WO2018098588A1 - Systèmes informatiques et procédés d'identification de matériaux non élémentaires sur la base de propriétés atomistiques - Google Patents

Systèmes informatiques et procédés d'identification de matériaux non élémentaires sur la base de propriétés atomistiques Download PDF

Info

Publication number
WO2018098588A1
WO2018098588A1 PCT/CA2017/051449 CA2017051449W WO2018098588A1 WO 2018098588 A1 WO2018098588 A1 WO 2018098588A1 CA 2017051449 W CA2017051449 W CA 2017051449W WO 2018098588 A1 WO2018098588 A1 WO 2018098588A1
Authority
WO
WIPO (PCT)
Prior art keywords
material property
data store
materials
quantitative
elemental
Prior art date
Application number
PCT/CA2017/051449
Other languages
English (en)
Inventor
Pawel Michal PISARSKI
Scott Richard Holloway
Original Assignee
Lumiant Corporation
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Lumiant Corporation filed Critical Lumiant Corporation
Publication of WO2018098588A1 publication Critical patent/WO2018098588A1/fr

Links

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C60/00—Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
    • G16C20/70—Machine learning, data mining or chemometrics
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
    • G16C20/90—Programming languages; Computing architectures; Database systems; Data warehousing

Definitions

  • TITLE COMPUTER SYSTEMS FOR AND METHODS OF IDENTIFYING NON-ELEMENTAL MATERIALS BASED ON ATOMISTIC PROPERTIES
  • the present disclosure relates to computer systems for and methods of identifying non-elemental materials. More in particular, the present disclosure relates to computer systems and methods involving the use of atomistic models to identify non-elemental materials having certain desired properties.
  • the present disclosure relates to a computer system for the identification of non-elemental materials having defined material property parameters.
  • the present disclosure provides, in at least one implementation, a computer system for the identification of non-elemental materials having defined material properties, the computer system comprising: a) a first data store assembled in electronic readable format and operable to receive:
  • a processing element coupled to the first, second, third, and fourth data stores, and configured to:
  • the processing element can be configured to predict the quantitative values of the material property parameters by:
  • the processing element can be configured to perform step (iii) using a machine learning algorithm.
  • the machine learning algorithm can form a regression model and trains the machine learning algorithm to perform step (iii) to thereby obtain a trained model, and applying the trained model to the second set of fingerprints to perform step (iv).
  • the material descriptors can include at least one of chemical formulas and crystalline forms.
  • the computer system can include a user output module coupled to the processing element.
  • the output module can be operable to receive a material property parameter associated with non- elemental materials for which a quantitative value of a material property parameter is absent in the second data store.
  • the processing element can automatically initiate prediction of the quantitative values upon receipt of a material property parameter associated with non-elemental materials for which a quantitative value of a material property parameter is absent in the second data store.
  • the computer system includes a user input module coupled to the processing element.
  • prediction of the quantitative values can be initiated by a user command to the input module.
  • the user input module can be operable to receive input for at least one of the first, second or third data stores.
  • the computer system can include user input and user output modules, coupled to the processing element, wherein the user input module is operable to receive user input in the form of a quantitatively defined value or range of values for a material property parameter, and the user output is operable to provide one or more non- elemental materials exhibiting the quantitatively defined value or range of values for the material property parameter in the form of user output.
  • the user output module can separately define the quantitative values of the material property parameters received by the computer system and those predicted by the computer system.
  • the present disclosure provides, in at least one implementation, a method of identifying non-elemental materials having defined material property parameters, the method comprising:
  • the method can include predicting the quantitative values of the material property parameters by:
  • the method can include performing step (iii) using a machine learning algorithm.
  • the method can include forming a regression model and training the machine learning algorithm to perform step (iii) to thereby obtain a trained model, and applying the trained model to the second set of fingerprints to perform step (iv).
  • the method can include receiving the material descriptors in the form of at least one of chemical formulas or crystalline forms.
  • the method can include receiving a material property parameter associated with non-elemental materials for which a quantitative value of a material property parameter is absent in the second data store by an output module.
  • the method can include automatically initiating prediction of the quantitative values upon receipt of a material property parameter associated with non-elemental materials for which a quantitative value of a material property parameter is absent in the second data store.
  • the method can include initiating prediction of the quantitative values by a user command.
  • the method can include receiving input for at least one of the first, second and third data stores.
  • the method can include receiving user input in the form of a quantitatively defined value or range of values for a material property parameter, and providing one or more non- elemental materials exhibiting the quantitatively defined value or range of values for the material property parameter in the form of user output.
  • the method can include of claim separately defining the quantitative values of the material property parameters received and those predicted in the form of user output. [0034] In at least some implementations, the method can include empirically testing the predicted quantitative values of the material property parameters to obtain an empirically determined quantitative value of the material property parameter.
  • the method can include providing the empirically determined quantitative value of the material property parameter in the form of input to the second data store.
  • the present disclosure provides, in at least one implementation, a use of a computer system of the present disclosure to identify non-elemental materials exhibiting a defined quantitative value of a material property parameter.
  • the present disclosure provides, in at least one implementation, a method of identifying and using a non-elemental material having a defined quantitative material property value, the method comprising:
  • the present disclosure provides, in at least one implementation, a method of identifying and using a non-elemental material having a defined quantitative material property value, the method comprising:
  • the present disclosure provides, in at least one implementation, a computer readable medium comprising a plurality of instructions that, when executed by a processing element, cause the processing element to perform any of the methods of the present disclosure.
  • FIGS. 1A, 1 B, 1 C, 1 D, 1 E and 1 F show schematic block diagrams illustrating examples of computer systems
  • FIGS. 2A, 2B, 2C, 2D, 2E, 2F and 2G show schematic flow charts illustrating examples of methods conducted by a computer system
  • FIG. 3 shows a schematic flow chart illustrating an example of a method conducted by a computer system
  • FIG. 4 shows a schematic block diagram illustrating an example of a computer system
  • FIG. 5 shows a schematic flow chart illustrating an example of a method conducted by a computer system
  • FIG. 6 shows a schematic flow chart illustrating another example of a method conducted by a computer system
  • FIG. 7 shows a schematic flow chart of an example method involving the use of a computer system
  • FIG. 8 shows an example of output by a computer system
  • FIG. 9 shows another example of output by a computer system
  • FIG. 10 shows an example neural network
  • FIGS. 1 1A and 1 1 B show results of training and validation of a computer system using two different material property parameters: density (FIG. 1 1 A) and enthalpy of formation (FIG. 1 1 B);
  • FIG. 12 shows results of an example method for generating atomistic fingerprints for titanium aluminide
  • FIG. 13 shows in example of an input to a computer system.
  • computer system refers to any device with an electronic processing element capable of executing instructions, including but not limited to, any supercomputer, mainframe computer, server, personal computer, desktop computer, hand-held computer, lap-top computer, tablet computer, cell phone computer, or smart phone computer or any other suitable electronic device.
  • the term "coupled” as used herein can have several different meanings depending on the context in which the term is used.
  • the term coupling can have a mechanical or electrical connotation depending on the context in which it is used; i.e. whether describing a physical layout or transmission of data as the case may be.
  • the term coupling may indicate that two elements or devices can be directly physically or electrically connected to one another or connected to one another through one or more intermediate elements or devices via a physical or electrical element such as, but not limited to, a wire, a non-active circuit element (e.g., resistor), wireless connections and the like, for example.
  • a physical or electrical element such as, but not limited to, a wire, a non-active circuit element (e.g., resistor), wireless connections and the like, for example.
  • data store refers to any device or combination of devices capable of storing, accessing, and retrieving data, including, without limitation, any combination and number of data servers, databases, tables, files, lists, queues, directories, data storage devices, data storage media, and the like.
  • element and "chemical element”, as may be used interchangeably herein, refer to chemical elements as set forth in the Periodic Table of Elements, as released and revised from time to time by the International Union of Pure and Applied Chemistry (lUPAC).
  • material descriptor refers to any and all information that can be used to describe and thereby identify a non- elemental material.
  • output device refers to any device that is used to output information and includes, but is not limited to, one or more of a terminal, a desk top computer, a laptop, a tablet, a cellular phone, a smart phone, a printer (e.g., laser, inkjet, dot matrix), a plotter or other hard copy output device, speaker, headphones, electronic storage device, a radio or other communication device, that could communicate with another device, or any other computing unit.
