Could Mendelev have Dreamt in OWL?

Robert StevensBioHealth Informatics GroupSchool of Computer Science

University of ManchesterRobert.Stevens@manchester.ac.uk

IntroductionThe challenge• Periodicity in physicochemical properties• Modelling inorganic chemistry• Defining a metal• Defining a salt• Alkali metals, alkaline earths• Halogens• Transition elements• Noble gases• What cannot be done

What the Ancients Knew• Iron, lead, copper, silver, gold, tin,

mercury, carbon, sulphur;• From 1660 to 1860 found sixty more

elements• Needed to organise them

What did the victorians Know?

• Electronic Structure of atom not known• Atomic weight (mass)• Basic physicochemical properties• Proportions in which elements combined

in salts: Moles• Ions: Charged particles

Moles

• 602,000,000,000,000,000,000,000.• 6.02 times 10^23• SI base unit for amount of substance• A Mole contains Avogadro’s number of

entities (atoms, ions, molecules, etc.)• The mass in grams of a Mole of an atom

equals the mass in AMU• 12g Carbon and Carbon has mass 12 amu

Dmitry Mendeleev

…credited as being the creator of the first version of the periodic table of elements.

- Wikipedia 2008

The Periodic Table

• Arrange elements in order of increasing atomic weight

• Notice periodicity in physicochemical properties

• Arrange similar elements to be vertically juxtaposed

• Predictive gaps• Infer properties of element in gaps from

surrounding elements

Periodic Table

Al

? ZnGe

In

Predicting Gallium

Ga

The Task

• How much of the grouping within the periodic table can we infer using OWL and automated reasoning?

• Use only kinds of physicochemical properties known in 1860 or modern equivalents (atomic mass not atomic weight)

The Modern Way• Mendeleev’s periodic table based on chemistry matches

the electronic structure of the atom• 1860’s vs 1920s• Alkali metal = valence shell is s with one electron• Alkaline earth metals = valence shell is s with two

electrons• Halogens = valence shell is P with five electrons• Transition metals = valence shell is d with one -- 10

electrons• Noble gases = valence shell is p with six electrons• Explains all chemistry (with a couple of wrinkles for

hydrogen and helium)

Thomas leaves the room

Thing

AmountOfSubstance PhysicalObject

Atom Ion Molecule

MoleOfIon MoleOfCompoundChemical MoleOfElementalChemicalMoleOfMolecule

MoleOfAtom

FundamentalSubAtomicParticle

Thomas enters the room

Assumptions

• Standard conditions• Substances have only one boiling point

etc.• No units• Functional data type properties• Data ripped from NIST, Wikipedia, the

Web, my school data book• Data a little flakey

Ontology Statistics• 721 classes• 28 defined classes• 18 data type properties• 12 object type properties• 3294 subclass axiom• 40,316 disjoint axiom• DL expressivity: ALCHQ(D)• 25 mins to classify with 64 bit FaCT++ on 2 x 2.66

Ghz Dual-Core Intel Xeon, 16 GB ram• With “anywhere blocking” takes 10 seconds

Class: SodiumAtom SubClassOf: Atom that hasPart exactly 11 Proton

and hasPart exactly 11 Electron and hasPart only (Proton or Neutron or Electron) and formsIon some SodiumIon and formsIon only SodiumIon and hasAtomicRadius value 77.2f and hasAtomicVolume value 23.7f and hasAtomicWeight value 22.98977f and hasElectronAffinity value 52.868f and hasElectronNegativity value 0.93f and hasFirstIonisation value 495.8f and hasSecondIonisation value 4562.4f and hasThirdIonisation value 6912.2f DisjointWith: <all other atoms>

Mole of AtomClass: MoleOfAtom

madeOfAtom exactly 602000000000000000000000 Atom

Mole of Sodium AtomClass: MoleOfSodiumAtom

SubClassOf: MoleOfAtomthat madeOfAtom onlysome SodiumAtomand hasColour some SilverColourand hasCrystalStructure some BodyCentredCubicStructureand hasState some SolidStateand hasMeltingPoint value 371.01and hasMeltingPoint value 371.01and hasBoilingPoint value 1154.6and hasDensity value 0.971and hasElectricalConductivity value 209.6and hasHeatOfAtomisation value 109.0and hasHeatOfFusion value 2.598and hasSpecificHeat value 1.23and hasThermalConductivity value 142.0and hasVapourisation value 96.960

Salts

• A salt is an ionic compounds of at least one metal ion and at least one non-metal ion

• That is, some cation and some anion• Distinction between metal and non-metal

fundamental to basic inorganic chemistry

Defining a Metal

• Formation of metal bonds; nuclei in sea of electrons;

• Freely moving electrons gives conductivity etc.

• So, defining indirectly...• Really an extensional definition: The

metals are those I say are metals• Almost everything a metal at high enough

pressure

Electrical Conductivity

Mole of Metal atomClass: MoleOfMetalAtom

equivalentTo: MoleOfAtomthat hasElectricalConductivity some double[>= 5.405]

• Infers all moles of metal• Conductivity proxy for “sea of electrons”

Mole of Metal Atom

• I need metal ion and mole of metal ion (plus non-metal equivalents)

• Therefore, need metal atom

Defining Metal AtomClass: MetalAtom

EquivalentTo: Atomthat formsMoleOfAtom onlysome MoleOfMetalAtom

• Somewhat dodgey• Assumption that all atoms are parts of moles—collectives of those atoms• Optionality or “may”...

