Professional Documents
Culture Documents
Anders Kühle
Danish Fundamental Metrology
Matematiktorvet 307 , DK-2800 Lyngby
Abstract:
This report documents the results of Nordtest project 1596-02 carried out by the Swedish
national metrology institute (SP), Toponova AB and the Danish Institute of Fundamental
Metrology (DFM) in year 2002. More than 8 years of research in calibration of atomic force
microscopes (AFM) has been condensed in to two procedures for calibration of metrology grade
AFM instruments. The greatest success of the project is that DFM in September 2002 has - as
one of the first laboratories in the world - successfully passed a technical assessment for using
one of these procedures for accredited (i.e. traceable) AFM measurements. It was found that the
calibration procedures in particular for the lateral plane with some limitations could also be used
for non-metrology grade AFMs if these are properly designed and linearised. In the vertical
direction this is more dubious since measurements of roughness in the 1-10 nm range gave
unacceptable differences in the results for two non-metrology grade AFMs and a metrology
AFM. The calibration procedures were successfully tested on an interference microscope and it
was shown that equivalence between AFM and interference microscope measurements can be
established through appropriate filtering, at least in the 1-10 nm range (vertical). In conclusion
there is a need to assess roughness measurement in the nanometer range with atomic force
microscopes more carefully. Also more research is required in order to establish standardised
procedures for making comparisons between AFM and interference microscope data.
Resumé:
Denne rapport gengiver resultat af Nordtest projekt nr. 1596-02 udført af Sveriges Provnings-
och Forskningsinstitut (SP), Toponova AB og Dansk Institut for Fundamental Metrologi (DFM)
i år 2002. Over 8 års forskning i kalibrering af atomic force mikroskoper (AFM) har i dette
projekt resulteret i to detaljerede procedurer for kalibrering af AFM instrumenter af metrologisk
kvalitet. Projektets største succes er, at DFM i september 2002 efter en teknisk evaluering har
opnået akkreditering til at benytte den ene af disse procedurer til sporbare AFM målinger.
Procedurerne blev også afprøvet på AFM’er af ikke metrologisk kvalitet. Resultatet er, at især
proceduren for lateral kalibrering med visse forbehold vil kunne kan anvendes på sådanne
AFM’er, såfremt disse er hensigtsmæssigt konstruerede og lineariserede. I den vertikale retning
er dette mere tvivlsomt, idet måling af ruhed i 1-10 nm intervallet gav resultater, der var
uacceptable forskelle for to ikke-metrologi AFM’er og et AFM af metrologisk kvalitet.
Procedurerne blev også afprøvet på et interferens mikroskop, og det blev bekræftet at der ved
brug af passende filtrering kan etableres korrespondance mellem ruhedsmålinger foretaget med
interferens mikroskoper og AFM instrumenter, i hvert fald i 1-10 nm området (vertikalt). Det
konkluderes, at der er behov for mere grundige undersøgelser af ruhedsmåling i nanometer
området med AFM for at kunne etablere troværdig ækvivalens mellem målinger. Der er også
behov mere forskning, der kan føre til standarder for hvordan interferens mikroskop målinger og
AFM målinger kan sammenlignes.
Danish Fundamental Metrology DFM- 02-R29 Phone 4593 1144
Matematiktorvet 307 3308 AK Fax 4593 1137
DK-2800 Kgs. Lyngby 2002 12 18 Web: www.dfm.dtu.dk
2
Contents
1 Project overview ..............................................................................................................................................3
1.1 Introduction..............................................................................................................................................3
1.2 Project background, objectives and main results .....................................................................................3
1.3 Partners ....................................................................................................................................................4
1.4 Course of the project ................................................................................................................................4
2 Lateral calibration ............................................................................................................................................5
2.1 Calibration of the metrology AFM ..........................................................................................................6
2.2 Calibration of Toponova’s standard AFM ...............................................................................................7
2.3 Calibration of SP’s standard AFM...........................................................................................................8
2.4 Calibration of Toponova’s interference microscope................................................................................9
3 Vertical calibration.........................................................................................................................................10
3.1 Calibration of the metrology AFM and of the TGZ02 grating...............................................................11
3.2 Calibration of Toponova’s and SP’s standard AFMs ............................................................................12
3.3 Measurement of the TGZ02 grating with an interference microscope...................................................14
4 Comparative roughness measurements on ball-joints for hip implants..........................................................15
4.1 AFM roughness measurements ..............................................................................................................15
4.2 Comparison between AFM and interference microscope roughness data .............................................17
5 Conclusion .....................................................................................................................................................18
6 List of equipment used...................................................................................................................................19
7 References......................................................................................................................................................19
Appendix Procedures for calibration of the three axes of an atomic force microscope using transfer standards
3308 AK 2002 10 29
3
1 Project overview
1.1 Introduction
Scanning Probe Microscopes (SPM) and in particular the family member called the Atomic Force Microscope
(AFM) - invented in 1986 [1] - are at present the most powerful instruments available for accurate nano- and
micrometer surface measurements including the characterisation of soft materials like polymers and biological
materials. In the AFM a sharp probe senses and follows the surface landscape while being scanned in an xy
pattern. It thereby gives a direct image of the surface area from the atomic level to about 0.1×0.1 mm2 in the
plane and up to about 6 µm for the height. The tip is an integrated part of a cantilever beam, which acts as the
force sensor in the AFM. The deflection of the cantilever, and hence the force, is most often measured by
sensing the displacement of a laser beam reflected on the back side of the cantilever on a photo detector.
