U.S. patent application number 17/581888 was filed with the patent office on 2022-05-19 for system and method for autonomous scanning probe microscopy with in-situ tip conditioning.
This patent application is currently assigned to Quantum Silicon Inc.. The applicant listed for this patent is The Governors of the University of Alberta, Quantum Silicon Inc.. Invention is credited to Mohammad Rashidi, Robert Wolkow.
Application Number | 20220155339 17/581888 |
Document ID | / |
Family ID | 1000006108103 |
Filed Date | 2022-05-19 |
United States Patent
Application |
20220155339 |
Kind Code |
A1 |
Rashidi; Mohammad ; et
al. |
May 19, 2022 |
SYSTEM AND METHOD FOR AUTONOMOUS SCANNING PROBE MICROSCOPY WITH
IN-SITU TIP CONDITIONING
Abstract
A method for assessing the quality of a tip of a scanning probe
microscope (SPM) includes recording an SPM image, extracting a
plurality of images of dangling bonds from the SPM image, feeding
the extracted images of dangling bonds into a convolution neural
network one image at a time, analyzing each of the plurality of
images of dangling bonds using the convolution neural network,
assigning each of the plurality of images of dangling bonds one of
a sharp tip status or a double tip status, and determining whether
the number of the plurality of images of dangling bonds of the SPM
image assigned the double tip status exceeds a predetermined
threshold. A method of automatically conditioning a tip of a
scanning probe microscope (SPM) during imaging of a sample and a
method of mass-producing atomistic quantum dots, qubits, or
particular atom orbital occupation are also provided.
Inventors: |
Rashidi; Mohammad;
(Edmonton, CA) ; Wolkow; Robert; (Edmonton,
CA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Quantum Silicon Inc.
The Governors of the University of Alberta |
Edmonton
Edmonton |
|
CA
CA |
|
|
Assignee: |
Quantum Silicon Inc.
Edmonton
CA
The Governors of the University of Alberta
Edmonton
CA
|
Family ID: |
1000006108103 |
Appl. No.: |
17/581888 |
Filed: |
January 22, 2022 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
16981512 |
Sep 16, 2020 |
|
|
|
PCT/CA2019/050333 |
Mar 19, 2019 |
|
|
|
17581888 |
|
|
|
|
62644974 |
Mar 19, 2018 |
|
|
|
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
G06N 3/08 20130101; G01Q
30/06 20130101; G01Q 80/00 20130101 |
International
Class: |
G01Q 30/06 20060101
G01Q030/06; G01Q 80/00 20060101 G01Q080/00; G06N 3/08 20060101
G06N003/08 |
Claims
1. A system comprising: a scanning probe microscope (SPM) with a
probe tip, where an automated assessment of the quality of the
probe tip is determined while the probe tip is scanning or
atomically manipulating a surface of a given sample, said system
further comprising: machine learning algorithms that monitor and
assess probe tip condition and determine when probe tip
reconditioning is required; a recording device configured to store
a series of SPM images; and wherein the system is configured to:
extract a plurality of images of dangling bonds from each of the
SPM images; feed the plurality of extracted images of dangling
bonds into a convolution neural network one image at a time;
analyze each of the plurality of extracted images of dangling bonds
using the convolution neural network; assign each of the plurality
of extracted images of dangling bonds one of a sharp tip status or
a double tip status; and determine whether the number of the
plurality of extracted images of dangling bonds of the SPM image
assigned the double tip status exceeds a predetermined
threshold.
2. The system of claim 1 wherein the convolution neural network
trains the machine learning algorithms.
3. The system of claim 1 wherein the SPM image is recorded at a
sample bias of -1.8 V.
4. The system of claim 1 wherein the SPM image is recorded at 50
pA.
5. The system of claim 1 wherein the plurality of extracted images
of dangling bonds appear as bright protrusions in the SPM image
recorded.
6. The system of claim 1 wherein each of the plurality of extracted
images of dangling bonds are 6.times.6 nm.sup.2.
7. The system of claim 1 wherein the SPM image recorded is
100.times.100 nm.sup.2.
8. The system of claim 1 wherein the sample is a
hydrogen-terminated Si(100) surface.
9. The system of claim 1 wherein the convolution neural network
includes a first convolution layer, a second convolution layer, a
pooling layer, a densely connected layer, and an output layer.
