Mechano-spectroscopy to identify sample material based on atomic force microscopy

Abstract

A method includes using an atomic-force microscope to train a machine-learning circuitry to classify a first pixel of an image as corresponding to a first material and a second pixel of the image as corresponding to a second material and to do so based on mechanical properties of said first and second materials. The image is an image of a sample that comprises materials selected from a finite group of materials that includes the first and second materials.

Description

[0001] MECHANO-SPECTROSCOPY TO IDENTIFY SAMPLE MATERIAL BASED ON ATOMIC FORCE MICROSCOPY

[0002] CROSS REFERENCE TO RELATED APPLICATIONS

[0003] This application claims the benefit of the filing date of and priority to U.S. Application No. 63 / 521,462, filed on June 16, 2023, the contents of which are hereby incorporated by reference in their entirety.

[0004] FIELD OF INVENTION

[0005] The invention relates to scanning probe microscopy, also known as atomic force microscopy, and in particular, to processing of the output of an atomic force microscope with the help of machine learning methods to identify the sample material using physico-mechanical properties of the sample.

[0006] BACKGROUND

[0007] Scanning probe microscopy, also known as atomic force microscopy, is a way to study force interactions between a probe and sample surface. It is broadly used to visualize features on a sample surface that have dimensions from hundreds of microns down to sub-nanometer scale.

[0008] In atomic force microscopy, a probe attached to a cantilever responds to forces that result from proximity or contact with a sample surface. This force influences the probe’s motion. The probe’s motion therefore provides a measure of the force. By scanning the probe's tip across the surface of a sample, one can obtain values of the extent of deflection as a function of spatial position. This allows one to build a two- dimensional map showing values of a surface parameter.

[0009] A difficulty that arises is that building such a map takes time because an entire region needs to be scanned. An obvious solution to the problem of slow scanning is to scan faster. But a cantilever arm has its own resonance characteristics. When one scans faster, the output signal of the microscope becomes a superposition of three components. The first component consists of the deflections caused by slow mechanical interaction with the surface. The second component consists of various second-order effects, such as deflections that arise from all other sources, including those that arise from the dynamic properties of the cantilever arm itself, and / or from the sliding interaction between the AFM probe and sample surface. The third component consists of instrumental and environmental noise. From the combination of these three components, one can filter out an output signal which is indicative of multiple physical and mechanical properties of the sample.

[0010] The deflection of the cantilever from its equilibrium position provides a signal from which a great deal of information can be extracted. As an example, by keeping either the loading force or the cantilever’s deflection constant, it is possible to obtain the sample’s topology at various points on the sample. The values collected at each point are then organized into an array in which the row and column identifies the location of a point in a two-dimensional coordinate system and the value at the row and column is representative of a property measured at that point. The resulting array of numbers can thus be viewed as a map. This makes it possible to make a map of the sample in which each point on the map indicates some property of the sample’s surface at that point. In some examples, the property is the height of the surface above or below some reference plane.

[0011] The cantilever’s deflection can be used to collect multiple images of the sample’s surface, with each image being a map of a different property of the surface. Recently introduced sub resonance tapping, in particular Ringing mode, allows extracting multiple physical and mechanical properties of the sample. Examples of just a few of these properties include adhesion between the probe and the surface, the stiffness of the surface, and viscoelastic energy loss.

[0012] A deficiency of atomic force microscopy is the absence of compositional information about the sample. To overcome this, combinations of atomic force microscopy and material-defining spectroscopy have been developed. Examples include combinations of atomic force microscopy with one or more of: Raman spectroscopy, infrared spectroscopy, electron microscopy elements analysis, all combined with atomic force microscopy.

[0013] SUMMARY

[0014] The invention provides a method for identifying a material composition of a sample surface using multidimensional images of physico-mechanical properties of samples obtained by an atomic force microscope. According to the invention, it is possible to obtain a multi-dimensional image of a surface with two of the dimensions corresponding to spatial dimensions and additional dimensions corresponding to different physical and mechanical properties that exist at the coordinate identified by the two spatial dimensions.

[0015] A question that arises is how one chooses and uses these different physical and mechanical properties for identification of the material of a surface. The identification of the surface material is based on application of a machine-learning circuitry applied to the multidimensional images of the sample surface. The machine-learning circuitry is trained based on a prerecorded database of multidimensional images of known materials. The machine-learning circuitry is applied to each pixel of the image, thereby allowing identifying the material at each pixel of the sample image.