Standard AFM instruments suffer from time dependent non-linearity and hysteresis in their actuating parts
which makes them difficult to use for metrology purposes. On- and off-line software linearisation schemes and
standardised instrument settings can improve this situation. The most reliable and reproducible results are,
however, achieved with metrology grade instruments, which have distance sensors on all axes. A couple of SPM
manufacturers offer a “Metrology” version of their instruments which allow for unmatched reproducibility and
linearity. Both standard AFMs and metrology AFMs require calibration in order to provide traceability to SI
units in measurements, which is the main purpose of this project.
3308 AK 2002 10 29
4
guidelines for internal quality documents at DFM and therefore have an appearance different to this report. As a
result DFM has in September 2002 successfully passed a technical assessment by the Danish accreditation body,
DANAK, which makes use of internationally acknowledged experts for such evaluations. Thus on the
background of the calibration procedures written in this Nordtest project DFM is now accredited - to our
knowledge as the first laboratory in the world - to calibrate the pitch of two dimensional gratings using AFM.
The second goal in this project was to assess how the procedures can be used for standard grade AFM
instruments and interference microscopes: what are the limitations and/or what are the requirements? The results
show that the non-linearity in the scanning devices of two standard grade AFMs used is two high for using the
procedure. However, it should be possible to linearise these instruments sufficiently well by adjusting their
internal parameters, so that the procedure is also applicable for these types of instruments – at least for lateral
calibration. The experiments showed that the procedures are highly applicable to interference microscopes.
The third goal was to compare measurements on industrial samples carried out with 1) a calibrated metrological
AFM (at DFM) and 2) two standard AFMs and 3) an interference microscope. The objective is to illustrate the
impact of the difference in the nature and quality of the instruments. Before comparing the measurements the
images were calibrated. Measurements were performed on two artificial hip joint balls from different
manufacturers. In order to facilitate direct comparison, Toponova using laser writing marked the measurement
areas on the two artefacts. It proved to be difficult to compare roughness in the 1-10 nm range for the used
AFMs, which are of different types. The main cause to discrepancies should probably be found in the non-linear
properties of the scanning devices in the instruments. However, it cannot be excluded that there are other factors
at play. It was also concluded that interference microscope images and AFM images can be compared given that
the surfaces do not have topographic variations in excess of the range covered by the phase shifting mode of the
interferometer (about 200-300 nm) and when applying appropriate filters to the AFM images.
1.3 Partners
Danish Institute of Fundamental Metrology (DFM) was the co-ordinator of the project. DFM has had as the
main task of writing and evaluating the calibration procedures (Work Package 2 and 3) and has also taken part
in the comparative measurements (WP 5). Moreover DFM has analysed all the results and compiled the final
report (WP 6). The staffs involved at DFM have been Dr. Jørgen Garnæs and Dr. Anders Kühle.
Swedish National Testing and Research Institute (SP) is the Swedish national metrology institute. SP’s role
was to evaluate the calibration procedures made by DFM (WP 3) and to perform calibration measurements with
a standard AFM instrument (WP 4). SP also measured the industrial samples provided by Toponova for the
common roughness measurement comparison (WP 5). SP’s staff has been Dr. Mikael Frennberg.
Toponova AB is a new small company which offers services and products related to surface roughness
measurement. Toponova has supplied the specimens for the comparative roughness measurements – two
artificial hip ball-joints from different manufacturers. With both a standard AFM and an interference
microscope, Toponova has measured the set of calibration standards and the hip joints. The staff involved at
Toponova has been Stefan Rosén and Dr. Bengt-Göran Rosén who is also lecturing at Halmstad Högskola.
3308 AK 2002 10 29
5
2 Lateral calibration
For lateral calibration a calibration grid produced within the EU project referred to in the introduction were
circulated between the partners. It was calibrated by DFM using the lateral calibration procedure described in
the appendix using a reference standard (of the same type), which had previously been calibrated and certified
by the National Physical Laboratory, NPL, in the United Kingdom. The general design of the standards is
described in the procedure (appendix), but in short they consist of a two dimensional grid etched in a silicon
substrate. The grid is described by its unit cell in terms of the side lengths, La and Lb of the unit cell and the
angle, γ, between the two corresponding component vectors. The parameters of the calibration grid are given in
Figure 3 together with a photograph of the artefact.
The calibration procedure yields three correction parameters, describing the microscope: Cx, Cy, and Cxy. The
metric coordinate system’s x-axis can be chosen so that it is parallel to the microscope coordinate system’s x*-
axis and the linear transformation can then be written as
x C x C xy x*
= *
y 0 C y y
The matrix elements can be interpreted as a correction parameter Cx for the x-direction, a correction parameter
Cy for the y-direction and a coupling term Cxy between the scanned x and y axes. If the x*- and y*-axes in the
microscope coordinate system are perpendicular, then Cxy = 0, Cx will be the scaling in the x-direction and Cy
will be the scaling factor in the y-direction. The angle between the x*- and the y*-direction in the microscope
coordinate system is equal to cot-1(Cxy/ Cy) for Cxy ≠ 0.