  • printer e.g., laser, inkjet, dot matrix
  • a plotter or other hard copy output device e.g., speaker, headphones, electronic storage device, a radio or other communication device, that could communicate with another device, or any other computing unit.
  • Output devices may include a two dimensional display, such as a TV or a liquid crystal display (LCD), a light-emitting diode (LED) backlit display, a mobile telephone display, a three dimensional display capable of providing output data in a user viewable format.
  • a two dimensional display such as a TV or a liquid crystal display (LCD), a light-emitting diode (LED) backlit display, a mobile telephone display, a three dimensional display capable of providing output data in a user viewable format.
  • LCD liquid crystal display
  • LED light-emitting diode
  • the term "input device”, as used herein, refers to any user operable device that is used to input information and includes but is not limited to, one or more of a terminal, a desk top computer, a laptop, a tablet, a cellular phone, a smart phone, a touch screen, a keyboard, a mouse, a mouse pad, a tracker ball, a joystick, a microphone, a voice recognition system, a light pen, a camera, a data entry device, such as a bar code reader or a magnetic ink character recognition device, sensor or any other computing unit capable of receiving input data.
  • a terminal such as a desk top computer, a laptop, a tablet, a cellular phone, a smart phone, a touch screen, a keyboard, a mouse, a mouse pad, a tracker ball, a joystick, a microphone, a voice recognition system, a light pen, a camera, a data entry device, such as a bar code reader or a magnetic ink character recognition device, sensor
  • Input devices may include a two dimensional display, such as a TV or a liquid crystal display (LCD), a light-emitting diode (LED) backlit display, a mobile telephone display, a three dimensional display capable of receiving input from a user, e.g., by touch screen.
  • a two dimensional display such as a TV or a liquid crystal display (LCD), a light-emitting diode (LED) backlit display, a mobile telephone display, a three dimensional display capable of receiving input from a user, e.g., by touch screen.
  • the user in accordance herewith may be any user including any scientist, technician, student or instructor.
  • a computer system can be configured that provides for the rapid identification of non-elemental materials based on one or more user-specified material properties.
  • the memory element of the computer system can be configured to receive and store information relating to non-elemental materials, including material property parameters associated with these materials.
  • the quantitative values of these material property parameters may be either known or unknown.
  • the computer system can predict a previously unknown quantitative value of one or more material property parameters of a non-elemental material.
  • the computer system can provide non-elemental materials exhibiting material properties defined by user input, even if the quantitative values of such material properties were previously unknown.
  • a computer system for the identification of non-elemental materials having defined material properties includes a memory element, a processing element and a user input module.
  • the memory element includes four data stores.
  • the first, second and third data store are each capable of receiving and storing certain input information relating to non-elemental materials.
  • the fourth data store is capable of receiving previously unknown quantitative values of material property parameters associated with the non-elemental materials. These previously unknown values can be predicted by the computer system using the information present in the first, second and third data store. The predicted quantitative values of the material property parameters are subsequently available to users wishing to employ the computer system to identify materials with certain desired material properties.
  • the first, second and third data stores can receive data via the user input module.
  • the processing element permits and controls communication between the data stores and the input module.
  • a computer system is configured to predict the quantitative values of material property parameters of non-elemental materials.
  • the computer system is configured to predict quantitative values of material property parameters in instances in which these values are absent from the data stores of the computer system.
  • a computer system initially receives a series of material descriptors. Each descriptor uniquely identifies certain non-elemental materials.
  • the computer system also receives one or more quantitative values of material property parameters associated with each of the materials. However, for several of the materials no quantitative value is known for a particular material property parameter. Thus, these quantitative values are not received by the computer system and identified as absent therefrom. The method then proceeds to predict the quantitative values.
  • the method first identifies two sets of atomistic fingerprints.
  • the first set of atomistic fingerprints corresponds with materials for which a quantitative value for the material property parameter is present in the data stores of the computer system.
  • the first set of atomistic fingerprints is then used to establish a correlation between certain distinct fingerprint features and specific quantitative values of the material property parameter present in the computer system. This can be achieved, for example, by employing a learning algorithm.
  • the second set of fingerprints is associated with the materials for which the quantitative value of the material property parameter is absent.
  • the quantitative values of the non-elemental material property parameter, which were initially unknown, are then predicted by the computer system by comparing the second set of fingerprints with the first set of fingerprints, with respect to the distinct fingerprint features.
  • the predicted quantitative value is then received by a data store.
  • An output module can also be included in the computer systems of the present disclosure.
  • a user can through an input module provide one or more desired quantitative values of a material property parameter.
  • An output module can be used to display to a user materials that exhibit the desired quantitative values.
  • the output module can indicate which materials are known to exhibit the desired quantitative values, and which materials have been predicted by the computer system to exhibit the desired quantitative values.
  • FIGS. 1A to 1 F details of various selected implementations of computer systems are provided (FIGS. 1A to 1 F), each in conjunction with an accompanying selected implementation of a method to be performed (FIGS. 2A to 2G).
  • FIG. 1A the present disclosure provides an example of a computer system 100 including user input module 125, processing element 1 15, memory element 120, and user output module 130.
  • Processing element 1 15 and memory element 120 can be housed together, for example, in terminal 1 10.
  • User input module 125 and user output module 130 are coupled to processing element 1 15 in order to permit communication between processing element 1 15 and user input module 125 and user output module 130.
  • Memory element 120 includes a tangible storage medium operable to store computer readable data and instructions.
  • the tangible storage medium may be any storage medium, such as, but not limited to, an electrical magnetic or optical storage medium, and the storage medium may be implemented using any suitable technique or methodology known to those skilled in the art, such as, but not limited to random access memory (RAM), disk storage, flash memory, solid state memory, CD-ROM, and so on.
  • Memory element 120 further includes first, second, third and fourth data stores 121 , 122, 123, 124, with each data store having data assembled in electronic readable format.
  • the term "assembled in electronic readable format" as used herein means any file, application, module or other data that is useable by processing element 1 15.
  • an application or module includes code executable by processing element 1 15 that may be run to carry out one or more functions associated with user input module 125 and the methods to implement the methods of the disclosure.
  • code executable by the processing element includes any computer- readable media or commands that may be interpreted by processing element 1 15, such as HTML or XML files, C, C++, SQL or other suitable computer files, that are rendered into user-viewable applications by an application executed by processing element 1 15.
  • Processing element 1 15, which is at least one processor or other suitable hardware, is configured to execute and perform computer readable instructions, which may be accessed from a disc, a memory element, or other device capable of storing instructions thereon, and can be arranged in one or more operable modules. These computer readable instructions cause processing element 1 15, upon user operation of the user input module 125, to acquire input data including, quantitatively defined values of material property parameters, and access and process, as herein further described, data from memory element 120 based on the input, and thereafter to transmit processed data to user output module 130 to permit user output module 130 to display output in the form of one more identified non-elemental materials.
  • Method 200 starts with the loading of an interface capable of receiving commands or information in the form of user inputs provided through a user input module (step 210), values or value ranges of material property parameters. After the method 200 is started, it waits for a user input command in the form of a selection of quantitatively defined values or value ranges of material property parameters (step 212).
  • a user may provide as quantitatively defined input material property parameters: a melting point of at least 600 °C (IP 1 ); a density of 1 mg/ml (IP 2); a melting point between 300 °C and 600 °C, and a density of up to 0.5 mg/ml (IP 3).
  • IP 1 a melting point of at least 600 °C
  • IP 2 a density of 1 mg/ml
  • IP 3 a density of up to 0.5 mg/ml
  • at least one input material property parameter is provided.
  • 2, 3, 4, 5, 6, 7, 8, 9, 10 or more input material property parameters or ranges of input material property values are provided.
  • the method 200 then screens (step 214) the entries of materials present in the data stores of the computer system for the presence of entries of non-elemental materials having material property parameters of which the quantitative values correspond with those provided by the user (step 210) and displays the materials to a user output module (step 216).