[0016] Among the embodiments of the invention are those that include using an atomic force microscope to acquire different maps corresponding to different properties of the surface and using combinations of physical and mechanical properties of the sample recorded in these maps, to identify the sample material at each pixel of the sample image. Such a method comprises recording atomic force microscope images of examples of surfaces of well-defined materials, forming a database in which such atomic force microscope maps are associated with the particular material, using the atomic force microscope maps thus obtained and the combinations thereof to train the machinelearning circuitry how to identify for sample material, for example by building a decision tree or neural network or a combination of thereof. The method is agnostic to the nature of the surface material as well as to the material of the microscope’s probe. For example, one might use the disclosed method to identify particular polymer phases in a complex composite mix to identify different components in composite materials, to identify different biomaterials, composition of biological membranes.

[0017] The invention features using atomic force microscope that can produce a multidimensional array of physical-mechanical properties, for example, when using subresonance tapping mode. In some practices, acquiring the set of images comprises using an atomic-force microscope in Ringing mode to carry out nanoscale-resolution scanning of the surfaces of a sample.

[0018] Among the embodiments of this invention are those in which the microscope is used in sub-resonance tapping mode. In yet other embodiments, the microscope is used in Ringing mode.

[0019] In one aspect, the invention features using an atomic-force microscope to train machine-learning circuitry to classify a first pixel of an image of a test sample as corresponding to a first material and a second pixel of the image as corresponding to a second material and to do so based on mechanical properties of the first and second materials. The test sample comprises materials selected from a finite group of materials that includes the first material and the second material.

[0020] Among the practices of the method are those that include obtaining the training samples in which the locations of the first and second materials are known and then obtaining multi-dimensional images of the training samples and storing the images in a database. In these practices, training the machine-learning circuitry comprises using the multidimensional images for training the machine-learning circuitry.

[0021] Still other practices include obtaining training samples that comprise the first and second materials at unknown locations on the sample, obtaining multi-dimensional images of the training samples, and storing the images in a database. In these practices, training the machine-learning circuitry likewise comprises using the multidimensional images for training the machine-learning circuitry.

[0022] In some practices, the first and second materials are polymers.

[0023] Also among the practices of the invention are those that include operating the microscope in sub-resonance tapping mode on the test sample and those that include obtaining the image by operating the microscope in Ringing mode on the test sample to carry out nanoscale-resolution scanning of the surfaces of the test sample.

[0024] Still other practices include those in which the image is obtained by the test sample at a controlled temperature, at a controlled humidity, or both.

[0025] Additional practices include obtaining the image comprises imaging the sample in a fluid, with or without a controlled temperature. Among the foregoing practices are those in which the fluid is a gas, those in which the fluid is a gas mixture having an organic vapor at a controlled partial pressure, those in which the fluid is an aqueous medium, and those in which the fluid is that comprises a mixture of an aqueous medium and at least one organic solvent that is miscible in the aqueous medium.

[0026] As used herein, “atomic force microscopy,” “AFM,” “scanning probe microscopy,” and “SPM” are to be regarded as synonymous. Sub-resonance tapping mode is regarded as a scanning mode of atomic force microscopy in which the distance between the microscope’s probe and the sample changes in an oscillatory sinusoidal manner at a frequency that is below the resonance frequency of the AFM cantilever.

[0027] The only methods described in this specification are non-abstract methods. Thus, the claims can only be directed to non-abstract implementations. As used herein, “nonabstract” is a deemed to mean compliant with the requirements of 35 USC 101 as of the filing of this application. These and other features of the invention will be apparent from the following detailed description and the accompanying figures, in which:

[0028] BRIEF DESCRIPTION OF THE FIGURES

[0029] FIG. 1 shows a simplified diagram of one example of an atomic force microscope;

[0030] FIG. 2 shows additional details from the processing system of FIG. 1 ;

[0031] FIG. 3 shows acquisition of images to be used for training machine-learning circuitry;

[0032] FIG. 4 shows a mix of three different polymers with polymer identification having been carried out by processing twelve different maps of physical or mechanical channels through the machine-learning circuitry;

[0033] FIG. 5 shows the results obtained using three different machine-learning methods;

[0034] FIG. 6 shows a mix of three polymers with particular polymer associated with each area having identifiable by either its morphology or by a method such as Raman microscopy;

[0035] FIGS. 7 and 8 show an example that highlights the resolution of the machinelearning circuitry when applied to detect a boundary between two materials; and

[0036] FIG. 9 shows contrast between two surfactant materials at different levels of humidity.