The procedure for lateral calibration has been developed for a metrology AFM, for which the non-linearity
(which is small) can be neglected in the uncertainty analysis. The procedure is designed for allowing subsequent
pitch measurements on grids with unknown pitch. For pitch measurements a small non-linearity can be
neglected when the same image size is always used both for the calibration of the AFM and for the measurement
of the grid with the unknown pitch. This is due to the fact that the calibration gives the correction factors for the
entire image – side to side, top to bottom. So as long one measures an average quantity over the entire image the
result is only affected by non-linearity if (1) the non-linearity prohibits detection of unit cells correctly (2) if the
non-linearity is so high that the inclusion/exclusions of unit cells at the boundary of the images will significantly
influence the average value.
Figure 3. Right: image of the calibration grid used. The chip is made from Silicon and is about 6 mm wide. The
grid itself is in the square shaped area in the middle. Left: The calibrated lengths of the unit cell vectors of the
grid, and the angle between the unit cell vectors.
3308 AK 2002 10 29
6
Figure 4. 60 × 60 µm2 (512×512 pixels) AFM image of the calibration grid using the metrology AFM. The two
graphs show the deviation (in x and y respectively) from the predicted positions of the unit cells in the image.
The unit cell vector components found by the procedure are: a = (999.86 nm, -28.019 nm), b = (26.027 nm,
999.17 nm).
3308 AK 2002 10 29
7
Figure 5. 30 × 30 µm2 (512×512 pixels) AFM image of the calibration grid using Toponova’s standard AFM (a
DME DualScope). The two graphs show the deviation (in x and y respectively) from the predicted positions of
the unit cells in the image. It is seen that the AFM is non-linear – in particular in the y direction. The unit cell
vector components found by the procedure are: a = (954.97 nm, -10.967 nm), b = (-13.825 nm, 905.63 nm).
3308 AK 2002 10 29
8
Figure 6. 40 × 40 µm2 (512×512 pixels) AFM image of the calibration grid using SP’s standard AFM (a DME
DualScope). The two graphs show the deviation (in x and y respectively) from the predicted positions of the unit
cells in the image. It is seen that the AFM is non-linear – in particular in the y direction. The unit cell vector
components found by the procedure are: a = (978.99 nm, 6.2481 nm), b = (-1.6432 nm, 906.18 nm).
The strong non-linearity in y for the two standard AFMs is caused by a combination of hysteresis and creep in
the piezo electric elements, which perform the tip motion in the microscopes. It is a commonly known
phenomenon, which can be reduced dramatically and accounted for if instead of returning to the upper left
corner after every scan down, the tip would instead reverse and scan from the bottom of the image and up.
3308 AK 2002 10 29
9
Figure 7. 39 × 39 µm2 (180×152 pixels) image of the calibration grid using Toponovas’s interference
microscope (MicroXAM). The two graphs show the deviation (in x and y respectively) from the predicted
positions of the unit cells in the image. It is seen that the interferometer is highly linear. The unit cell vector
components found by the procedure are: a = (1008.1 nm, 7.1933 nm), b = (6.8768 nm, 1004.0 nm).
3308 AK 2002 10 29
10
3 Vertical calibration
The calibration procedure in the appendix yields three correction parameters: Czx*2, Czz*2, and Czz*. The latter,
Czz*, is the scaling (or calibration) factor of the z axis scale of the microscope at x, y, z ≈ 0 nm, Czz*2 quantifies
the weak dependence of Czz* on z to order one, and Czx*2 represents the coupling between the z-axis and the x-
axis, i.e. the image bow. The coupling between the z- and y-axes is not taken in to account since most often
images have to be levelled line-wise and thus any image bow along the y-axis will be flattened out. This implies
that step heights should always be measured with the x-axis perpendicular to the step (see Figure 8 - Figure 11).
Czx*2 is used to correct the images for the bow caused by the coupling between the x- and z-axes. Besides this,
neither Czx*2 or Czz*2 are used directly for the calculation of Czz*, but only play a role in the uncertainty budget.
The procedure for vertical calibration has been developed for a metrology AFM, for which the sensed tip
position and the physical tip position in the vertical direction is a single valued function, i.e. there is no
hysteresis. For the calibration of the metrology AFM a certified reference standard with a nominal step height
(groove) of 800 nm is used. Using the traceability provided by this standard the MDT-TGZ02 grating from
µMasch used for this project was calibrated.
Figure 8. The calibrated depth of the grooves of the standard. The groove is approximately 28 µm wide. Right:
AFM image (3D projection) of the groove.
3308 AK 2002 10 29
11
Czx*2 = (3.9 ± 1.1) m-1 Czz*2 = (0.0035 ± 0.001) µm-1 Czz* = 1.058 ± 0.011
The uncertainties are expanded standard uncertainties (k=2). Note that the parameter Czz*2 is not to be used
directly, but is used to estimate the uncertainty when calibrating an object with unknown height: Assuming that
calibrations are always carried out such that the mean z-value is within the ±1µm range, the maximum error on
Czz* due to a height dependent sensitivity will be 1 µm × 0.0035 µm-1 = 0.0035 (brick-wall) corresponding to a
standard uncertainty of 0.0020.
An AFM image acquired near the centre of the TGZ02 grating used for circulating between partners is shown in
Figure 9. The average step height within the imaged area was derived by first calculating the average line profile
in the x-direction. Then the procedure (appendix), which is in agreement with the ISO5436 standard, was used to
measure the depth of the grooves, as indicated by the profile shown in the figure. The observed step-height is
then corrected with Czz* (above), and the uncertainty evaluated using the principles above. The result is hTGZ02 =
(104.1 ± 1.2) nm.