  • method 200 may provide in the form of output materials M1 , M2 and M3 having a melting points of 700 °C, 900 °C, and 1 , 100 °C, respectively; in response to IP 2, method 200 may provide material M4 and M5, each having a density of 1 mg/ml; and in response to IP 3, the method may provide a material M6 having a melting point between of 350 °C and a density of 0.4 mg/ml.
  • data stores 121 , 122, 123 of computer system 100 are initially assembled, as is described hereinafter, with respect to the example implementations of computer systems 101 , 102, 103, 104 as illustrated in FIGS. 1 B to 1 E, and example methods 201 , 202, 203, 204 as illustrated in FIGS. 2B to 2E.
  • the computer system In order for the computer system to predict quantitative values of material property parameters and include these predicted values and associated materials in data store 124, at least one quantitative value of a material property parameter associated with a non- elemental material must be absent in the second data store. Upon identification of such absence, a user may initiate the prediction, or the system may automatically initiate prediction of these values. This is further illustrated in FIG.
  • a first data store 121 is configured to receive material descriptors via user input module 125. Each material descriptor is uniquely associated with a non-elemental material and the material descriptors are received in such a form that a unique identification or separate entry of each of the materials in the data store can be achieved.
  • Processing element 1 15 is configured to comprise input module 140 operable to communicate with first data store 121 , and to provide the material descriptors to first data store 121 .
  • First data store 121 is configured to receive and store the material descriptors.
  • Method 201 starts with the loading of an interface capable of receiving commands or information in the form of user inputs provided through a user input module.
  • the user input consists of material descriptors associated with a plurality of materials, and, as noted, is provided in such a form that the unique identification or separate entry for each of the materials can be achieved (step 220).
  • the material descriptors can include a chemical formula for a non-elemental material, for example, a chemical formula or representation based on the Periodic Table of Elements.
  • the material descriptors can include information relating to crystalline phases of a non-elemental material, which can, for example, be obtained from the Inorganic Structure Crystal Database (ICSD) (http://www.cds.dl.ac.uk), the crystallography open database (COD) (http://www.crystallograaphy.net), Pauling File (http://www.paulingfile.com) or Pearson's Crystal Data (http://www.crystalimpact.com/pcd/).
  • ICSD Inorganic Structure Crystal Database
  • COD crystallography open database
  • Pauling File http://www.paulingfile.com
  • Pearson's Crystal Data http://www.crystalimpact.com/pcd/
  • the material identifying information can include separate entries for two or more crystalline phases, when a material, although chemically identified by the same formula, is known to exist in multiple crystalline phases. It is noted that the quantitative values for various material property parameters of each of these crystalline phases can vary.
  • method 201 After method 201 is started, it waits for a user input command in the form of material descriptors (step 222).
  • a user can provide input in the form of material descriptors relating to non-elemental materials titanium dioxide and aluminum oxide by providing chemical formulas for each of the materials, i.e. Ti0 2 and AI2O3, respectively.
  • the method can then create unique and separate entries in for each of the non-elemental materials.
  • crystalline structures of titanium dioxide are known.
  • a user can provide a material descriptor by including each of the crystalline forms known as rutile, anatase, and brookite. The method can then create separate unique and separate entries for each of the rutile form, anatase form and brookite form of titanium dioxide.
  • Method 201 then provides the material descriptors to a first data store (step 224).
  • a data store comprising material descriptors for a plurality of materials can be assembled.
  • a data store can include, for example, material descriptors associated with 1 ,000, 10,000, 100,000, 1 ,000,000, 5,000,000 or even more non-elemental materials.
  • second data store 122 is configured to receive one or more material property parameters associated with the non-elemental materials in the first data store.
  • Processing element 1 15 is configured to comprise input module 150 operable to communicate with second data store 122 and provide the material property parameters to second data store 122.
  • Second data store 122 is configured to receive and store the material property parameters associated with the non-elemental materials.
  • Method 202 starts with the loading of an interface capable of receiving commands or information in the form of user inputs provided through a user input module.
  • the user input consists of one or more material property parameters associated with the non-elemental materials in the first data store (step 230).
  • the method 202 waits for a user input command in the form of one or more material property parameters (step 232).
  • a user can provide a known melting point of 600 °C associated with a material M7, or a user can provide a known melting point of 300 °C, and a known density of 0.7 mg/ml associated with a material M8.
  • a known melting point of 600 °C associated with a material M7
  • a known melting point of 300 °C can be provided.
  • a known density of 0.7 mg/ml associated with a material M8.
  • the method 202 then provides the material property parameters to the second data store (step 234).
  • material property parameters include: acoustical properties, for example, acoustic absorption; chemical properties, for example, corrosion resistance; electrical properties, for example, dielectric constant; magnetic properties, for example, Curie temperature; manufacturing properties, for example, castability; mechanical properties, for example, brittleness, tensile strength, Young's modulus, hardness, viscosity, fracture toughness and elasticity; optical properties, for example, luminosity; radiological properties, for example, neutron cross section; thermal properties, for example, boiling point, melting point, eutectic point, and vapour pressure.
  • acoustical properties for example, acoustic absorption
  • chemical properties for example, corrosion resistance
  • electrical properties for example, dielectric constant
  • magnetic properties for example, Curie temperature
  • manufacturing properties for example, castability
  • mechanical properties for example, brittleness, tensile strength, Young's modulus, hardness, viscosity, fracture toughness and elasticity
  • optical properties for example, luminosity
  • Quantitative values for each property parameter for a non- elemental material can be expressed in various units known to those of skill in the art.
  • melting temperature can be expressed in Centigrade (°C) or Degrees Kelvin (K)
  • density can be expressed in milligram per cubic centimeter (mg/cm 3 ) or kilogram per cubic meter (kg/m 3 ), and so forth.
  • SI Units International Systems of Units
  • the processing element can include operational modules for conversion of units.
  • Known quantitative values of material property parameters are generally values that have been empirically determined using suitable experimental methods or techniques for the determination of these quantitative values. For example, the melting temperature of a solid material can be experimentally determined by heating the material, and measuring the temperature at which the material melts.
  • Known quantitative values of material property parameters can also include quantitative values that have been indirectly determined, for example, theoretical quantitative values deduced from empirical observations. Methods that can be used to obtain theoretical quantitative values of material property parameters include, for example, molecular dynamics based methods or methods based on density functional theory.
  • any and all material property parameters applicable to a non-elemental material are intended to be included herein, and, it should be clear that the present disclosure is not intended to be limited with respect to the actual material property parameters that can be used in accordance herewith.
  • a second data store can be assembled comprising at least one quantitative value of a material property for each material in the first data store. It will be appreciated to those of skill in the art that a data store including more than one quantitative value associated with a plurality of materials can be assembled.
  • the second data store can comprise 5, 10, 15, 20, 25, 50, 100, or more quantitatively defined material properties for one or more materials in the second data store.
  • the first data store includes a plurality of material descriptors, each material descriptor uniquely identifying a non-elemental material.
  • the second data store includes for each of these materials at least one quantitatively defined material property parameter.
  • data store 123 is configured to receive atomistic fingerprints via user input module 125.
  • the atomistic fingerprints in third data store 123 generally corresponds with the material property parameters and associated non-elemental materials in first data store 121 , so that memory element 120 includes for each non-elemental material a material descriptor and at least one material property parameter in first or second data store 121 , 122 and an atomistic fingerprint in third data store 123.
  • the stored atomistic fingerprints can be stored in any kind of format, including, text files, HTML files, images, movies and the like.
  • the atomistic fingerprints can be assembled using any suitable methodology, including, for example, a methodology based on crystal structure, a methodology based on electronic band structure, or a methodology based on vibrations of the crystal lattice.
  • Processing element 1 15 is configured to comprise input module 160 operable to provide the atomistic fingerprints to third data store 123, and third data store 123 is configured to receive the atomistic fingerprints.
  • atomistic fingerprints can be entered into a third data store in an assembled fashion.
  • Assembled atomistic fingerprints of non-elemental materials are available for example in: Acta Cryst. 66, 507 (2010); J. Chem. Phys. 130, 104504 (2009); Chem. Phys. Lett. 395, 210 (2004); Phys. Rev. Lett. 108, 058301 (2012); or J. Chem. Theory Comput. 9, 3404 (2013).
  • Method 203 starts with the loading of an interface capable of receiving commands or information in the form of user inputs provided through a user input module.
  • the user input consists of one or more atomistic fingerprints associated with the non-elemental materials (step 240).
  • the method 203 waits for a user input command in the form of one or more atomistic fingerprints associated with the non-elemental materials (step 242).