[0037] DETAILED DESCRIPTION

[0038] FIG. 1 shows an atomic force microscope 10 having a scanner 12 that supports a cantilever 14 having a probe 16 attached at a distal end thereof. The scanner 12 moves the probe 16 along a scanning direction that is parallel to a reference plane of a sample’s surface 18. In doing so the scanner 12 scans a region of a sample’s surface 18. As the scanner 12 moves the probe 16 in the scanning direction, it also moves it in a vertical direction, i.e., perpendicular to the reference plane of the sample surface 18. This causes the distance from the probe 16 to the surface 18 to vary as the probe 16 scans the sample’s surface 18.

[0039] As the probe 16 scans the sample’s surface 18, a laser 20 directs an illumination beam 22 towards a mirror 24 formed by a reflective portion of the cantilever 14. This mirror 24 directs a reflected beam 26 towards a photodetector 28.

[0040] As the probe 16 moves, so too does the mirror 24. This causes the reflected beam 26 to illuminate different parts of the photodetector 28. The position of the illuminated portion of the photodetector 28 is encoded in a probe signal 30, which therefore encodes information about the probe’s movement.

[0041] A processing system 32 receives this probe signal 30 as well as a scanner signal 34 that indicates the probe’s position on the sample’s surface 18. In some embodiments, the processing system 32 comprises FPGA electronics to permit real time calculation of surface parameters based on physical or geometric properties of the surface.

[0042] The processing system 32 then uses the probe signal 30 and the scanner signal 34 to calculate certain surface parameters 36 using methods described herein. The processing system 32 then outputs these surface parameters 36 to a storage medium 38.

[0043] FIG. 2 shows the processing system 32 in detail. The processing system 32 features a power supply 40 having an AC source 42 connected to an inverter 44. The power supply 40 provides power for operating the various components described below. The processing system further includes a heat radiator 46 to dissipate excess heat. In a preferred embodiment, the processing system 32 further includes a user interface 48 to enable a person to control its operation.

[0044] Within the processing system 32, first and second A / D converters 50, 52 receive the probe signal 30 and the scanner signal 34, respectively. These signals are placed on a bus 54. A program storage section 56, a working memory 58, and CPU registers 60 are also connected to the bus 54. A CPU 62 for executing instructions 66 from the program storage section 56 connects to both the registers 60 and an ALU 64.

[0045] A non-transitory computer-readable medium stores these instructions 66. When executed, the instructions 66 cause the processing system 32 to calculate any of the foregoing parameters based on inputs received through the first and second A / D converters 50, 52. These calculations comprise non-mental steps that are not executed by a human being.

[0046] The processing system 32 further includes machine-learning circuitry 68 and a database 70 that includes training data 72 and testing data 74. The machine-learning circuitry 68 uses the training data 72 and the testing data 74 for implementing the method described herein to train the machine-learning circuitry 68.

[0047] FIG. 3 shows a process that uses the atomic force microscope 10 of FIG. 1 to acquire images 78 and to provide those images 78 to the machine-learning circuitry 68 in FIG. 2 for use in identification of a particular material in the sample.

[0048] The process shown in FIG. 3 includes a preparation step 80 in which samples of known materials are prepared. This is followed by a scanning step 82 in which the samples are scanned to form images 78 of the known materials. In a storing step 84, these images 78 are stored in the database 70. The images 78 shown in FIG. 3 serve as a nonrestrictive example of three materials imaged through twelve different channels. The process continues with a training step 86 in which the images 78 that were stored in the database 70 are used to train the machine-learning circuitry 68.

[0049] Each image 78 is an array in which each element of the array represents a property of the surface 18 at a particular location. The location of the element within the array corresponds to a spatial location on the sample’s surface 18. Thus, the image 78 defines a map corresponding to that property. Such a map shows the values of that property at different locations on the sample’s surface 18 in much the same way a soil map shows different soil properties at different locations on the Earth’s surface. Such a property will be referred to as a “mapped property.”