Figure 9. 20× 20 µm2 (512×512 pixels). Corrected image and average profile of the TGZ02 grating acquired
with the metrology AFM at DFM. The groove depth is an average over 6 parallel grooves each evaluated
according to the ISO 5436 guideline and corrected according to the procedure. The result is (104.1 ± 1.2) nm.
3308 AK 2002 10 29
12
Figure 10. 20× 20 µm2 (512×512 pixels). Levelled image and average profile of the TGZ02 grating acquired
with the DualScope at Toponova. The groove depth is an average over 6 parallel grooves each evaluated
according to the ISO 5436 guideline. The result is 102.4 nm. The hysterises is seen as overshoots on the left side
of the lines (scanning from left to right) and as dips in the left side of the grooves.
3308 AK 2002 10 29
13
Figure 11. 20× 20 µm2 (512×512 pixels). Levelled image and average profile of the TGZ02 grating acquired
with the DualScope (standard AFM) at SP. The groove depth is an average over 6 parallel grooves each
evaluated according to the ISO 5436 guideline. The result is 114.4 nm. The hysterises is seen as overshoots on
the left side of the lines (scanning from left to right) and as dips in the left side of the grooves. Due to the slight
angle of the pattern of the grating to the xy-axes of the AFM this is not as clear as in Figure 10.
3308 AK 2002 10 29
14
Figure 12. 20× 20 µm2 (92×78 pixels). Levelled image and average profile of the TGZ02 grating acquired with
the MicroXAM interferometer at Toponova. Due to the limited resolution of the interferometer the lines are not
imaged as rectangular features but appear rounded. In fact they appear so rounded that the peak to valley height
is only about 80 nm – which should be compared to the expected height of 104.1 nm.
3308 AK 2002 10 29
15
Table 6. Summary of the roughness data calculated for the images in Figure 13.
It is seen that though the roughness values are within the same range they do not coincide very will with up to
almost a factor of two between the smallest values and the highest values. It is difficult to point at one
straightforward explanation, especially because the measurements of the vertical calibration grating gave results,
which agreed within about 10%. We have identified three possible causes to the discrepancies:
1) Deviations within 1-2 nanometers can be explained by the fact that all AFMs used in this project are
operating in dynamic mode, i.e. using the cantilever as a resonator. It is well known that this mode of
operation is sensitive to adsorbed water layers and other contamination on the surfaces, which the tip may
track instead of the real surface. But as the images do all have a comparable resolution in xy (showing no
blur due to contamination) this is probably not the cause.
2) In the dynamic mode the cantilever sometimes is un-stable switching between attractive and repulsive
modes [6]. When this is happening fast it is difficult to see this as other than random noise in the images –
increasing the roughness with up to 1-5 nm. Another cause to increased roughness can simply be
mechanical or electrical noise in the microscopes.
3) When comparing profiles taken over specific holes on the sample “1DP” for the three microscopes we
observe that the measured depth of the holes actually do vary from microscope to microscope in the same
way as the roughness data in Table 6. This first of all excludes the possibility that it is the image processing
(filtering) which disturbs the data. It also shows that the calibration performed with the 104.1 nm grating is
not very useful. This is most probable due to the non-linearity and hysteresis of the piezoelectric scanners in
the two standard AFMs.
In conclusion we have found if difficult to compare roughness in the 1-10 nm range for AFMs of different types.
The main cause to discrepancies should probably be found in the non-linear properties of the scanning devices
in the instruments of the standard type. However, it cannot be excluded that there are other factors at play.
3308 AK 2002 10 29
16
DFM
Topo-
nova
SP
Figure 13. AFM images, of the two artificial hip joint balls, named “1DP” (left), and “2BM” (right), taken with
the AFMs at the three partners; top: DFM, middle: Toponova, bottom: SP.
3308 AK 2002 10 29
17
Figure 14. Filtered AFM images (left) and interference microscope images (right) of the two hip joint balls.
Top: 1DP, bottom: 2BM. The AFM images were low pass filtered with a two-dimensional Gaussian envelope
with a cut-off wavelength of 2600 nm.
3308 AK 2002 10 29
18
In conclusion (supported by [5]) we can say that interference microscope images and AFM images can be
compared given that the surfaces do not have topographic variations in excess of the range covered by the phase
shifting mode of the interferometer (about 200-300 nm). We also conclude that the method proposed in [4] and
used in section 3.3 for determining the effective cut-off wavelength of the interference microscope within the
framework of Gaussian weighting functions is useful.
Table 7. Summary of the roughness data calculated for the AFM and interference microscope
images in Figure 14.
5 Conclusion
In this project more than 8 years of research in calibration of atomic force microscopes has been summarised
and brought in to the form of two procedures for calibration of metrology grade AFM instruments: one for the
lateral plane and one for the vertical axis. The procedures include a full uncertainty analysis. As a result DFM
has in September 2002 successfully passed a technical assessment by the Danish accreditation body, DANAK,
for using the lateral calibration procedure to calibrating the pitch of two dimensional transfer standards for
clients. To our knowledge this is one of, if not the first accredited use of AFM in the world. The challenge for
the future is to disseminate traceable measurements in the nanometer scale to industry and to extend the
catalogue of traceable measurement types through research in targeted areas as f. ex. micro optics.
It was investigated how the calibration procedures work for standard AFMs and interference microscopes. The
conclusion is that the non-linearity in the scanning devices of the two standard grade AFMs used is two high for
using the procedure for lateral calibration. A better software linearisation and better control of the piezo electric
scanners in the instruments could change this. The vertical calibration procedure could be applied, but
unfortunately it was not found useful as explained below. The experiments showed that the procedures are
highly applicable to interference microscopes, only the grating used for vertical calibration should have a higher
pitch than the one used, preferably about 20 µm.
