
between the control sample and samples being calibrated for Design of an atomic force microscope with interferometric
sale to customers. When the new AFM measurements are position control J. Vac. Sci. Technol. B 12 3561–6
[6] Meli F and Thalmann R 1998 Long range AFM profiler used
incorporated into the uncertainty analysis, it should be possible for accurate pitch measurements Meas. Sci. Technol.
to reduce the reported uncertainties on the smallest feature 9 1087–92
widths by about 40% [39]. [7] Gonda S, Doi T, Kurosawa T, Tanimura Y, Hisata N,
Yamagishi T, Fujimoto H and Yukawa H 1999 Real-time,
interferometrically measuring atomic force microscope for
5. Conclusions direct calibration of standards Rev. Sci. Instrum.
70 3362–8
For more than two decades, NIST has been at the forefront of [8] Picotto G B and Pisani M 2001 A sample scanning system
SPM-based dimensional metrology, adapting and expanding with nanometric accuracy for quantitative SPM
our efforts as needed to respond to growth and migration of the measurements Ultramicroscopy 86 247–54
industrial application space for SPM dimensional metrology. [9] Misumi I et al 2005 Sub-hundred nanometer pitch
measurements using an AFM with differential laser
An overview of these efforts has been presented. The bulk interferometers for designing usable lateral scales
of our activities include special measurements and calibrations Meas. Sci. Technol. 16 2080–90
for internal and external users, and research aimed at advancing [10] Dai G, Pohlenz F, Dziomba T, Xu M, Diener A, Koenders L
the state of the art. We also work closely with industry to and Danzebrink H U 2007 Accurate and traceable
identify new applications of SPM dimensional metrology, and calibration of two-dimensional gratings Meas. Sci. Technol.
18 415–21
to develop rigorously traceable methods for existing uses. [11] Kramar J A 2005 Nanometre resolution metrology with the
SPM-based dimensional metrology remains a critically Molecular Measuring Machine Meas. Sci. Technol.
important area for continued NMI nanometrology research, 16 2121–8
especially in light of the rapid growth of nanotechnology and [12] Li J, Shen Y-L, Jeong J, Scire F E and Kramar J A 2008
nanomanufacturing in industry. A compact, compound actuator for the Molecular
Measuring Machine Proc. Am. Soc. Precis. Eng.
44 312–5
References [13] Lan K-J, Yen J-Y and Kramar J A 2004 Active vibration
isolation for a long range scanning tunneling microscope
[1] Binnig G and Rohrer H 1982 Scanning tunneling microscopy Asian J. Control 6 179–86
Helv. Phys. Acta 55 726–35 [14] Lan K-J, Yen J-Y and Kramar J A 2004 Sliding mode control
[2] Binnig G, Quate C F and Gerber Ch 1986 Atomic force for active vibration isolation of a long range scanning
microscope Phys. Rev. Lett. 56 930–3 tunneling microscope Rev. Sci. Instrum. 75 4367–73

10
Meas. Sci. Technol. 22 (2011) 024001 J A Kramar et al

[15] ISO 1995 Guide to the Expression of Uncertainty in [28] Liu H-C, Osborne J R, Osborn M and Dahlen G 2007
Measurement (Geneva, Switzerland: ISO) Advanced CD-AFM probe tip shape characterization for
[16] Taylor B N and Kuyatt C E 1994 Guidelines for evaluating and metrology accuracy and throughput Proc. SPIE
expressing the uncertainty of NIST measurement results 6518 65183K
NIST Technical Note 1297 [29] Bao T, Fong D and Hand S 2007 Automated AFM as an
[17] McClelland J J, Scholten R E, Palm E C and Celotta R J 1993 industrial process metrology tool for nanoelectronic
Laser-focused atomic deposition Science 262 877–80 manufacturing Applied Scanning Probe Methods X (Berlin:
[18] Dixson R, Köning R, Fu J, Vorburger T and Renegar B 2000 Springer) pp 359–412
Accurate dimensional metrology with atomic force [30] Dixson R, Guerry A, Bennett M, Vorburger T and Bunday B
microscopy Proc. SPIE 3998 362–8 2003 Implementation of a reference measurement system
[19] Dixson R, Köning R, Tsai V W, Fu J and Vorburger T V 1999 using CD-AFM Proc. SPIE 5038 150–65
Dimensional Metrology with the NIST calibrated atomic [31] Dixson R, Guerry A, Bennett M, Vorburger T and Postek M
force microscope Proc. SPIE 3677 20–34 2002 Toward traceability for at line AFM dimensional
[20] Dixson R, Köning R, Vorburger T V, Fu J and Tsai V W 1998 metrology Proc. SPIE 4689 313–35
Measurement of pitch and width samples with the NIST [32] Orji N G, Dixson R G, Martinez A, Bunday B D, Allgair J A
calibrated atomic force microscope Proc. SPIE and Vorburger T V 2007 Progress on implementation of a
3332 420–32 reference measurement system based on a
[21] Meli F 2001 International comparison in the field of critical-dimension atomic force microscope
nanometrology: pitch of 1D gratings (Nano4) J. Micro/Nanolith. MEMS MOEMS 6 023002
Proc. 2nd EUSPEN (European Society for Precision [33] Cresswell M, Dixson R, Guthrie W, Allen R, Murabito C,
Engineering) Int. Conf. (Turin, Italy, 27–31 May 2001) Park B, Martinez de Pinillos J and Hunt A 2005 Critical
pp 358–61 dimension reference features with sub-five nanometer
[22] Koenders L et al 2003 Comparison on nanometrology: nano uncertainty Proc. SPIE 5752 288–303
2—step height Metrologia 40 04001 (Technical [34] Potzick J, Pedulla J M and Stocker M 2003 Updated NIST
Supplement) photomask linewidth standard Proc. SPIE 5038 338–49
[23] Garnaes J and Dirscherl K 2008 Nano 5—2D gratings—final [35] Dixson R, Fu J, Orji N, Guthrie W, Allen R and Cresswell M
report Metrologia 45 04003 (Technical Supplement) 2005 CD-AFM reference metrology at NIST and
[24] Hackley V 2008 Nanoparticle standards at NIST: gold SEMATECH Proc. SPIE 5752 324–36
nanoparticle reference materials and their characterization [36] Dixson R G, Allen R A, Guthrie W F and Cresswell M W
2nd Tri-National Workshop on Standards for 2005 Traceable calibration of critical-dimension atomic
Nanotechnology (Gaithersburg, MD, 6 February 2008) force microscope linewidth measurements with nanometer
[25] Villarrubia J S, Dixson R G, Jones S N, Lowney J R, uncertainty J. Vac. Sci. Technol. B 23 3028–32
Postek M T, Allen R A and Cresswell M W 1999 [37] Orji N G and Dixson R G 2007 Higher order tip effects in
Intercomparison of SEM, AFM, and electrical linewidths traceable CD-AFM-based linewidth measurements Meas.
Proc. SPIE 3677 587–98 Sci. Technol. 18 448–55
[26] SIA 2000 International Technology Roadmap for [38] Bodermann B et al 2009 Results of an international photomask
Semiconductors (ITRS) (San Jose, CA: SIA) linewidth comparison of NIST and PTB Proc. SPIE
Metrology Chapter 7488 74881H
[27] Martin Y and Wickramasinghe H K 1994 Method for imaging [39] Dixson R, Potzick J and Orji N 2008 Recalibration of the SRM
sidewalls by atomic force microscopy Appl. Phys. Lett. 2059 master standard using traceable atomic force
64 2498–500 microscope metrology Proc. SPIE 7122 71222Q

11