10. The system of claim 1 wherein the analysis of each of the
plurality of extracted images of dangling bonds includes pixilation
of each of the images of dangling bonds.
11. The system of claim 1 further comprising an interface to alert
a user of the SPM when the number of the plurality of extracted
images of dangling bonds of the SPM image assigned the double tip
status exceeds the predetermined threshold.
12. The system of claim 1 wherein the machine learning algorithms
are further configured to set an image frame and a location within
the image frame to automatically condition the tip when it is
determined that a number of the plurality of extracted images of
dangling bonds of the SPM image within the image frame assigned a
double tip status exceeds the predetermined threshold.
13. The system of claim 12 wherein an assessment of the quality of
the tip of the scanning probe microscope (SPM) is performed within
the image frame on the sample until it is determined that the
number of the plurality of extracted images of dangling bonds of
the SPM image within the image frame assigned the double tip status
exceeds the predetermined threshold.
14. The system of claim 12 wherein the location of tip conditioning
is a location on the sample outside of the image frame.
15. The system of claim 12 wherein the location for tip
conditioning is a location where the tip is conditioned when it is
determined that the number of the plurality of images of dangling
bonds of the SPM image within the image frame assigned the double
tip status exceeds the predetermined threshold.
16. A system that mass produces atomistic quantum dots, qubits, or
particular atom orbital occupation, the system comprising: a
scanning probe microscope (SPM) having a probe tip that scans or
atomically manipulates the surface of a given sample, said system
further comprising: machine learning algorithms that monitor and
assess probe tip condition and determine when tip reconditioning is
required; a recording device configured to store a series of SPM
images; and wherein the system is configured to: extract a
plurality of images of dangling bonds from each of the SPM images;
feed the extracted images of dangling bonds into a convolution
neural network one image at a time; analyze each of the plurality
of extracted images of dangling bonds using the convolution neural
network; assign each of the plurality of extracted images of
dangling bonds one of a sharp tip status or a double tip status;
and determine whether the number of the plurality of extracted
images of dangling bonds of the SPM image assigned the double tip
status exceeds a predetermined threshold; and wherein the system is
further configured to: selectively sense or modify an orbital
occupation state of a given atom with the SPM; repeat the selective
sensing or modifying of the orbital occupation state of a plurality
of additional individual atoms with the SPM; repair the SPM or
replace the SPM with a new SPM; and repeat the selective sensing or
modifying the orbital occupation state of a second plurality of
additional individual atoms with the repaired SPM or the new
SPM.
17. The system of claim 16 wherein the convolution neural network
trains the machine learning algorithms.
18. The system of claim 16 wherein the convolution neural network
includes a first convolution layer, a second convolution layer, a
pooling layer, a densely connected layer, and an output layer.
19. The system of claim 16 wherein the plurality of images of
dangling bonds appear as bright protrusions in the SPM image
recorded.
20. The system of claim 16 wherein the sample is a
hydrogen-terminated Si(100) surface.
Description
RELATED APPLICATIONS
[0001] This application is a continuation of U.S. application Ser.
No. 16/981,512, filed Sep. 16, 2020, which in turn is a US National
Phase of PCT Application Serial Number PCT/CA2019/050333, filed
Mar. 19, 2019, which in turn claims priority benefit of U.S.
Provisional Application Ser. No. 62/644,974, filed Mar. 19, 2018;
the contents of the aforementioned applications are hereby
incorporated by reference.
FIELD OF THE INVENTION
[0002] The present invention is directed generally to scanning
probe microscopes (SPMs), and more specifically to scanning probe
microscopes designed to use machine learning technology to predict
failure of a probe tip and to automatically re-functionalize or
condition the probe tip in-situ.
BACKGROUND
[0003] Scanning probe microscopes (SPMs) are a family of tools used
to make images of nanoscale surfaces and structures, including
atoms. They use a physical probe to scan back and forth over the
surface of a sample. During this scanning process, a computer
gathers data that is used to generate an image of the surface.
[0004] As shown in FIG. 1, an SPM has a probe tip that can be as
sharp as a single atom. The tip is moved precisely and accurately
back and forth across the surface, even atom by atom. When the tip
is near the sample surface, SPMs can measure tip-surface
interactions caused by many kinds of effects, including tunneling
current, electrostatic forces, magnetic forces, chemical bonding,
van der Waals forces, and capillary forces. SPMs can detect
differences in height that are a fraction of a nanometer, about the
diameter of a single atom. A computer combines the data to create
an image.