  • atomistic fingerprints can be provided associated with material M7, M8, and M9. It is noted that, in some implementations, atomistic fingerprints can be provided to the third data store in this manner, for which certain quantitative material property values are absent in the second data store.
  • the method 203 then provides the atomistic fingerprints to a third data store (step 244).
  • the assembly of atomistic fingerprints is performed by processing element 1 15 of the computer system of the present disclosure.
  • the computer system can include a fifth data store 126 including atomistic fingerprint assembly data, or operable to receive fingerprint assembly data via input module 125.
  • Processing element 1 15 includes atomistic fingerprint processing module 170 to assemble atomistic fingerprints using the fingerprint assembly data. Once assembled, the atomistic fingerprints can be provided to third data store 123 via fingerprint processing module 170. Third data store 123 can be configured to receive the assembled atomistic fingerprints.
  • Atomistic fingerprints can be assembled in the form of numerical matrices on the basis of, for example, crystal structures, electronic band structures, vibrations of the atomic lattice, or a combination thereof.
  • An example assembly of an atomistic fingerprint in the form of a Coulomb matrix is further described in Example 1 .
  • Atomistic fingerprints can also be assembled in the form or vectors, including one-dimensional vectors. Further examples of methods for the assembly of atomistic fingerprints, including in the form of matrices or vectors, are described in: Acta Cryst. 66, 507 (2010); J. Chem. Phys. 130, 104504 (2009); Chem. Phys. Lett. 395, 210 (2004); Phys. Rev. Lett. 108, 058301 (2012); or J. Chem. Theory Comput. 9, 3404 (2013).
  • one or more atomistic fingerprints can be present in the third data store for a given non-elemental material.
  • Method 204 starts with the loading of an interface capable of receiving commands or information in the form of user inputs provided through a user input module.
  • the user input consists of one or more atomistic fingerprint assembly data, which can be used to assemble atomistic fingerprints associated with the first and second set of materials (step 250).
  • method 204 waits for a user input command in the form of one or more atomistic fingerprint assembly data which can be used to assemble atomistic fingerprints associated with the first and second set of non-elemental materials (step 252).
  • the data is provided to fifth data store (step 254).
  • the atomistic fingerprint assembly data is then used to assemble atomistic fingerprints (step 256).
  • atomistic fingerprint assembly data can be provided to assemble atomistic fingerprints associated with material M7, M8, and M9. It is noted that in some implementations atomistic fingerprint assembly data can be provided to the fifth data store in this manner, for which no quantitative values for material property parameters are present in the second data store.
  • the method 204 then provides the atomistic fingerprints to the third data store (step 258).
  • the first data store comprises descriptors to uniquely identify a plurality of non-elemental materials.
  • the second data store comprises quantitatively defined material property parameters associated with each of the non- elemental materials.
  • the third data store includes atomistic fingerprints for each of the non-elemental materials.
  • the quantitative values of some material properties associated with the non-elemental materials can be absent from the second data store.
  • first data store 121 assembled in electronically readable format, includes one or more material descriptors uniquely associated with a set of non-elemental materials.
  • Second data store 122 assembled in electronically readable format, includes one or more material property parameters associated with the materials, wherein the value of at least one material property parameter is non-quantitatively defined for at least one of the materials.
  • Third data store 123 is further assembled in electronically readable format to include a first set of atomistic fingerprints associated with each of the materials of non-elemental materials.
  • Each of the data stores can be assembled employing user input module 125 and input module 180.
  • Processing element 1 15 is coupled to first, second, third and fourth data stores 121 , 122, 123, 124. Processing element 1 15 is further configured to access first and second data store 121 , 122, and identify material property parameters for which quantitative values are absent in the second data store. Processing element 1 15 is further also configured to access third data store 123 and identify a first set of fingerprints associated with non-elemental materials for which the quantitative values are present in the data store and second set of atomistic fingerprints, associated with non-elemental materials for which the quantitative values of the selected material property parameters are absent from the data store, respectively. Processing element 1 15 is further configured to, upon identification of material property parameters for which quantitative values are absent in the second data store, predict the quantitative value of the absent material property parameters.
  • processing element 1 15 is further configured to identify in the first set of atomistic fingerprints one or more distinct fingerprint features that correlate with the quantitative values of the selected material property parameters. [001 15] In some implementations, processing element 1 15 is further configured to compare the second set of atomistic fingerprints with the first set of atomistic fingerprints with respect to the distinct fingerprint features to thereby predict the quantitative values of the selected material property parameters.
  • Processing element 1 15 is further configured to provide the predicted quantitative value of the material property parameters to fourth data store 124, and fourth data store 124 is configured to receive the predicted quantitative value of the material property parameters.
  • the present disclosure provides an example method 205.
  • the method 205 starts with the loading of an interface capable of receiving commands or information in the form of user inputs provided through a user input module.
  • the user input consists of one or more material property parameters associated with non-elemental materials, wherein the quantitative value of at least one material property parameter is absent from the second data store (step 260).
  • method 205 waits for a user input command in the form of non-elemental material descriptors and associated material property parameters, wherein for a material, the quantitative value of at least one material property parameter is absent from the second data store (step 262).
  • the prediction of the absent quantitative value can be initiated (step 363).
  • the method includes receiving the identified absence of a quantitative value and providing the identified absence of a quantitative value to an output module.
  • a user may be notified thereof.
  • the output module may receive a material M10 for which the boiling point is unknown.
  • a user accessing the output module may receive notice on the output module that the computer system has identified that no boiling point is known for material M10.
  • the method includes initiating the prediction of the quantitative value via an input module in the form of a user command to initiate the prediction.
  • a user may opt to initiate the prediction by providing a user command initiating the prediction (step 363).
  • FIG. 2G shown therein is an example method 206.
  • the method starts in a manner similar to method 205 by identifying by identifying the absence of a quantitative value for a material property parameter (step 262). Upon such identification the method provides output to an output module there (step 381 ). The method then waits for an input command (step 382) in the form of an instruction to the method to initiate the prediction (step 363). The method thereafter continues in a manner similar to method 205.
  • the method includes an automatic initiation of the prediction of the absent quantitative value upon identification of the absence of the quantitative value of a material property from the second data store.
  • step 363 can automatically follow step 262.
  • Method 206 then identifies a set of materials for which the quantitative values of the material property has been defined (step 364) and creates a first set of non-elemental materials (step 366). Method 206 then identifies a first set atomistic fingerprints associated with the elemental materials in the first set (step 368).
  • the first set of atomistic fingerprints then is used to correlate fingerprint features with specific quantitatively defined values of material properties of materials in the first fingerprint set (step 370).
  • this can be achieved by using a machine learning algorithm.
  • this can be achieved by, for example, a support vector regression based learning algorithm, a boosted regression tree based learning algorithm, or a neural network based learning algorithm.
  • the machine learning algorithm includes an algorithm capable of forming a regression model, and training the machine learning algorithm.
  • a trained model capable of correlating a finger print feature associated with a non-elemental material to a specific quantitative value of material property parameter is obtained.
  • a certain distinct fingerprint feature present in fingerprints in the first material set may become correlated with a particular quantitative value of a melting temperature of a non-elemental material, or a particular quantitative value for the density of a non-elemental material.
  • FIG. 3 shown therein is an example method 300 for the performance of a learning algorithm to train a regression model.
  • the method starts (step 310) by creating pairs 1 -1 connections or references between fingerprints and quantitative material property parameter values (step 315).
  • a set of pairs of fingerprints and quantitative material property parameter values is then assigned for training purposes, and a set of pairs of fingerprints and quantitative material property parameter values is assigned for validation purposes.
  • pairs of fingerprints and quantitative material property values included in the training set are not included in the validation set and vice versa.
  • the set of training fingerprints and values is then used for machine learning (step 320).
  • the validation set of fingerprints and quantitative material property parameter values is applied to the model to test and validate the model (step 325).
  • the method will end (step 340). It is noted that the degree of accuracy can vary and be controlled using a correlation coefficient, for example a Pearson correlation, and relative error of predictions.