Roughness measurements (1-10 nm range) on two artificial hip joint balls with three AFMs did not give data
which were satisfactory consistent, even though the vertical axes of the instruments were calibrated in the
100 nm range. The cause should probably be found in the non-linear and hysteretic nature of the piezo electric
scanners in the two standard AFMs. But no definite cause was identified. In order to understand this better and
establish equivalence there is thus a need to assess roughness measurement in the nanometer range with atomic
force microscopes more carefully.
Last it was confirmed that metrology AFM and interference microscope roughness measurements can be
compared when applying Gaussian filters to the AFM images given that the surfaces have small topographic
variations, e.g. less than 50 nm. The results indicate that the properties of the filter can be determined from
images of rectangular structures. These conclusions are very promising, however, more work is required in this
field in order to establish standardised procedures for making comparisons.
3308 AK 2002 10 29
19
7 References
[1] Atomic Force Microscope, G. Binnig, C.F. Quate, CH. Gerber, Phys. Rev. Lett. 56, 930 (1986).
[2] SMT4/CT95/2018.
[3] “SURFSTAND”, SMT4/CT98/2209.
[4] W. Hillmann, U. Brand, M. Krystek, Measurement 19 (2), 95-102 (1996).
[5] J. Garnaes, N. Kofod, A. Kühle, C. Nielsen., K. Dirscherl, L. Blunt, Precision Engineering 27 (1), 91-98
(2003).
[6] A. Kühle, A. H. Sørensen, J. B. Zandbergen, and Jakob Bohr, Appl. Phys. A 66, 329-332 (1998).
3308 AK 2002 10 29
Page 1 of 31
2002-12-18
Appendix to
Contents
1. Objective ..................................................... 3
2. Description of equipment ...................................... 4
2.1 The microscope ................................................ 4
2.2 Clean room bench .............................................. 4
2.3 Software ...................................................... 5
2.4 The two-dimensional standard .................................. 5
2.5 The step height standard ...................................... 6
3. Lateral calibration using transfer standards .................. 7
3.1 Theory for lateral calibration ................................ 7
3.2 Measurement of reference standard ............................ 12
3.3 Image processing ............................................. 13
3.4 The certificate .............................................. 14
3.5 Uncertainty analysis ......................................... 14
3.6 Uncertainty budget ........................................... 18
4. Vertical calibration using transfer standards ................ 19
4.1 Theory for vertical calibration .............................. 19
4.2 Measurement of reference standard ............................ 24
4.3 Image processing and calculations ............................ 25
4.4 The certificate .............................................. 26
4.5 Uncertainty analysis ......................................... 26
5. Addendum - Estimation of a reference plane, levelling ........ 31
AFM calibration procedures Page 3 of 31
1. Objective
The objective of this calibration procedure is to calibrate a metrology
atomic force microscope (AFM), in units of metre,
1. in the lateral plane for a nominal scan area of 60 µm x 60 µm.
2. in the vertical direction for a nominal length up to 6 µm
at a reference temperature of Tref = (23º± 2)°C and a relative humidity of
Href = (50 ± 10)%,
The lateral calibration is based on a two dimensional grid with a cali-
brated pitch in the x and y direction. The vertical calibration is based on
a calibrated step height and a flat surface. The result of the vertical
calibration is a 2 x 2 correction matrix and for the vertical calibration
it is a set of three correction parameters. Using these corrections a re-
corded image can be described in a metric coordinate system off-line.
The equipment is described in section 2, the theory for the lateral cali-
bration is presented in section 3 and the theory for the vertical calibra-
tion in section 4.
2. Description of equipment
Vertical and
lateral capac-
Parallelogram ity position
sensors
Reference cu-
be
Cantilever
and tip Sample
Figure 1 Sketch of the metrology system. The metrology frame is indicated for one
axis only. The figure is modified from [1].
During calibration it is indirectly the sensitivity of the capacitive sen-
sors which is determined. The relationship between distance and measured
capacitance may deviate from theory due to a number of causes [2]. These
are bow of capacitor plates, plate tilt (that is the plates are not 100%
parallel), and stray capacitance. Stray capacitance can be reduced by the
use of guard rings; tilt errors can be reduced by using large gaps (plate
spacing). The draw-back is a reduction in the signal to noise ratio of the
sensor. Calibrating the sensors and applying corrections to the signals can
account for most of these errors causing deviations from linearity. How-
ever, for practical capacitive sensors a small amount of non linearity is
inevitably present. For the microscope used the integral non linearity is
specified by the manufacturer to less than 1 ‰ of the scan range in the
lateral plane and 1 % in the vertical direction. For the lateral plane we
observe a non linearity less than 0.4 ‰.
2.3 Software
The software used (Scanning Probe Image Processor, SPIP [10]) for the de-
scribed analysis is an image processing program with special tools for ac-
curate characterisation of image structures and calibration of in particu-
lar scanning probe microscopes. A number that is stored in each data file
generated identifies the current version of the software. Every new release
of the software taken in use is validated according to ISO 17025 require-
ments and [3].For precautions and details about access to the programme a
Software Validation Checklist should be maintained.
For calculating some of the correction parameters and all the uncertainty
budget the program DFM-GUM is used. Every release of DFM-GUM is validated
according to ISO 17025 and [3] by the manufacturer. For precautions and de-
tails about access to the programme a Software Validation Checklist is
maintained.