[0005] There are several types of SPMs. Atomic force microscopes
(AFMs) measure the electrostatic forces between the cantilever tip
and the sample. Magnetic force microscopes (MFMs) measure magnetic
forces, and scanning tunneling microscopes (STMs) measure the
tunneling current flowing between the tip and the sample.
[0006] In addition to visualizing nanoscale structures, SPMs are
used to manipulate individual atoms and molecules and move them to
make specific patterns. For example, SPMs can be used to add or
remove hydrogen atoms from a silicon surface to manipulate multiple
dangling bonds (DBs) on an otherwise H-terminated silicon surface
that form quantum dots. A DB corresponds to a desorbed single
hydrogen atom from the otherwise passivated silicon surface. It is
approximately an sp3 hybrid orbital that can be occupied by 2, 1,
or 0 electrons resulting, respectively, in a negative, neutral, or
positively charged DB. Thus, a DB behaves essentially as a single
atom quantum dot controlled atom-by-atom lithography, i.e.,
hydrogen desorption, on the H--Si surface allows creation of DB
based circuits for next generation ultimately miniaturized low
power nanoelectronic devices beyond complementary metal oxide
semiconductor (CMOS) technology. In principle, the capacity of this
technology provides smaller, more efficient, faster, and cheaper
electronics compared to the currently dominant CMOS technology.
[0007] Hydrogen-terminated silicon surfaces are a promising
platform for atomic circuitry. This is due, among other things, to
its inertness, thermal stability, very low defect density, and
particularly the unique properties of its surface dangling bonds.
Fundamental properties of isolated dangling bonds have been studied
extensively. It has been shown that they are electronically
isolated from the host substrate and can hold charge, thereby
acting as atomic quantum dots. They can be precisely placed and
erased. Various functional elements such as
quantum-cellular-automaton cells, binary transmission wires, and
binary computational gates can be made by arranging dangling bonds
close to each other.
[0008] SPM techniques and in particular atomic manipulation rely on
atomically sharp tips. The tip preparation is done in different
steps. First, a metal wire, commonly tungsten, is electrochemically
etched in ambient condition, resulting in few tens of nanometer
large tips. This step is usually followed by thermal annealing in
ultra-high vacuum to clean the oxide layer formed at the apex of
the tip during the etching process. The tip can be additionally
sharpened down to the single atom via field ion microscopy. The
carefully ex-situ prepared atomically sharp tips occasionally
become non-ideal or fail (generally called "double tips") during
imaging and atomic manipulation due to the interaction with the
surface. SPM data collected with a failed tip is generally
unreliable and unusable. In-situ tip treatments are necessary to
return the tip to its ideal (sharp) condition. In-situ tip
sharpening can be performed but is usually the most time-consuming
process for SPM operators. It is often done by applying voltage
pulses between the tip and the sample or by a controlled
indentation of the tip into the sample. This process most often
needs to be repeated several times before obtaining the desired tip
again.
[0009] Accordingly, current SPM techniques are too slow to be
practical. Therefore, there exists a need for a rapid and
commercially feasible way to mass produce electronic components
with atomic precession.
SUMMARY
[0010] A method for assessing the quality of a tip of a scanning
probe microscope (SPM) is provided. The method includes recording
an SPM image, extracting a plurality of images of dangling bonds
from the SPM image, feeding the extracted images of dangling bonds
into a convolution neural network one image at a time, analyzing
each of the plurality of images of dangling bonds using the
convolution neural network, assigning each of the plurality of
images of dangling bonds one of a sharp tip status or a double tip
status, and determining whether the number of the plurality of
images of dangling bonds of the SPM image assigned the double tip
status exceeds a predetermined threshold.
[0011] A method of automatically conditioning a tip of a scanning
probe microscope (SPM) during imaging of a sample is also provided.
The method of automatically conditioning a tip includes setting an
image frame and a location for tip conditioning, performing the
method for assessing the quality of a tip, as described above,
within the image frame on the sample, conditioning the tip when it
is determined that the number of the plurality of images of
dangling bonds of the SPM image within the image frame assigned the
double tip status exceeds the predetermined threshold.