  • the training set can be expanded (step 345). These instances can, for example, be encountered when the set of materials in the database corresponding with the user input parameters is limited, and no meaningful correlation between fingerprint features and quantitative values of material property parameters can be established. Generally during validation, the model is then unable to correlate fingerprint features with quantitative values of material property parameters in the validation set with a sufficient degree of accuracy.
  • method 205, 206 can infer quantitative values of material property parameters which are unknown (FIG. 3F and 3G, step 378).
  • the computer system upon completion of method 300, the computer system is configured such that one or more quantitative values of material property parameters which are absent from the computer databases can be predicted. These values thereafter can be stored and used upon query of the computer system by a user to provide a material with desired quantitative values material property parameters. Further examples and details of learning algorithms are provided in Example 1 .
  • Atomistic fingerprint features that may be used to correlate to quantitative values of material property parameters may be any distinct characteristic in a fingerprint, including, for example, any arrangement or pattern of numerical values in a fingerprint assembled in the form of a numerical matrix, which correlates with a quantitative value of a material property parameter.
  • method 205 identifies the non- elemental materials for which the value of a material property parameter are absent the second data store (step 372) to create a second set of non- elemental materials (step 374).
  • a second set of atomistic fingerprints associated with the second set of non-elemental materials is then identified in the third data store (step 376).
  • the second set of atomistic fingerprints is then applied to the trained model, by presenting and applying each atomistic fingerprint in the second set of fingerprints to the trained model. Similarities in fingerprint features between an atomistic fingerprint included in the second set of fingerprints and atomistic fingerprints in the first set of fingerprints are used to predict the unknown quantitative value of a material property parameter of a fingerprint included in the second set of finger prints.
  • a set of five non-elemental materials can be received by the data stores where for non-elemental materials M12-M16 the melting point is unknown.
  • the first data store includes 1 ,000 non- elemental materials (M17-M1 ,016) for which the melting point is known.
  • the third data store includes the atomistic fingerprints of materials M17-M1 ,016, as well as the atomistic fingerprints of M12-M16.
  • a machine learning algorithm is formed and trained to correlate melting temperatures to fingerprint features of materials M17-M 1 ,016. The method then proceeds to predict the melting temperature of the non-elemental materials M12-M16 by applying the trained regression model to fingerprints of M12-M 16.
  • the method may provide, for example, a predicted melting temperature of M12 of 1 ,000 °C, M13 of 200 °C, M 14 of 350 °C, M 15 of 390 °C and M 16 of 1 ,200 °C.
  • the predicted quantitative value of a material parameter is provided to the fourth data store, and received by the fourth data store (step 379). In this manner, the amount of data present in the computer system can be increased each time the method exemplified by method 205 is performed. It will be clear to those of skill in the art that the accuracy with which the computer system of the present disclosure can predict quantitative values can increase as the quantity of material property parameters and the associated materials in the data stores increases.
  • user output is provided to a user output module in the form of materials that exhibit certain user specified material characteristics.
  • the user output is provided in a manner in which the user can identify whether the material characteristics are known, or predicted by the computer system.
  • the predicted quantitative values of a material property parameter can be tested experimentally.
  • a material M1017 is presented as having a predicted melting temperature of 300 °C, a quantity of the material may be obtained and the melting temperature may be determined experimentally.
  • a predicted quantitative value for a material property parameter such value may be entered into the second data store, and output may thereafter no longer be presented as a predicted quantitative value.
  • FIG. 6 shown therein is an example method 600, for determining the quantitative value of a material property parameter following prediction by a computer system of the present disclosure.
  • the method starts with providing an input value or range of values of material property parameters (step 605).
  • the computer system then predicts one or more quantitative values (step 610) and provides output (step 615) in the form of non-elemental materials exhibiting quantitative values of material property parameters corresponding with the input.
  • steps 605, 610 and 615 together correspond with example method 205.
  • the non-elemental materials are then synthesized or otherwise obtained (step 620) and tested experimentally (step 625), typically in a laboratory, in order to empirically determine the quantitative value of the material property parameter (step 630).
  • the empirically determined quantitative value of the material property parameter can be compared with the predicted value, and the empirically determined quantitative value is provided as input to the computer system (step 635). It is noted that step 635 corresponds with example method 202.
  • the computer system may be configured to retain both the predicted value and the empirical value, or in other implementations, the predicted value may be removed from the data store upon the addition of an experimental value.
  • Example method 500 starts with the loading of an interface capable of receiving commands or information in the form of user inputs provided through a user input module.
  • User input can consist of: (i) one or more desired quantitative values of material property parameters (step 502); (ii) material descriptors associated with a non-elemental material (step 512); (iii) quantitatively defined values for material property parameters associated with a material (step 522); (iv) atomistic fingerprints, or data for the assembly thereof associated with a material (step 532).
  • the computer system can identify that a non-quantitatively defined value for material property parameters associated with a material is absent in the computer system (step 542).
  • method 500 After method 500 is started, it waits for a user input command in the form of any one of the aforementioned (i), (ii), (iii) or (iv). Then, method 500 proceeds depending on the input provided.
  • method 500 will identify materials corresponding with the desired quantitative values of material property parameters (step 504) and display the materials to an output module (step 506). If input (ii) is provided, method 500 proceeds with providing the material descriptors to a first data store (step 514). If input (iii) is provided, method 500 proceeds with providing the quantitatively defined values for material property parameters associated with the material to a second data store (step 524). If input (iv) is provided then method 500 proceeds with providing atomistic fingerprints, or data for the assembly thereof associated with the material to a third data store (step 534).
  • method 500 proceeds with prediction of these quantitative values (step 544), and inclusion of the predicted material property parameters to the fourth data store (step 546).
  • the method relating to the provision of input (i) corresponds with the exemplified method 200, hereinbefore described; the method relating to the provision of input (ii) corresponds with exemplified method 201 , hereinbefore described; the method corresponding with the provision of input (iii) corresponds with exemplified method 202, hereinbefore described; and the method relating to the provision of input (iv) corresponds with the exemplified methods 203, 204, hereinbefore described.
  • the method relating to the prediction of quantitative values corresponds with exemplified methods 205, 206, hereinbefore described.
  • a computer system can be configured to include a plurality of user input modules, or a plurality of user output modules.
  • FIG. 4 the present disclosure provides, in an implementation, an example of a computer system 400 including user input module 125a, 125b, 125c, processing element 1 15, memory element 120, and user output modules 130a, 130b and 130c.
  • Processing element 1 15 and memory element 120 can be housed together, for example, in terminal 1 10.
  • User input modules 125a, 125b and 125c and user output modules 130a, 130b, and 130c are coupled to processing element 1 15 via network 410 and 420, respectively, in order to permit communication between processing element 1 15 and user input modules 125a, 125b and 125c, and user output module 130a, 130b, and 130c, respectively.
  • user input modules 125a, 125b, and 125c can be configured so that one or more of the methods described herein, and, as hereinbefore exemplified, by method 200, 201 , 202, 203, 204, 205 or 206 can be performed.
  • a user input module e.g., user input module 125a
  • a user input module e.g., 125b
  • a user input module e.g., 125b
  • a variety of configurations involving a plurality of user input modules or user output modules can be assembled. It will be clear to those of skill in the art that, depending on the specific configuration selected, some of the herein described parts can become superfluous and can be omitted in such configuration.
  • the processing element does not need to be configured to be operable to access the third data store.
  • the computer system and methods of the present disclosure permit the identification by a user of non-elemental materials having one or more specified desired material property parameter values.
  • the non- elemental materials may be obtained and used in a material application, notably advantageously, in an application in which it is of significance that the material exhibits the desired quantitative values of the material property parameters.
  • material applications can range across an extremely wide spectrum, including industrial or domestic applications, depending on the desired function and performance of the material.
  • Material applications include, without limitation, engineering applications, such as mechanical, chemical and civil engineering applications; electronic applications; metallurgical applications; and medical applications.
  • the present disclosure includes a method of identifying and using a non-elemental material having a defined quantitative material property value in a material application.
  • the method can include the steps of (a) identifying a non-elemental material having a defined quantitative material property value using a computer system or method according to the present disclosure; (b) obtaining the non- elemental material; and (c) using the non-elemental material in a material application. This method is now further illustrated with reference to FIG. 7.