2.4.1 Traceability
The three quantities, La , Lb, and the angle γ , have been calibrated by a
diffraction technique using a helium-neon laser, calibrated against an io-
dine-stabilized reference laser at NPL [4]. The expanded uncertainties of
the three quantities are given in the calibration certificate. The artefact
has La = Lb = 895 nm and γ ≈ 91°.
Figure 3 To the left is a part of the step height standard showing the reference
field and the positions E1 to E4 chosen for calibration. To the right a photogra-
phy of the standard with the same orientation.
2.5.1 Traceability
The depth of the groove in the reference area was measured at four spots
named E1 to E4 (see Figure 4). The groove depth was determined by fringe
evaluation in a calibrated interference microscopy [6]. As this method al-
lows only fringe fractions to be determined, a calibrated contact stylus
instrument estimated the number of whole interference fringe.
AFM calibration procedures Page 7 of 31
where z*, x*i and y*j are the vertical and lateral positions in the z*, x*
and y* direction which would be displayed by the microscope at the refer-
ence conditions, with the dimension length, and where c*x and c*y are the
microscopes scaling factors, with the dimension length, used by the micro-
scope. This reference coordinate system will be referred to as the “micro-
scope coordinate system” and a “*” will indicate that the basis for a coor-
dinate set is the coordinate axes of the microscopes coordinate system. The
microscope often uses the same scaling factor, and the same number of pixel
and along the x*i and y*j directions corresponding to a square image. Cor-
rect microscope scaling factors, and a correct angle between the x* and y*
direction can be calculated from the estimated correction parameters cx cy,
and cxy , however, they are not necessary to estimated.
To simplify the notation the discrete values of the lateral position x*i
and y*j, are substituted with only x*, and y* and redefined as a continuous
variable when, for example, analysis is done at sub pixel level.
AFM calibration procedures Page 8 of 31
x cxx * cxy * x*
= * ( 2 )
y 0 cyy * y
where cxx*, cyy*, and cxy* are the correction parameters to be estimated by
the calibration procedure. By choosing the x and x* direction to be paral-
lel the matrix element cyx* become zero.
The matrix elements can be interpreted as a correction parameter cxx* for
the x-direction, a correction parameter cyy* for the y-direction and a cou-
pling term cxy* between the scanned x and y axes.
If the x*- and y*-axes in the microscope coordinate system are perpendicu-
lar, then cxy* = 0, cxx* will be a scaling in the x-direction and cyy* will be
a scaling factor in the y-direction. The angle between the x*- and the y*-
direction in the microscope coordinate system is equal to cot-1(cxy*/ cyy*)
for cxy* ≠ 0.
The correction parameters are functions of same parameters as the sampled
coordinates (see section 3.1.1) The measurand during calibration is defined
as the values at equilibrium at the reference temperature Tref = 23ºC and a
relative humidity of Href = 50%, and a reference height of the probe
∆hp,ref = 0, that is
cxx * ≡ cxx *(Tref, H ref, ∆hp,ref)
( 3 )
cyy * ≡ cyy *(Tref, H ref, ∆hp,ref)
cxy * ≡ cxy *(Tref, H ref, ∆hp,ref)
x* x* / cos β x
→
(4)
y* y* / cos β y
1
The basal plane is defined as the bottom part of the surface in case of a continuous surface
with bumps and as the top part of the surface in the case of a continuous surface with pits.
AFM calibration procedures Page 10 of 31
where the zero point for o and p has been chosen appropriately.
3. The peaks, p*kl, which are the local maximum in the cross correlation
function, represent positions of the unit cell in the microscope coordi-
nate system. Mathematically it is expressed as
min (∑ e kl2)
a*,b*,p *00
kl (9)
In wording the co-ordinates of the average unit cell vectors a*, b* are
calculated by averaging the nearest neighbour vectors between peaks in
the cross-correlation function, see Figure 4. All positions and vectors
are identified at sub-pixel level.
0.94µm × 0.86µm
20µm × 20µm 20µm × 20µm
Figure 4. Example of an image recorded of a standard, the identified unit cell tem-
plate, and the cross-correlation function. In order for the structures to be dis-
cernable, this example is from a 20µm × 20 µm scan and not for a full range scan.
AFM calibration procedures Page 11 of 31
Figure 5 Result of the calculation reported by the software. The numbers named
“a vector – x”, “a vector – y”, “b vector – x” and “b vector – y” correspond to the
“observed” unit cell vector coordinates ax*' , a*' *'
y , bx
*'
and b y for the a*' and b *'
vectors used in the calculations of the correction parameters. The program can also
correct the image for non-linearity. This option is not used in this procedure. The
check box “Include Border Region” should remain unchecked. Note that the program
can not distinguish between a*' and b *' . Further the program uses a y-axis which is
-90° rotated from the x-axis. It is therefore up to the user to sort out which vec-
tor is which, and to transform the orientations into a “right handed” coordinate
system.
ax c cxy * a*x
= xx * (10)
ay 0 cyy * a*y
bx c cxy * b*x
= xx * (11)
by 0 cyy * b*y
where cxx*, cyy* and cxy* are the unknown correction parameters to be esti-
mated.
Though it has no implication for the actual calculations, we choose by con-
vention the a* =(a*x, a*y) vector to be the one with the largest x*-
coordinate, i.e. the vector most parallel to the x*-axis.