[0012] A method of mass producing atomistic quantum dots, qubits,
or particular atom orbital occupation is also provided that
includes selectively sensing or modifying the orbital occupation
state of a given atom with an SPM, repeating the selectively
sensing or modifying the orbital occupation state of a plurality of
additional individual atoms with the SPM, performing a method of
assessing the quality of a tip of a scanning probe microscope (SPM)
and/or conditioning the tip of a SPM, repairing the SPM or
replacing the SPM with a new SPM, and repeating the selectively
sensing or modifying the orbital occupation state of a second
plurality of additional individual atoms with the repaired SPM or
the new SPM.
BRIEF DESCRIPTION OF THE DRAWINGS
[0013] FIG. 1 shows a prior art schematic drawing of the components
of a scanning probe microscope;
[0014] FIG. 2 shows a compilation of SPM images recorded under
similar conditions and used for training the tested machine
learning models;
[0015] FIGS. 3A and 3B show two SPM images of FIG. 2 labeled as a
good tip and a double tip, respectively;
[0016] FIG. 4 shows images of dangling bonds extracted from an SPM
image of FIG. 2;
[0017] FIG. 5 is a graph of the result of the grid search for a
support vector machine;
[0018] FIG. 6 is a table showing the best parameters determined
from the graph of FIG. 5;
[0019] FIG. 7 is a table showing the best parameters determined
from the random forest;
[0020] FIG. 8 shows a collection of misclassified dangling bond
images of the random forest of FIG. 7;
[0021] FIG. 9 is a schematic representation of a fully connected
neural network;
[0022] FIG. 10 is a graph showing the accuracy score of the fully
connected neural network (FCNN) in FIG. 9 for different number of
hidden layers;
[0023] FIG. 11 is a table showing the determined best depth of the
FCNN based on the results of the train and test data;
[0024] FIG. 12 is a table showing the nearest neighbor result;
[0025] FIGS. 13A-13D show the workflow of a tip quality analysis
process using a CNN according to an embodiment of the present
disclosure;
[0026] FIG. 14 shows a CNN according to an embodiment of the
present disclosure;
[0027] FIG. 15 shows a compilation of dangling bond images
misclassified by the CNN of FIG. 14;
[0028] FIG. 16 shows a user interface for automating SPM tip
conditioning according to an embodiment of the present
disclosure;
[0029] FIG. 17 shows an image frame and location for tip
conditioning selected by a user;
[0030] FIG. 18 shows a SPM image recorded within the image frame of
FIG. 17 with the dangling bonds that selected by the automation
routine (identified in dashed line boxes);
[0031] FIG. 19 shows the identified dangling bonds with double tip
statuses assigned by the CNN;
[0032] FIG. 20 shows the outcome of image analysis by the CNN in
which it is determined the tip is a double tip;
[0033] FIGS. 21A-21F show a sequence of SPM images during an
autonomous tip reconditioning process according to an embodiment of
the present disclosure;
[0034] FIGS. 22A and 22B show the outcome of dangling bond image
analysis by the CNN of a conditioned tip having a sharp status
compared to a double bond tip SPM image, respectively; and
[0035] FIGS. 23A-23E show a sequence of SPM images captured and
analyzed by the CNN during an automated tip conditioning process
according to an embodiment of the present disclosure.
DESCRIPTION OF THE INVENTION
[0036] The present invention has utility as a system and method for
autonomous scanning probe microscopy with an automated process for
determining the condition of the tip and in-situ tip conditioning
for a probe tip when it is determined that the probe tip has failed
or is no longer optimal. The present invention provides scanning
probe microscopes designed to use machine learning technology to
predict failure of a probe tip and to automatically
re-functionalize or recondition the probe tip in-situ thereby
reducing the need for user intervention and failed experiments or
failed scans due to failed probe tips. The present invention
provides the ability to automate a vast number of atomic precision
robots to manipulate and analyze atoms in parallel with limited
human intervention.
[0037] The present invention has utility in the automated
production of atomistic quantum dots, qubits, or selective bonding
to a dangling bond (unoccupied orbital extending from a surface
atom). Such structures have been detailed for example in
PCT/IB2006/000510 and PCT/IB2009/006400, the contents of which are
hereby incorporated by reference. Besides facilitating mass
production, or operation of the quantum computing elements, it is
appreciated that a given atomic orbital functions as a binary
storage element as unoccupied or occupied. As a result of a given
atom functioning as an atomic scale transistor operative for data
storage. With reliable modification or sensing of orbital
occupation, large scale atomic computation and data storage becomes
practical, as compared to the current labor intensive
processes.