[0050] In some cases, the mapped surface properties are physical properties. In other cases, the surface properties are geometrical properties. An example of a geometrical property is the height of the surface 18. Examples of physical properties include the surface’s adhesion, its stiffness, and energy losses associated with contacting the surface 18.

[0051] A multi-channel atomic force microscope 10 has the ability to map different properties at the same time. Each mapped property corresponds to a different “channel” of the microscope 10. An image 78 can therefore be regarded as a multidimensional image array where the channel index, k, is an integer in the interval [1, K], where K is the number of channels.

[0052] When used in a sub-resonance tapping mode, a multi-channel atomic force microscope 10 can map the following surface properties: height, adhesion, deformation, stiffness, viscoelastic losses, feedback error. This results in six channels, each of which corresponds to one of six mapped surface properties. When used in Ringing mode, the atomic force microscope 10 can map, as an example, one or more of the following additional properties in addition to the previous six properties: restored adhesion, adhesion height, disconnection height, pull-off neck height, disconnection distance, disconnection energy loss, dynamic creep phase shift, and zero-force height. This results in a total of fourteen channels in this example, each of which corresponds to one of fourteen mapped properties.

[0053] The scanner 12 defines discrete pixels on the reference plane. At each pixel, the microscope’s probe 16 makes a measurement. For convenience, the pixels on the plane can be defined by Cartesian coordinates (x,, y7). The value of the kthchannel measured at that pixel is Zy®. With this in mind, an image array that represents a map or image of the kthchannel can be formally represented as:

[0054] ^={xi, yj, Zik)} (1) where “z” and “j” are integers in the intervals [1, Ni] and [1, Nj], respectively and where Ni and Nj are the numbers of pixels available for recording an image 78 in the x and y directions respectively. The values of Ni and Nj can be different. However, the methods described herein do not depend significantly on such a difference. Hence, for purposes of discussion, Ni=Nj=N.

[0055] The number of elements in a sample’s image array would be the product of the number of channels and the number of pixels. For a relatively homogeneous surface 18, it is only necessary to scan one region of the surface 18. However, for a more heterogenous surface 18, it is preferable to scan more than one region on the surface 18. By way of analogy, if one wishes to inspect the surface of the water in a harbor, it is most likely only necessary to scan one region because other regions would likely be similar anyway. On the other hand, if one wishes to inspect the surface of the city that the harbor serves, it would be prudent to scan multiple regions. After all, some regions would be paved, others would be grass, and still others might be water.

[0056] With this in mind, the array acquires another index to identify the particular region that is being scanned. This increases the array’s dimensionality. A formal representation of the image array is thus: where the scanned-region index 5 is an integer in the interval [1, 5] that identifies a particular scanned region within a sample. Note that this causes the number of elements in the image array for a particular sample to grow by a factor equal to the number of scanned regions. Preferably, the number of such scanned regions is large enough to represent the sample as a whole. One way to converge on an appropriate number of scanned regions is to compare the distribution of deviations between two such scanned regions. If incrementing the number of scanned regions does not change this in a statistically significant way, then the number of scanned regions is likely to be adequate to represent the surface as a whole. Another way is to divide what is considered to be a reasonable testing time by the amount of time required to scan each scanned region and to use that quotient as the number of areas.

[0057] In some cases, it is useful to split each of the scanned regions into partitions. For the case in which there are P such partitions in each scanned region, the array can be defined as: where the partition-index p is an integer in the interval [1, P], In the case of a square scanned area, it is convenient to divide the square into four square partitions, thus setting P to be equal to four.

[0058] The ability to divide a scanned region into partitions provides a useful way to exclude image artifacts. This is particularly important for inspection of hard-to-prepare samples. This makes it possible to compare one partition against the others to identify which, if any, deviate significantly enough to be excluded. A partition can also be the area of known material, which can be used for training the classification algorithm.

[0059] On the other hand, the addition of a new index further increases the dimensionality of the array.

[0060] To identify a class to which a sample belongs based on the image arrays M acquired by the atomic force microscope 10, the machine-learning circuitry 68 relies in part on building a suitable database 70 that includes images of surfaces that are known a priori to belong to particular classes C®, i.e., training data 72. Each class corresponds to a particular material. A suitable formal representation for such a database 70 is: where k is a channel index that represents a property or channel, .v is a scanned- region index that identifies a particular scanned region, p is a partition index that represents a particular partition of the sthscanned region, n is a sample index that identifies a particular sample, and I is a class index that identifies a particular class from a set of L classes. The overall size of the array is thus the product of the number of classes, the number of samples, the number of scanned regions, the number of partitions per scanned region, and the number of channels.