In the metric coordinate system (x, y) the relationship between the average
length La and Lb and the angle γ between closest features on the reference
standard can be written as:
AFM calibration procedures Page 12 of 31
a = La b = Lb a ⋅ b = La ⋅ L b ⋅ cos(γ ). (12)
By inserting (10) and (11) into (12), a restraint between the unknown ma-
trix elements cxx*, cyy* and cxy*, and the certified dimensions La, Lb and γ is
obtained resulting in three equations with three unknowns [8, 9,12]:
A
c xx * = − (13)
*
a y ⋅ b *
x − a* x ⋅ b * y
La ⋅ Lb ⋅ sin(γ )
cyy * =
A (14)
b*x ⋅ b*y ⋅ L2a + a*x ⋅ a*y ⋅ L2b − (a*y ⋅ b*x + ax* ⋅ b*y ) ⋅ La ⋅ Lb ⋅ cos(γ )
cxy * =
A ⋅ (a*y ⋅ b *x − a*x ⋅ b*y ) (15)
where
2 2
A = b*y ⋅ L2a + a*y ⋅ L2b − 2 ⋅ a*y ⋅ b*y ⋅ La ⋅ Lb ⋅ cos(γ ) .
(16)
y x y
x
° °
At 90 scan rotation At 0 scan rotation
Ibsen La, La 60µm × 60µm 512 × 512 0° or 90° 0.14Hz Trace down
~ 895 nm
where the matrix element δxM is due to the fact that the drift will cause
the start of each line in the x-direction to shift (for further details see
[11]).
The microscope coordinate set (x*i, y*j) as function of the actual traced
coordinates can then be expressed as
ax C C xy 1 − δ xC − δ xA − δ xM /m − δ xM a*'x
= x * (20)
ay 0
C y 0 1 − δ yC − δ yA − δ yM a '
y
La ⋅ Lb ⋅ sin(γ)
cyy * ≅
(
A ⋅ 1 − δyM − δyC − δyA ) (23)
where
(25)
A = b *'y2 ⋅ L2a + a*'y2 ⋅ L2b − 2 ⋅ a*'y ⋅ b*'y ⋅ La ⋅ Lb ⋅ cos(γ )
and where
La, Lb, γ dimensions of the transfer standard, from the cer-
tificate
a*’x, a*’y, b*’x, b*’y observed average lattice vector coordinates ex-
pressed in the microscope coordinate system.
Detailed descriptions are given in separate paragraphs below.
= u2p + u((ekl)x )2
Anticipating the standard is approximately aligned with its rows and col-
umns along the x- and y-axes, this is estimated to be a function of scan
size and pitch.
For calibrations using the 895 nm reference standard (see section 2.4) at
scan areas of 50µm × 50µm to 70µm × 70µm we estimate:
cause the tip to be observed to drift over the surface when it is supposed
to stand still. The average position change is estimated to be zero and the
long term average drift speed is estimated to be 0 nm/s. Hence, δxM = δyM =
0.
However, during the time it takes to record one image a finite drift can be
observed. We interpret this as the uncertainty on δxM and δyM. From hundreds
of measurements we estimate average drift speeds (vdx and vdy) in x and y
during the time of the recording of one image to:
vdx = vdy = 0.2 nm/s during initial operation of the microscope.
vdx = vdy = 0.05 nm/s after 1 hour of scanning in a stable environment.
The terms δxM and δyM are corrections to the recorded pixel positions and
quantify the total drift in x and y, respectively, during the scanning of
one image, relative to the respective side lengths, S x* and S *y ,of the im-
age. Taking these numbers as the standard uncertainties of δxM and δyM we
get:
v dx(m / fs ) v dy(m / fs )
u(δ xM ) = , u(δ yM ) =
S x* S *y
where m is the number of scanned lines in the image fs [Hz] is the rate at
which lines are recorded. Thus 1/fs is the time it takes the microscope to
scan one line forward and one line back (trace and retrace).
For images with nominally equal side lengths in the range of 50µm × 50µm to
70µm × 70µm, with 512 lines, and for scan rates of 0.10-0.15Hz, after one
hour of scanning we estimate:
u(δxM) = u(δyM) = 0.005
Model: Y =-(SQRT((X7^2)*(X1^2)+(X5^2)*(X2^2)-2*X5*X7*X1*X2*COS(RADIANS(X3))))/((X5*X6-X4*X7)*(1-X8/X11-X9-X10))
The results for all three correction parameters are summarised in Table 2.
The uncertainties corresponds to the combined standard uncertainties multi-
plied by the coverage factor k = 2, in accordance with EAL-R2 [15].
15
5
z [nm]
-5
-15
0 20 40 60
x [µm]
Figure 7 The top is the definition of a step height in analogy to the ISO stan-
dard 5436. The bottom is the interpretation of the definition for an average
profile. The solid lines is the three parts of two parallel lines fitted to the
profile in the three segment A, B and C .
The auxiliary z-coordinate z(A ) , z(B ) and z(C ) is defined as the average z*-
coordinate of the segment A, B and C of the profile (see Figure 7). The av-
erage off-set z*offset for the profile segment is then defined as
z(C) + 1
z(A) + 1
z(B)
z*offset = 2 2
(28)
2
The reference offset z*offset,ref is an average offset z*offset with an absolute
value of less than 500 nm. Let ∆h be a calibrated step height and ∆z* the
uncorrected difference in z*-coordinate values for the top and bottom line.
