Hello. I'm  Dr  Michael  Nazarkovasky,

Ukrainian  researcher  in  chemistry and  materials  science  from  Brazil.

Evolved  data  driven  solutions   in  my  area  of  knowledge  and  expertise.

The  presentation  is  made   in  collaboration  with  Ukrainian  colleagues

from  National  Academy  of  Sciences  of  Ukraine,

supporting  and  promoting their  scientific  research  programs

during  such  a  difficult  period.

The  subject  says  on  statistical   and  data  analysis  approaches,

deepen  conception  behind the  experimental  results  and  phenomena.

In  particular,   unsupervised  methods  are  helpful

for  multivariate  cases  like  this

when  two  or  more  classes  of  materials   are  characterized  by  a  large  body

of  the  parameters  measured  or  calculated in  the  course  of  the  lab  analysis.

This  case  is  about  hybrid  materials

which contain mixed  nanoxides   and  carbon  phases.

The  nano hybrids  combine  properties  of  both  components;

well  ordered  micro  and  meso porosity,

a large  surface  area   and  high  porosity  in  general,

high  corrosion  resistance,

thermal  and  mechanical  stability,

hydrophobicity  and  high  conductivity due  to  the  presence  of  carbon,

developed for active sites  attributed  to  the  metal  oxide  nanoparticles.

Hence  reasons  of  the  subject   Nanomaterials' I mportance

exists in  the  variety   of  their  applications;

Catalysis, adsorption,  sensors, energy  field,  and  hydrogen  adsorption.

Typical  preparation  of  binary  oxides, non carbon  oxide  nanocomposites

consists  in  three  stages.

On  the  first  stage,  preparation of  the  homogeneous  dispersion  of  silica

in  the  aqueous  solution of  the  corresponding  metal  acetate

under  stirring  at  room  temperature  was  conducted.

On  the  second  stage,

dispersions  were  dried   at  130  degree  C  during  five  hours

and  sift through the  0.5  millimeter  mesh.

At  the  last  stage  all  of  the  10  powders were   [inaudible 00:02:26]   for  2  hours

at  600  degree  C  in  air.

Also  the  reference  sample  of   fumed silica  without  metals

was  treated  under  the  same  conditions,

by  bringing  all  these  three  stages;

homogenization of  the  aqueous  dispersion,

drying,  grinding,  sieving,  and  carbonization   at  the  same  temperature.

The  process  of  impregnation   of  fumed  silica  with  an  aqueous  solution

of  metal  acetate and  subsequent  removal  of  the  solvent,

the  adsorbed acetates,  are  distributed uniformly  over  the  matrix  surface.

Modification  of  reserves   of  resorcinal  formaldehyde polymer

by  oxide  nano composite  was  carried  out

by  easing the  process mixing  resorcinal formaldehyde,

and  this  binary pre-synthesized nanocomposites

reference  silica  the  weight  ratio

of  an  aqueous solution  under  stirring  at  room  temperature.

All  these  samples  were  sealed  and  placed

in  a  thermostatic  oven  for  polymerization,

and  all  synthesized  composites  were  seized

with  further  drying   at  90  degrees  C  for  10  hours.

Carbonization  of  the  samples   was  carried  out

in  a  tubular  furnace   under  nitrogen  atmosphere

upon  heating  from  room  temperature  up  to  800  degrees  C

at  a  heating  rate   of  5  degrees  C  per  minute,

and  kept  at  a  maximum temperature  for  two  hours.

Hence  the  carbonized  samples  are  labeled  as  C

and  the  initial  materials  which  do  not  contain  carbon

are  [inaudible 00:04:29].

Actual properties  of the… Or  in  other  word,

parameters  of  porosity  were  calculated using  modified Nguyen-Do  method

from  the  low  temperature  nitrogen  adsorption- desorption.

This  is  a  standard   [inaudible 00:04:49]   method  for  porosity

and  it's  called  the   [inaudible 00:04:51]   .

The  calculated  parameters   are  assigned  as  variables

for  further  data  processing  using JMP.

To  be  more  exact,

the  specific surface  area  and  total pore  volume

were  derived  from  the  BET  measurements

and  then served  to  calculate

respective  portions  of  micro, meso,  and  macro porosity.

In  this  case  we  have  a  set of inward  variables.

For  example, Nguyen-Do  method   was  developed  initially

for  calculation  of  carbon  materials with  a  sleeve-like  porosity,

afterward  by  one  of  the  co  authors  of  the presentation,  Professor  Vladimir Gun'ko .

The  method  was  modified  and  extended

through  a  large  variety  of  other  materials

which may  contain  cylinders,  also  slits, and  voids  among  the  particles

within  the  aggregates and  agglomerates  of  aggregates.

Not  only  for  carbon  materials

and  the  method  serves  also  for  hybrids,

as  for  individual  materials as  for  hybrids  also.

So  let's  start  from  the  basics.

In  multivariate  analysis  indicates  that  not  all  the  parameters  around  the  health

are  well  related  with  each  other   in  case  of  non  carbon  materials.

Specifically  there  a meso porosity   dominates  overly  serious  as  shown

on  the  heat  map and  from  the  table.