  • Method 700 starts with the identification of desired material properties for an application (step 705).
  • a desired quantitative value or range of values of one or more material property parameters is then provided in the form of input to the computer system (step 710).
  • the computer system identifies materials exhibiting the desired values of material property parameters, which as will be clear from the foregoing, can be either previously experimentally determined or predicted by the computer system (step 715), and provides non-elemental materials in the form of output (step 720).
  • the material can then be obtained (step 725), for example by synthesizing the material, and tested and used in the material application (step 730).
  • At least some of the elements of the various computer systems described herein are implemented via software and may be written in a high- level procedural language such as object oriented programming or a scripting language. Accordingly, the program code may be written in C, C++ or any other suitable programming language and may include modules or classes, as is known to those skilled in object oriented programming. Alternatively, at least some of the elements of the various computer systems described herein that are implemented via software may be written in assembly language, machine language or firmware. In either case, the program code can be stored on a storage media or on a computer readable medium that is readable by a general or special purpose electronic device having a processor, an operating system and the associated hardware and software that implements the functionality of at least one of the implementations described herein. The program code, when read by the electronic device, configures the electronic device to operate in a new, specific and defined manner in order to perform at least one of the methods described herein.
  • the methods described herein are capable of being distributed in a computer program product including a transitory or non-transitory computer readable medium that bears computer usable instructions for one or more processors.
  • the medium may be provided in various forms such as, but not limited to, one or more diskettes, compact disks, tapes, chips, USB keys, external hard drives, wire-line transmissions, satellite transmissions, internet transmissions or downloads, magnetic and electronic storage media, digital and analog signals, tablet (e.g., iPad) or smartphone (e.g., iPhones) apps, and the like.
  • the computer useable instructions may also be in various forms, including compiled and non-compiled code.
  • An example computational system was assembled to include an input and output module, as well as a processing element and access to data stores, as shown in FIG. 1 .
  • a user can provide a query to the input module of the computer system.
  • a user can specify a desired property, or set of material properties, as well as a value or range of values for each one of the material properties he/she wishes to investigate.
  • the output module displays materials exhibiting the specified quantitative values of the material property or properties provided by the user.
  • the computer system includes data stores which include a set of quantitative values of material properties associated with a wide range of materials. Values for material property parameters, which are empirically known, or theoretically derived, have been stored in a second data store, and material property values as predicted by machine learning algorithm have been stored in a separate data store namely a fourth data store (see FIG. 1 ).
  • a first data store holds information to uniquely identify each material present in the computer system.
  • These data stores, specifically the second and fourth data store are searched for materials in response to a user query. Materials corresponding with the desired quantitative property values provided are displayed to the user in the form of results on the output module.
  • the fourth data store is populated with predicted values of material parameters. Predictions are made by the computer system on the basis of atomistic fingerprints of each of the materials in first data store by a machine learning algorithm. In the present example a neural network algorithm is further described. However, one can also use boosted regression trees, or support vector regression, or any other machine learning algorithms.
  • the inventors have built training and validation sets for a neural network algorithm from a dataset comprising entries from the second and third data stores entries as described below. Once the neural network has been trained and the accuracy of the results were evaluated, based on a validation set, the inventors used it to predict the same set of properties for other compound for which fingerprints have been identified and stored in the third data store.
  • the database comprising of first, second, third and fourth data stores, is an integral part of a broader computing platform.
  • the data and information contained within the computing platform includes:
  • the data can originate from various sources like for example ICSD, COD, Pauling File, Pearson's Crystal Data. This data can be further augmented with computationally derived descriptors like electronic band structure, or atomic lattice vibration information computed on the basis of Density Functional Theory (DFT) calculations (included in the first data store).
  • DFT Density Functional Theory
  • This data can originate from various sources (included in the second data store).
  • Information about theoretically derived values of material properties for example on the basis of DFT, or Molecular Dynamics, or any other computational method enabling theoretical prediction of material properties (included in the second data store).
  • Fingerprints can be based on many types of data and contain variety of information about crystal structure, electronic band structure, vibrations of the atomic lattice, or even the combination of these groups. Their purpose is to build features for the machine learning algorithms.
  • a single atom is selected and based on individual positions of neighboring atoms Euclidian norms are constructed and used to radially sort all atoms in the crystal. Having an ascending list of Euclidian distances between neighboring atoms, the number of unit cell repetitions needed is calculated to generate the Coulomb matrices of a previously specified size. In this example, matrices of size 20x20 have been used. This is the size of the fingerprint matrices.
  • An example fingerprint of titanium aluminide is shown in FIG. 12.
  • a neural network algorithm mimics a biological network of neurons interconnected together to form a structure. Many simulated neurons, which in the algorithm can be considered as signal processors, are connected and analogously form an artificial neural network structure. Each neuron receives input signals and transmits a response when the total input exceeds threshold value. One can think about this signal processor as a neuron with a Heaviside activation function. Neural networks are one of the most popular methods used to solve classification problems, but they can also be applied in regression (T. Hastie, R. Tibshirani, J. Friedman, The elements of statistical learning, (Springer series in statistics Springer 2008).
  • a neural network takes feature vectors as input and using multiple layers of neurons the algorithm creates nonlinear features which can be further used for regression.
  • the neurons are organized in layers for which output of one layer becomes an input for next layer. Only adjacent layers are connected together through weights and activation functions and the parameters are obtained by optimizing the loss function using training data.
  • An example structure of a single layer neural network is shown in FIG. 10.
  • the action of each layer of neurons given an input vector of features (1) (l, x 1 , x 2 , x 3 , ... , x u ... , x p ) is expressed algebraically by the following equations:
  • the input feature vector When constructing the input feature vector one needs to add 1 as the bias feature to account for the constant term in regression.
  • non-linear activation function g(x) is used and a transformation parametrized by coefficients derived feature vector (2) is obtained from feature vector ( 1) .
  • the output layer serves as the prediction vector.
  • the elements of all weight matrices ⁇ as the model parameters, can be determined during training (learning) process.
  • the number of layers and number of neurons in the layer define the network architecture.
  • Example results shown in FIGS. 1 1A and 1 1 B were obtained using a neural network of 400-600-200-40-1 nodes respectively.
  • the first layer was constructed to accept the fingerprints.
  • the elements of 20x20 matrices as shown in FIG. 10, were rearranged to form a vector. To accomplish this, the matrix was rewritten so that all columns of the fingerprint matrix were assembled together to form a 400 entries long vector.
  • Nonlinear activation function was applied to the next two hidden layers, as defined by equation (3). To the last two layers linear activation functions were applied.
  • FIG. 1 1A shows results obtained for mass density for both training and validation data sets.
  • FIG. 1 1 B shows results for formation enthalpy.
  • An example computer system was assembled and provided with an example user input to identify materials having a density of ⁇ 3 g/cm 3
  • the computer system provided the following materials: distrontium chromium antimonate, silico, tetrasodium dimanganese (II I) oxide, calcium magnesium bis(cyclo-trioxopentahydroxotriborate) hexahydrate, potassium cerium (I II) cerium (IV) tetrakis(sulfate(VI)), as shown in FIG. 8.
  • An example computer system was assembled and provided with an example user input to identify materials having a density of ⁇ 3 g/cm 3 and an enthalpy of formation of ⁇ -200 kJ/mol.
  • the computer system provided the following materials: potassium sodium iron cyclo-tetrasilicate; lanthanum gold indium, neodymium copper oxide, zinc-dilaluminum sulfide (alpha) and ammonium tetrachloroplatinate (II), as shown in FIG. 9. It is noted that some materials are predicted (see: under column labeled "Type") to exhibit a density of ⁇ 3 g/cm 3 and an enthalpy of formation of ⁇ -200 kJ/mol (e.g.
  • An example computer system was assembled and provided with an example input for inclusion in the data store.
  • two quantitative values of two material property parameters density and enthalpy of formation
  • two materials cobalt antimonide and copper phosphide
  • the data store of the computer comprised a value of 7.61 g/cm 3 and -62.761 kJ/mol for the density and enthalpy of formation, respectively, for cobalt antimonide, and a value of 4310 kg/m 3 and -121.001 kJ/mol, respectively, for the density and enthalpy of formation, respectively, for copper phosphide.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Analysing Materials By The Use Of Radiation (AREA)