[0038] While the present disclosure describes automated SPM atomic
manipulation and tip re-conditioning in the context of
hydrogen-terminated silicon surfaces, it will be readily understood
that the system and methods described herein are not limited to
such hydrogen-terminated silicon substrates and can be adapted for
other systems as well.
[0039] The present disclosure provides a system and automated
method for detecting the condition of a probe tip of an SPM device
and changes in the condition. The system and method can detect
conditions such as the sharpness of the probe tip and other
qualities contributing to SPM image quality like chemical,
magnetic, or optical qualities. Upon detecting loss or decrease of
a desired quality such as sharpness (commonly called a "double
tip") the automated system and method of the present disclosure
reconditions the probe tip, for example by sharpening the tip by
some objective criteria, or by re-establishing magnetic, chemical,
or optical quality. Embodiments of the system and method for
automated probe tip condition detection provide for the automated
detection of the condition of the probe tip during operation of the
SPM device, that is while the probe tip is scanning or atomically
manipulating the surface of a given sample, the condition of the
probe tip is monitored for degradation. Upon detection of a
degraded probe tip condition, the operation of the SPM device is
paused and the probe tip is automatically reconditioned without
human intervention. The system and method also provide an automated
awareness or detection of when reconditioning of the probe tip is
achieved, thereupon the operation of the SPM device resumes with
the probe tip continuing its scanning or atomic manipulation
operation from the location at which the operation was paused.
[0040] The inventive system and methods of the present disclosure
implement machine learning algorithms to monitor and assess probe
tip condition and determine when tip reconditioning is required.
The automation process is trained using machine learning. A
convolution neural network (CNN) model is used in the training. The
CNN includes two hidden layers. As shown in FIG. 2, the first step
of the training process includes compiling a number of SPM images
recorded under similar conditions, for example those images
recorded under conditions of -1.8V and greater than 10.times.10 nm.
Next, as shown in FIGS. 3A and 3B, each of the extracted SPM images
are labeled as being a good tip scan or an undesirable degraded tip
scan, for example a double tip. Next, from the extracted and
labeled SPM images, images of the dangling bonds are further
extracted. The dangling bonds appear in the SPM images as bright
protrusions, such as those shown in FIG. 4. Next, the images of the
dangling bonds are split into a train set and a test set. The
images of the dangling bonds in the train set are augmented by
rotating the images four times by 90.degree. and then mirrored each
image, such that the images in the train set are multiplied by
eight times. Next, the images of the dangling bonds in the train
set are input into the CNN one at a time such that the CNN learns
to recognize a good tip image and a double tip image. FIG. 9 shows
a fully connected neural network such as the one used herein. Next,
the test set of dangling bond images are used to test the training
of the CNN machine learning. The most time consuming part of the
training is to establish the best parameters to fit the data and to
get the best performance on the test set. This is done by grid
search and cross validation. The result of the grid search for two
support vector machine (SVM) parameters is shown in FIG. 5. FIG. 6
shows the result of the best parameters. FIG. 7 shows the
parameters established from the random forest. Upon completion of
the training, the machine learning algorithm of the CNN obtains a
precision score of 97%. FIG. 9 is a graph showing a fully connected
neural network. As shown in FIG. 10, the best depth based on the
comparison of the train and test data is at a depth of 18.
[0041] FIGS. 13A-13D show a workflow of the tip quality analysis
using a CNN. During a SPM scanning or atomic manipulation
operation, the inventive method records and SPM images of a sample
surface as shown in FIG. 13A. The surface in FIG. 13A is a
hydrogen-terminated Si(100) recorded at -1.8 V and 50 pA. According
to various embodiments, the SPM image recorded is 100.times.100
nm.sup.2. Dangling bonds on the surface are shown as bright
features. Here the bright features have been segmented from the SPM
image as shown by the dotted line boxes shown in FIG. 13A. The
method continues with the images of the dangling bonds of FIG. 13A
being automatically extracted and fed into the CNN one at a time.
FIG. 13B shows one of the images of the dangling bonds from the
recorded SPM image shown in FIG. 13A. According to some embodiments
each of the images of dangling bonds are 6.times.6 nm.sup.2.