The coefficient
∂z ∆h
Czz * = ≅ (29)
∂z * z* = 0 ∆z * z *offset = 0
∂z
C zx * = (30)
∂x * x* = 0
is zero as the X axis and the X* axis are chosen to be parallel and as the
Z axis is perpendicular to the X axis.
The coefficient
AFM calibration procedures Page 21 of 31
∂ 2z
c zz* 2 = (31)
2∂z*2 z*offset=0
is a measure for the linear variation in the linear correction factor as
function of the average off-set. It is estimated as the slope of the cor-
rection factor Czz* as function of the average z-offset, similar to the pro-
cedure in [16].
The coefficient
∂ 2z
Czx *2 = (32)
2∂x*2 x* = 0
is a measure for the image bow. It is estimated as minus the second order
coefficient in a polynomial fit to the average profile z’(xi) of a horizon-
tal and flat surface. To get a more useful and intuitive scale, this cor-
rection factor is quantified as the minimum to maximum value εp of the fit-
ted second order polynomial, over the maximal scan length xlmax , that is
2
x
ε p = Czx*2 l max (33)
2
The term Czx*2 z*(x)x can be ignored as the observed step height is found
not to depend on the horizontal position. The possible significant contri-
bution from higher order terms will be assessed in the uncertainty budget.
If the Z* axis is perpendicular to the X*Y* plane and the microscope coor-
dinate system was a linear function of the position then czz* would be a
scaling factor in the z-direction and all other coefficients would be zero.
The correction coefficients czz* , … are functions of the same parameters as
the sampled coordinates (see section 3.1.1). The measurand during calibra-
tion is defined as the values at equilibrium, at the reference temperature
Tref, the reference relative humidity Href, and a reference off-set
z*offset,ref , that is,
czz * ≡ czz *(Tref, H ref, z*offset,ref)
( 34 )
czz *2 ≡ czz *2(Tref, H ref, z*offset,ref)
czx *2 ≡ czx *2(Tref, H ref, z*offset,ref)
where βx and βy are the tilt angles of the surface of the standard with re-
spect to the scanning plane along the x-axis and the y-axis, respectively.
In practice the tilt angles, βx and βy, are often below 0.5°, which implies
the cosines to deviate from unity with less than 8×10-5. Therefore the
cos(βx) and cos(βy)terms are considered unity in the analysis and the asso-
ciated uncertainty vanishing. For the y-axis βy is estimated as the average
of the up and down scan, for the x-axis βx is estimated as the average of a
trace and retrace scan. If it is higher than 0.5° the images are corrected
by multiplying the z’ with cos(βx)cos(βy) before the plane correction is ap-
plied
2
The basal plane is defined as the bottom part of the surface in case of a continuous surface
with bumps and as the top part of the surface in the case of a continuous surface with pits.
AFM calibration procedures Page 23 of 31
min ∑ (z* (xi ) − ho ) 2 + ∑ z* (xi ) 2 (38)
ho
x *i ∈ C x *i ∈ A
x *i ∈ B
where the profile segment A, B and C are defined in Figure 7, The step
height is calculated using “ISO 5436” under “Z-calibration” under “Process-
ing”. In the windows, which give the parameters for the calculation “Middle
Length”, “Side Length” and “Edge Distance” should all be set to 33.00.
“Groove” or “Line” should be checked as appropriated. The button “Profile”
should be marked.
y = -0.0018x + 1.0019
1.006
R2 = 0.9584
1.004
1.002
H800
1.000
H80
0.998 Linear (H800)
Linear (H80)
0.996
0.994
0.992
-3 -2 -1 0 1 2 3 4
Figure 8 Example of the measured correction factor czz* as function of the average
observed z-coordinate. The squares are correction factors for ca step height of
nominal 800 nm; the diamonds are correction factors for a nominal step height of 80
nm. The non-linear variation of the correction factors as function of the average
z-position is insignificant.
The size of czz*2 is compared with the statement about the nonlinearity of
the z-sensitivity in the registrations for the instruments (“apparatur
kort”) and based on the measurements it is assessed if this statement is
still valid. If the value is significant different from the value in the
registrations for the instruments proper action is taken and documented on
the registration for the instrument or on the internal certificate.
href
czz * = (39)
hobs,ref
Table 4 Parameters for imaging the step height for measuring czz*2
Table 5. Parameters for imaging the step height for measuring czz*2
where the corrections δzt*2 and δzz* are defined in the following sections.
AFM calibration procedures Page 27 of 31
u( czz*2 ) = 0.00002
AFM calibration procedures Page 29 of 31
5. References to literature
Restricting the estimation volume to the basal plane3 of the image using
the “Inside Color Range” option in SPIP means that only observed heights
z'j (xi ) in the basal plane is included in the minimising.
The reference plane is then given by
η (xi , y j ) = A 0j + A1jxi + A 2jxi 2 +... .
(43)
A corrected image z(xi, y j) is then calculated by subtracting the reference
plane from the observed height of the surface
z(xi, y j) = z'(xi, y j) − η(xi, y j)
(44)
If the highest coefficient for x is 1, that is, η j(xi ) = a0j + a1jxi the ref-
erence line is named a 1. order line estimated based on the least square
method. If the highest coefficient for x is 3, that is,
2 3
η j(xi ) = a0j + a1jxi + a2jxi +a3jxi the reference line is named a 3. order line
estimated based on the least square method.
3
The basal plane is defined as the bottom part of the surface in case of a continuous surface
with bumps and as the top part of the surface in the case of a continuous surface with pits.