Abstract

L'invention concerne des systèmes informatiques et des procédés d'identification de matériaux non élémentaires sur la base de propriétés atomistiques. Les systèmes informatiques et procédés peuvent permettre l'identification de matériaux non élémentaires ayant une ou plusieurs propriétés de matériaux souhaitées qui peuvent être spécifiées sous la forme d'une entrée utilisateur. Les matériaux identifiés peuvent être connus ou prédits par les systèmes informatiques et les procédés pour présenter les propriétés de matériaux souhaitées.
PCT/CA2017/051449 2016-12-02 2017-12-01 Systèmes informatiques et procédés d'identification de matériaux non élémentaires sur la base de propriétés atomistiques WO2018098588A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US201662429529P 2016-12-02 2016-12-02
US62/429,529 2016-12-02

Publications (1)

Publication Number Publication Date
WO2018098588A1 true WO2018098588A1 (fr) 2018-06-07

Family

ID=62241135

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CA2017/051449 WO2018098588A1 (fr) 2016-12-02 2017-12-01 Systèmes informatiques et procédés d'identification de matériaux non élémentaires sur la base de propriétés atomistiques

Country Status (1)

Country Link
WO (1) WO2018098588A1 (fr)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111986740A (zh) * 2020-09-03 2020-11-24 平安国际智慧城市科技股份有限公司 化合物分类方法及相关设备
EP3757909A1 (fr) * 2019-06-24 2020-12-30 Tata Consultancy Services Limited Procédé et système de conception de produits formulés
CN113454728A (zh) * 2019-02-12 2021-09-28 Jsr株式会社 数据处理方法、数据处理装置以及数据处理系统
CN113505527A (zh) * 2021-06-24 2021-10-15 中国科学院计算机网络信息中心 一种基于数据驱动的材料性质预测方法及系统
JP2021174402A (ja) * 2020-04-28 2021-11-01 株式会社日立製作所 材料の特性値を推定するシステム
JP2021174403A (ja) * 2020-04-28 2021-11-01 株式会社日立製作所 材料の特性値を推定するシステム
WO2022251233A1 (fr) * 2021-05-24 2022-12-01 Amatrium Inc. Assistant de matériaux à intelligence artificielle