[0042] Each of the extracted images of dangling bonds is fed into
the CNN, which is shown at FIG. 13C. The CNN includes two
convolution layers followed by a pooling layer, a densely connected
layer, and an output layer. As indicated in FIG. 13C, the two
convolutional layers of the CNN have 30 and 40 kernels,
respectively. The CNN uses its machine learning training to analyze
the images of the dangling bonds for characteristics indicative of
the tip quality. Further details of the CNN processing of a
dangling bond image are shown in FIG. 14. The CNN pixilates the
dangling bond image to examine each dangling bond image for
characteristics indicative of a good tip or a double tip. For each
image of a dangling bond, the CNN makes a determination of a sharp
tip or a double tip. For sharp tips, the output layer of the CNN
assigns a dangling bond image the number "0." For double tips, the
output layer of the CNN assigns a dangling bond image the number
"1." FIG. 13D displays the output of the CNN for all the dangling
bonds in the SPM image of FIG. 13A. The method then continues by
determining how many dangling bond images of the SPM image have
been assigned the double tip, "1" status, for example by performing
a majority voting operation. Implementing the majority voting
operation increases the operational accuracy of the CNN to nearly
100%. When the number of dangling bond images of the SPM image that
have been assigned the double tip, "1" status exceeds a
predetermined threshold, a user may be alerted by a light or a
sound on the SPM device and/or an automatic tip reconditioning
process is undertaken.
[0043] The present disclosure further provides a method of
automatically conditioning a tip of a scanning probe microscope
(SPM) during imaging of a sample. The method beings with a user
setting an image frame and a location for tip conditioning, such as
that shown in FIG. 17. The location of tip conditioning is a
location on the sample outside of the image frame. In FIG. 17, the
image frame is shown within the dashed line, and the location of
tip conditioning is shown by the "X." The method continues by
imaging the sample within the image frame using SPM and performing
the above described method for assessing the quality of a tip of a
scanning probe microscope within the selected image frame on the
sample. As described above, images of the dangling bonds are
extracted, as shown in FIG. 18, and the CNN assess quality of the
tip based on characteristics of the dangling bond images. The CNN
assigns a good tip status of "0" or a double tip status of "1" to
each dangling bond image, as shown in FIG. 19. The CNN then does a
voting operation to determine if the probe tip is sharp or double.
Depending on the outcome of this assessment, the program either
stops operation (for sharp tips) or repeats the same procedure (for
double tips). The tip assessment is repeated within the image frame
on the sample until it is determined that the number of images of
dangling bonds of the SPM image within the image frame assigned the
double tip status exceeds the predetermined threshold, such as the
CNN output shown in FIG. 20.
[0044] When it is determined that that the number of the dangling
bond images of the SPM image within the image frame assigned the
double tip status exceeds the predetermined threshold, such as the
CNN output shown in FIG. 20, an automatic in-situ tip conditioning
process is undertaken. In-situ tip reconditioning is done by
applying voltage pulses between the tip and the sample or by a
controlled indentation of the tip into the sample. This process
most often needs to be repeated several times before obtaining the
desired condition of the tip again. FIGS. 21A-21F show an example
of autonomous tip reconditioning according to an embodiment of the
present disclosure. FIG. 21A shows an initial SPM image within the
image frame that is used to judge the quality of the tip based on
the automated tip condition assessing process described above. The
tip is automatically re-conditioned by a controlled indentation at
the location of tip conditioning on the surface set by the user.
FIGS. 21B-21E shown the sequence of images recorded automatically
to judge the condition of the tip after each tip conditioning
operation. Each image of FIGS. 21B-21E is taken after the
controlled indentation at the location of tip conditioning set by
the user. The majority vote outputs of the CNN is "1" (double tip)
for all of the images in FIGS. 21B-21E, indicating that each round
of the tip conditioning was not successful. As shown in FIG. 21F,
after a successful tip conditioning, the majority vote output of
the CNN becomes "0" (sharp tip) and the program stops its
operation. FIGS. 22A and B shows the outcome of dangling bond image
analysis by the CNN of a conditioned tip having a sharp status as
compared to the double tip SPM image.
[0045] Further embodiments of the present disclosure provide a SPM
device that is further automated such that the samples are fed into
the device for SPM analysis automatically, thereby further reducing
the need for human intervention and observation.
[0046] As a person skilled in the art will recognize from the
previous detailed description and from the figures and claims,
modifications and changes can be made to the preferred embodiments
of the invention without departing from the scope of this invention
defined in the following claims.
* * * * *