The  parameters  corresponding to  microporosity

are  demonstrating  correlation only  within  their  group

and  group  and  with  macroporosity. Yes,  surprisingly.

Contrastingly,  the  heat  map of  the  carbonized  nano  hybrid

speaks  for  more  consistent   and  more  ordered  structure

with  almost  complete  correlation between  all  the  types  of  the  porosity,

whereas  the  role  of  microporosity is  significantly  increased.

Comparison  between  parameters  or  variances  

reveals  the  differences especially  for  micro porosity.

For the specific  surface  area, as  for a total pore  volume,

for example, total pore volume and  volume  of  the  mesop ores

can  be  also  considered  as  factors

to  claim  the  difference  between both  types  of  the nanomaterials.

All  eight  parameters  were  taken  to  perform  hierarchical  clustering

and  it  is  easy  to  see  that the  minimal  optimal  in  the  same  time  model

can  be  offered  for  three  clusters on  the  cluster  path

and  on  the  constellation  plant.

Think oxide  sample, it cannot  separate within  the  non  carbon  group

but  can  be  attributed in  other  carbonized  cluster.

Well,  main  parameters  as  seen   from  the  summary  are  the  volume

and  surface  of  the  micro pores. In  other  words,  micro porosity.

Indeed,  some  parameters; surface  and  volume  of  macro  and  mesopores

are  out  of  the  group  samples  of  the  non  carbon  nanomaterials.

I'm  talking  about,   namely  for  a  sync oxide.

Anyway,  the  mean  values  for  both  parameters

do  not  match  over  the  whole parallel  plot  of  their  clusters.

Principal  components  analysis   help  to  represent  all  variables

in  three   linear  combinations.

According  to  the   [inaudible 00:09:02]   ,  the eigen values  less  than 1

are  not  taken  into  consideration.

This  is  why  we  have  only  three  components whose  values  are  higher  than  1.

As  the  two  first  describe  almost  80%   of  the  samples  or  nine  samples  from  12,

with  the  help  of  the  third  component, the  least  important,

almost  all  the  samples  fit  such  a  model.

The  first  component  comprises   both  micropores  parameters

[inaudible 00:09:39]   and  volume  of  the  meso pores.

The   other  variables  take  a  secondary  role in  the  second  and  third  components,

describing  together   another  half  of  the  temples.

Mapping  the  points   over  the  score  plots  in  3D,

it  is  easy  to  conclude  that  both  groups, carbonized  and  non  carbonized,

can  be  separated  into  two  clusters   defined  by  three  principal  components

and two  almost  flat  clusters  comprising  the  points

situated  on  the  plane are  set  by  two  main  clustering  algorithms

and described,  yes,  by  these  three  components.

Taking  a  closer  look  at  the  results of  predictor  screening  by  boosting,

again  microporosity  is  referred to  be  the  main  property.

The simplified  analysis,   two  variables  can  be  extracted  and  plotted

with  the  PCA  cluster, I mean using  the  same  PCA,

however,  with  completely  different  amounts,

one  instead  of three  components,  and  yes,

whereas  a  single  principal  component serves  to  describe  the  cluster  model.

The  cluster  formula  and  equation for  the  principal  component  are  provided.

It  can  be  recommended   for  future  classification

or  for  a  qualitative  analysis   of  synthesized  samples.

As  conclusions, I  can  say  that

the  presented  synthesis  method  makes  it  possible

to  obtain  mesoporous  carbon nanocomposites  uniformly  filled  with  metal

and  metal  oxide  phases  which  were pre-synthesized  in  silicon  matrix.

With  the carbonization,  the  portion  of  micro pores  is  growing,  yes,

the  specific  surface  area  increased

with  decreased  total  porosity, total  pore  volume.

High  order  hybrid carbon  oxide  nanocomposites

with large specific  surface  area.

The  controlled  size  distribution of  the  modifier,

which  is  important   from  the  clinical  point  of  view,

and  significantly  expands

the  scope  of  application of  the  synthesized  materials.

Parameters  of  textural  properties   are  effective  variable

helpful  to  identify classified  nanooxide  materials.

Data visualization  has  given  insights   to  select  adequate  approaches

to  analyze  the  experimental  data.

K-Means  Clustering, Self  Organized  Map,

and  Hierarchical C lustering have  proven  to  be  powerful  tools

for  classification   of  the  subject  materials

 by  actual  properties.

Principal  Component  Analysis  in  turn  had  reduced  the  set  of  variables

for  a  definite  and  simple  classification.

The  study  claims  two  cluster  models  described  by  three  or  even  one

principal  component  to  classify  carbonized and  carbon- free  hybrid  nano composites.

I'm  thankful  also  for  the  financial  support

for  Brazilian  agency.

I'm  very  thankful  to  my  colleague

David  Kirmayer   from  the  Hebrew  University  of  Jerusalem

and  one  of  the  co  authors  Maria  Galaburda,

who actually  synthesized  these  samples.

Thanks to P olish  Foundations and  International  Visegrad  Fund.

Thank  you  very  much  for  your  attention.

It  was  a  pleasure  for  me to  make  such  a  presentation.