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2235160A1 (fr) * 1995-10-18 1997-04-24 Shell Canada Limited Procede de prediction d'une propriete physique d'un materiau hydrocarbone residuel
CA2269669A1 (fr) * 1996-11-04 1998-05-14 3-Dimensional Pharmaceuticals, Inc. Systeme, procede et programme produit informatique pour identifier des composes chimiques presentant des proprietes desirees
WO2003048759A1 (fr) * 2001-11-30 2003-06-12 Exxonmobil Research And Engineering Company Procede permettant d'analyser une matiere inconnue sous forme de melange de matieres connues calcule de maniere a correspondre a certaines donnees analytiques et de predire des proprietes de la matiere inconnue sur la base du melange calcule
JP2007018444A (ja) * 2005-07-11 2007-01-25 Yamato Hiroshi 新規材料の構成物質情報探索方法、及び新規材料の構成物質情報探索システム

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2235160A1 (fr) * 1995-10-18 1997-04-24 Shell Canada Limited Procede de prediction d'une propriete physique d'un materiau hydrocarbone residuel
CA2269669A1 (fr) * 1996-11-04 1998-05-14 3-Dimensional Pharmaceuticals, Inc. Systeme, procede et programme produit informatique pour identifier des composes chimiques presentant des proprietes desirees
WO2003048759A1 (fr) * 2001-11-30 2003-06-12 Exxonmobil Research And Engineering Company Procede permettant d'analyser une matiere inconnue sous forme de melange de matieres connues calcule de maniere a correspondre a certaines donnees analytiques et de predire des proprietes de la matiere inconnue sur la base du melange calcule
JP2007018444A (ja) * 2005-07-11 2007-01-25 Yamato Hiroshi 新規材料の構成物質情報探索方法、及び新規材料の構成物質情報探索システム

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
BOTU ET AL.: "A learning scheme to predict atomic forces and accelerate materials simulations", PHYS. REV. B, vol. 92, no. 094306, 25 September 2015 (2015-09-25), pages 1 - 5, XP080792487, Retrieved from the Internet <URL:https://dasher.wustl.edu/chem430/readine/phvsrevb-92-094306-15.pdf> *
DI PASQUALE ET AL.: "Optimization Algorithms in Optimal Predictions of Atomistic Properties by Kriging", J. CHEM. THEORY COMPUT., vol. 12, no. 4, 2016, pages 1499 - 1513, XP055508176, Retrieved from the Internet <URL:http://pubs.acs.ore/doi/pdfl0.1021/acs.jctc.5b00936> *
FERNANDEZ ET AL.: "Data Analytics and Machine Learning in Nanomaterials Discovery", 2016 AICHE ANNUAL MEETING: PROCEEDINGS : COMPUTATIONAL MOLECULAR SCIENCE AND ENGINEERING FORUM: DATA MINING AND MACHINE LEARNING IN MOLECULAR SCIENCES II, 18 November 2016 (2016-11-18), pages 1 - 35, XP055508168, Retrieved from the Internet <URL:http://vowconference.com.au/slides/vowdata2016/Fernandez-DataAnalvtics.pdf> *
RAMPRASAD ET AL.: "Machine Learning and Materials Informatics: Recent Applications and Prospects", NPJ COMPUTATIONAL MATERIALS 3, 13 December 2017 (2017-12-13), pages 1 - 27, XP080778817, Retrieved from the Internet <URL:https://arxiv.org/pdf/71707.07294.pdf> *

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113454728A (zh) * 2019-02-12 2021-09-28 Jsr株式会社 数据处理方法、数据处理装置以及数据处理系统
EP3757909A1 (fr) * 2019-06-24 2020-12-30 Tata Consultancy Services Limited Procédé et système de conception de produits formulés
JP2021174402A (ja) * 2020-04-28 2021-11-01 株式会社日立製作所 材料の特性値を推定するシステム
JP2021174403A (ja) * 2020-04-28 2021-11-01 株式会社日立製作所 材料の特性値を推定するシステム
WO2021220776A1 (fr) * 2020-04-28 2021-11-04 株式会社日立製作所 Système estimant une valeur caractéristique de matériau
WO2021220775A1 (fr) * 2020-04-28 2021-11-04 株式会社日立製作所 Système permettant d'estimer une valeur caractéristique d'un matériau
JP7339923B2 (ja) 2020-04-28 2023-09-06 株式会社日立製作所 材料の特性値を推定するシステム
JP7339924B2 (ja) 2020-04-28 2023-09-06 株式会社日立製作所 材料の特性値を推定するシステム
CN111986740A (zh) * 2020-09-03 2020-11-24 平安国际智慧城市科技股份有限公司 化合物分类方法及相关设备
CN111986740B (zh) * 2020-09-03 2024-05-14 深圳赛安特技术服务有限公司 化合物分类方法及相关设备
WO2022251233A1 (fr) * 2021-05-24 2022-12-01 Amatrium Inc. Assistant de matériaux à intelligence artificielle
CN113505527A (zh) * 2021-06-24 2021-10-15 中国科学院计算机网络信息中心 一种基于数据驱动的材料性质预测方法及系统

Similar Documents

Publication Publication Date Title
WO2018098588A1 (fr) Systèmes informatiques et procédés d&#39;identification de matériaux non élémentaires sur la base de propriétés atomistiques
Frady et al. A theory of sequence indexing and working memory in recurrent neural networks
Vasudevan et al. Machine learning for materials design and discovery
Zeng et al. Revealing high-fidelity phase selection rules for high entropy alloys: A combined CALPHAD and machine learning study
Xue et al. Accelerated search for BaTiO3-based piezoelectrics with vertical morphotropic phase boundary using Bayesian learning
EP3757995A1 (fr) Procédé et appareil de recommandation de musique, et dispositif informatique et support
CN110362723B (zh) 一种题目特征表示方法、装置及存储介质
US9652558B2 (en) Lexicon based systems and methods for intelligent media search
JP6069419B2 (ja) データベース検索方法、システムおよびコントローラ
US20060074594A1 (en) Systems and methods for predicting materials properties
JP7353874B2 (ja) 材料特性予測装置および材料特性予測方法
JP2000276474A (ja) データベース検索装置および方法
Žitnik et al. Matrix factorization-based data fusion for gene function prediction in baker’s yeast and slime mold
Naden Robinson et al. On the chain-melted phase of matter
CN110728047A (zh) 一种基于机器学习性能预测含能分子计算机辅助设计系统
Longere et al. Enlarged finite strain modelling incorporating adiabatic shear banding and post-localization microvoiding as shear failure mechanisms
Jiang et al. Hierarchical multiscale simulations of crystalline β-octahydro-1, 3, 5, 7-tetranitro-1, 3, 5, 7-tetrazocine (β-HMX): Generalized interpolation material point method simulations of brittle fracture using an elastodamage model derived from molecular dynamics
Yao et al. Artificial neural network modelling to predict hot deformation behaviour of as HIPed FGH4169 superalloy
CN115952307A (zh) 基于多模态图对比学习的推荐方法、电子设备及存储介质
CN114144110B (zh) 合成化合物的方法
Yeh et al. An efficient content-based time series retrieval system
Reddy Addula et al. Review of the Frenkel–Ladd technique for computing free energies of crystalline solids
KR102437392B1 (ko) 물리 기반 심층 신경망 모델을 이용한 분자 특성 예측 방법 및 분석 장치
Zhao et al. An efficient active set method for optimization extreme learning machines
Zheng et al. Image and Data Mining in Reticular Chemistry Using GPT-4V

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17876294

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 17876294

Country of ref document: EP

Kind code of ref document: A1