[0061] The entire database Dn^l;k;s;p^ is usable for training a machine-learning circuitry 68 for recognizing a particular class, or material in the database. A suitable way to estimate the effectiveness of such training is that of splitting the database into training and testing subsets. Examples of suitable ratios between training and testing subsets include 8:2, 7:3, and 6:4, with the training subset in each case being the larger. A confusion matrix is then calculated for the testing subset.

[0062] There is no restriction on the particular machine learning methods used to train the machine-learning circuitry 68. Examples of suitable algorithms include, but are not limited to, a neural network, including regular, deep learning, and convolution neural networks, various regression algorithms, decision trees, and support vector machines.

[0063] For spectroscopy, the machine-learning circuitry 68 developed as specified above is applicable to images 78 of surfaces that comprise an unknown mix of the materials that were used to create the database 70 and to train the machine-learning circuitry 68.

[0064] The machine-learning circuitry 68 is applied to each pixel of the image 78. As noted above, the image 78 is one that contains the mix of the materials used to create the database 70 and to train the machine learning algorithm 76. In the case of regression methods, the machine-learning circuitry 68 algorithm outputs a value indicative of a probability that the pixel corresponds to a particular one of the materials. In the case of classification methods, the machine-learning circuitry 68 provides the class or type of the material for the pixel under analysis.

[0065] It should be noted that a regression algorithm can be converted into classification algorithm by introducing an additional algorithm. An example of a suitable additional algorithm to carry out such a conversion is one that classifies a class assignment as being definite if its probability of being correct is above some threshold value. Alternatively, one can choose the assignment of the class for the maximum probability among probabilities for all possible classes.

[0066] In the following examples, the atomic force microscope 10 was operated using both sub-resonant tapping modes, such as PeakForce QMN as implemented by Bruker, Inc., and Ringing modes (implemented by NanoScience Solutions, LLC.). Both modes make it possible to record two or more channels simultaneously. In particular, both modes make it possible to simultaneously record a height channel and an adhesion channel. The use of Ringing mode, however, is significantly faster than sub-resonant tapping for collecting images. As noted above, these modes allow many channels to be recorded simultaneously.

[0067] In a first example, three polymer materials were imaged to create the database m aterja]wasimaged with just one 5x5 micron area using twelve different channels. These channels were: regular height, height at zero force, height and adhesion, height at the disconnection, peak force, adhesion, restored adhesion, viscoelastic elongation, neck height, disconnection length, energy losses during disconnection, and dynamic phase. Each image was recorded with 512x512 pixels. This information was used to train the machine-learning circuitry 68. For the purpose of mechano-spectroscopy, it is possible to apply the developed machine-learning circuitry 68 to a similarly imaged unknown mix of the materials used to create the database Dn^l;k;s;p

[0068] FIG. 4 shows images 88 for an unknown mix of three polymers. These images 88 were used to train the machine-learning circuitry 68. By passing the images 88 through the machine-learning circuitry 68, it is possible to identify which polymer is present at each pixel of the image 88. The result is presented in an identification result 90 formed by as superposition of the AFM imaged geometry of the sample with each polymer having been highlighted by a color that corresponds to that polymer.

[0069] FIG. 5 shows identification results 90 resulting from three different types of machine-learning circuitry 68: neural network results 92, decision tree results 94, and random forest results 96. As is readily apparent, the results 92, 94, 96 are substantially identical. Accordingly, it appears that the identification results are not sensitive to whether training occurs using a neural network or a bootstrapped decision tree method.

[0070] A second example trains machine-learning circuitry 68 using only those parts of an image for which it is possible to definitely identify a material. In this second example, the results of which follow, the imaging techniques are the same as those used in the first example and the size of the images is twenty square micrometers.

[0071] FIG. 6 shows an image 88 example of a mix of three polymers with rectangles showing particular areas 98 that are identifiable as being made of a particular polymer. Such identification is carried out by examining morphology or carrying out Raman microscopy or an equivalent thereof. It is the portions of the image 88 within these rectangles 98 that are used to create the database Dn(l;k;s;p^ for training the machinelearning circuitry 68. Passing the rest of the image through the machine-learning circuitry 68 results in an identification result 92 similar to that shown in FIG. 4. The illustrated example offers significantly higher resolution than known methods, such as confocal Raman microscopy.