• Assumption of purity etc.

Metalloids

• Some properties of metals and some of non-metals

• Not equivalent to semi-conductor (carbon semi-conducts)

• Tend to be amphoteric• Another extensional definition

Non-Metals

• Don’t conduct (except carbon in its graphite isoform)

• NOT (MetalAtom or MetalloidAtom)

Once we have Metals

• Metal Atoms• Metal Ions• Moles of Metal Ion, non-metal ion• Cation = positively charged ion (metal ion)• Anion = negatively charged ion• Metal/non-metal cation/anion basis of

much of the element’s chemistry

Mole of Sodium ChlorideClass: MoleOfSodiumChloride

SubClassOf: MoleOfCompoundChemicalthat madeOfMoleOfIon exactly 1 MoleOfChlorideIonand madeOfMoleOfIon exactly 1 MoleOfSodiumIonand madeOfMoleOfIon only (MoleOfChlorideIon or MoleOfSodiumIon)and hasMolarMass value 58.442and hasDensity value 2.16and hasColour some WhiteColourand hasState some SolidStateand hasBoilingPoint value 1465.0and hasMeltingPoint value 801.0and hasSolubilityInWater value 35.9

Defining a Salt• “A salt”, not salt• Some metal ion and some anion: NaCl, KCl, CaCO3

Class: MoleOfSaltEquivalentTo: MoleOfChemicalCompoundthat madeOfMoleOfIon some MoleOfMetalIonand madeOfMoleOfIon some MoleOfAnionand madeOfMoleOfIon only (MoleOfMetalIon or MoleOfAnion)

• Also, crystal structure, etc.

Defining Alkali Metal Salts• Form halid in ratio 1:1 • Form oxide in ratio 2:1

Class: MoleOfAlkaliMetalChlorideEquivalentTo: MoleOfCompoundChemicalthat madeOfMoleOfIon exactly 1 MoleOfMetalIonand madeOfMoleOfIon exactly 1 MoleOfChlorideIonand madeOfMoleOfIon only (MoleOfMetalAtom or MoleOfChlorideIon)

• Want to say “and in no other ratio”• Really like to use other chemical properties if necessary

Salt Density (g/cm3)

Caesium chloride 3.99 Copper (I) chloride 4.140 Gold (I) chloride 8.25 Lithium chloride 2.07 Potassium chloride 1.987 Rubidium chloride 2.8 Silver (I) chloride 5.56Sodium chloride 2.16Thallium (I) chloride 7.004

Alkali Metal Chloride Data

Defining Alkaline Earth Salts

• Same technique• Works less well• Ratio, density, colour get all of the alkaline

earths, but include too much• Need trends and reactions

Alkaline Earth OxidesClass: MoleOfAlkaliEarthOxide

EquivalentTo: MoleOfCompoundChemicalthat madeOfMoleOfIon exactly 1 MoleOfMetalIonand madeOfMoleOfIon exactly 1 MoleOfOxygen2Ionand madeOfMoleOfIon only (MoleOfMetalIon or MoleOfOxygen2Ion)and hasDensity some double[<= 6.0]and hasColour some WhiteColour

• Again light, but barium and strontium confound; use colour• Leaves only Zinc(II) oxide• Very soluble; alkaline earths moderately soluble or react with water

…This works better

Class: AlkalineEarthAtomEquivalentTo: AtomThat formsIon onlysome (Ion that hasCharge exactly 1 PositiveCharge)

• Not really based on 1860’s chemistry and is not ontologically good

Defining Halides

• Same methods workClass: HalogenAtom

EquivalentTo: Atomthat formsIon some

(Ion that hasCharge exactly 1 PositiveCharge and hasCharge only PositiveCharge

and hasCharge min 1 PositiveCharge)

• Gets all and only halogens• Works for group 1 and group 2• Maybe others too

Defining Noble Gases

• They don’t really do anything• Don’t form ions (except under extreme

conditions)• Might look at ionisation energy• This is where the open world assumption

doesn’t help

Transition Elements

• Variable oxidation state• Dense• Coloured compounds• Not quite exclusive conditions (lead has

multiple oxidation states)• Trends confound simple definitions

• More sophisticated relationships?

Counting Oxidation State

• Give me atoms that have more than one oxidation state

CopperAtom formsIon (CopperIon that hasCharge exactly 1 PositiveCharge)CopperAtom formsIon (Copper2Ion that hasCharge exactly 2 PositiveCharge)

• Can’t just ask the simple question• Have to ask for all combinations: 1 or 2; 1

or 3; 1 or 4;...; 2 or 3;...3 or 4; ...

What do I need?

• Describe trends nicely• Describe relationships between values for

data types (reactivity, volume, mass, radius, etc)

• Reactions• Counting

What’s Next?

• Cannot describe giant structure of salts in OWL

• So, describe their features• Molecules• Molecular structure hard to describe• Describe features or describe the

molecule• Description graphs offer a way out

Carbon dioxide (CO2)

C

OO

Cyclohexane

Image from http://www.chemvc.com/~tim/organic%20alkanes.html

Conclusions

• It is possible to make a lot of inferences that could help with formulating a periodic table

• Definite limitations• Prediction not on the cards...• Scalability a issue• Modelling compromises• Should I do this in OWL?

Acknowledgements

• Simon Jupp• Dmitry Tsarkov• Alan rector• Uli Sattler