[0072] FIG. 7 shows an example of the application of the machine-learning circuitry 68 as applied to a higher resolution area of a sample having first and second regions 100, 102. The first region 100 is dominated by silica particles; the second region 102 is dominated by epoxy resin.

[0073] FIG. 8 shows result of having the machine-learning circuitry 68 determine the silica probability 104, i.e., the probability of encountering silica particles, as one traverses a cross section along the roughly twenty-nanometer line segment 106 in FIG. 7. A transition region 108 between the first region 100 and second region 102 provides an indication of the lateral spectroscopic resolution of the atomic force microscope 10. In this case, the transition region 108 has a width of about 1.6 nanometers. This is two orders of magnitude greater than the resolution provided by conventional methods, such as confocal Raman spectroscopy.

[0074] In some cases, it is useful to immerse the sample 18 in a suitable fluid medium to increase contrast, and hence spectroscopic resolution, between different materials.

[0075] Examples of suitable media include liquid media, such as an aqueous medium, a medium that consists of an organic solvent, a hydrophobic medium, such as oil, and a mixture of an aqueous medium and one or more organic solvents that are miscible in that aqueous medium.

[0076] Other examples include gaseous media, such as mixtures of gases that include water vapor and vapors from various organic materials. In the case of gaseous media, it is possible to have different partial pressures of the gases, including water vapor. This leads to different humidities. FIG. 9 in particular shows the effect of three different humidity levels in the contrast between two surfactant materials that have been placed on a flat surface and exposed to air at different humidities.

[0077] To promote stability in measurement, it is useful to image a sample at a controlled temperature. This arises because mechanical properties of the sample depend on temperature to an extent such that even small variations have a noticeable effect on measurements.

[0078] Having described the invention and a preferred embodiment thereof, what is claimed as new and secured by letters patent is:

Claims

CLAIMS1. A method comprising using an atomic-force microscope to train machine-learning circuitry to classify a first pixel of an image of a test sample as corresponding to a first material and a second pixel of said image as corresponding to a second material and to do so based on mechanical properties of said first and second materials, wherein said test sample comprises materials selected from a finite group of materials that includes said first material and said second material.

2. The method of claim 1, further comprising obtaining training samples in which locations of said first and second materials in said training samples are known, obtaining multi-dimensional images of said training samples, and storing said images in a database, wherein training said machine-learning circuitry comprises using said multidimensional images for training said machine-learning circuitry.

3. The method of claim 1, further comprising obtaining training samples that comprise said first and second materials, wherein there exist regions in said training samples for which locations of said first and second materials are unknown, obtaining multi-dimensional images of said training samples, and storing said images in a database, wherein training said machine-learning circuitry comprises using said multidimensional images for training said machine-learning circuitry.

4. The method of claim 1, further comprising selecting said first and second materials to be polymers.

5. The method of claim 1, further comprising obtaining said image by operating said microscope in sub-resonance tapping mode on said test sample.

6. The method of claim 1, further comprising obtaining said image by operating said microscope in Ringing mode on said test sample to carry out nanoscale -resolution scanning of the surfaces of said test sample.

7. The method of claim 1, further comprising obtaining said image, wherein obtaining said image comprises imaging said test sample at a controlled temperature.

8. The method of claim 1, further comprising obtaining said image, wherein obtaining said image comprises imaging said sample in a fluid.

9. The method of claim 1, further comprising obtaining said image, wherein obtaining said image comprises imaging said test sample in a gas.

10. The method of claim 1, further comprising obtaining said image, wherein obtaining said image comprises imaging said test sample under conditions of controlled humidity.

11. The method of claim 1 , further comprising obtaining said image, wherein obtaining said image comprises imaging said test sample under conditions of controlled partial pressure of an organic vapor.

12. The method of claim 1, further comprising obtaining said image, wherein obtaining said image comprises imaging said test sample in an aqueous medium.

13. The method of claim 1, further comprising obtaining said image, wherein obtaining said image comprises imaging said test sample in a liquid that comprises a mixture of an aqueous medium and at least one organic solvent that is miscible in said aqueous medium.

Images