by  the  way .

I 'm  not  sure  if  it 's  just  really  laggy   or  what 's  going  on  there .

You 're  not  seeing  the  title  slide ?

Now  I 'm  seeing  the  title  slide .

I 'm  going  to  go  ahead  and  unmute  myself .

Yeah ,  I  don 't  know  if  there 's  anything we  do  about  that  either .

That 's  got  to  be  like  a  Zoom or  potentially  an  Internet  thing .

Well ,  let 's  go  ahead  and  do  this

and  then  if  I  see that  things  look  really  janky ,

I  might  be  messaging  you .

But  I  think  for  the  most  part ,

if  you  just  maybe  take   a  quick  pause  between  slides ,

we 'll  be  okay .

I 'll  go  ahead  and  mute ,

and  then  you  can  go  ahead   and  get  started ,  Tom .

I 'm  really  glad  to  be  here  today   and  presenting  to  you .

I 'll  be  discussing  Chemical Mechanical  Planarization  or  CMP .

This  is  an  integral  step in  the  semiconductor  manufacturing  process

that  I  will  give  a  little  bit   of  background  on

as  I  move  through  my  slides .

Let 's  go  to  the  first  slide  of  the  presentation .

Quick  overview  of  CMP .

Once  again ,  that  stands  for   Chemical  Mechanical  Planarization ,

and  that  describes  the  two  processes  involved

with  planarizing  certain  layers   of  circuits  in  the  semiconductor  field .

It 's  crucial  to  planarize  each  layer  of  circuits

so  that  the  one  that  goes on  top  of  the  preceding  layer

will  not  have  any  shorts or  opens   or  short  circuits .

The  chemical  part  of  chemical  mechanical  planarization

is  when  we  introduce  a  liquid

to  react  with  the  top  atomic  layer of  the  material  that  we  are  planarizing .

That  reaction  softens  the  layer ,

whether  it 's  a  metal  or  an  oxide , whatever  the  case  may  be ,

we  choose  that  chemical  so  that   it  will  soften  the  top  atomic  layer .

Then  the  mechanical  part  comes

with  the  abrasives  that  are  carefully selected  for  that  polishing  slurry .

These  abrasives  then  remove that  softened  top  layer ,

exposing  the  next  atomic  layer   to  be  softened  by  the  chemical  reaction .

This  process  is  repeated atomic  layer  after  atomic  layer

until  the  desired  level  of  planarization is  reached .

Like  I  was  saying ,

this  planarization  is  crucial  to  do in  between  each  layer  of  circuitry

to  ensure  that  we  have   solid  electrical  isolation

without  any  shorts  or  opens .

This  process  enables  us  to  pile   many  layers  on  top  of  one  another

and  reduce  the  footprint of  whatever  chip  we 're  making .

When  we  do  the  CMP  processes ,

many  of  them  consist  of  a  two  polish  table process  to  enable  maximum  throughput .

Typically ,  the  first  table   consists  of  a  bulk  polish

where  most  of  a  substance   is  planarized  and  removed ,

and  then  a  final  polish   or  second  polish  table

where  we  do  the  final  touch -ups

and  ensure  that  we  have  good electrical  isolation  between  the  circuits .

Each  polish  table  has  a  polish  pad

where  we  press  the  wafer   containing  the  chips  against ,

and  this  enables the  polish  process  to  occur

and  it  enables  the  slurry  to  be  pressed against  the  wafer  that  contains  the  chips .

The  polish  pad ,  unfortunately ,

it  wears  down  as  each  wafer   is  polished  on  it ,

which  inhibits  the  transport  of  this chemical  slurry  to  the  wafer  surface .

Typically ,  as  the  polish  pad  wears ,   we  see  a  reduction  in  polish  rate ,

which  causes  longer and  longer  polish  times .

If  we  don 't  take   that  wear  factor  into  account

and  we 're  setting  up the  two -table  process ,

then  we  will  have  one  polish  table   begin  to  predominate ,

and  it  can  actually  develop such  a  long  polish  time

that  it  will  become  a  limiting  factor in  our  throughput .

For  my  presentation  today ,

I 'm  explaining  a  process  whereby   we  can  take  into  account  the  pad  wear  rate

and  still  ensure  that  we  have   optimized  throughput .

Approach  and  the  results .

I  gathered  and  analyzed  data  in  JMP

to  show  the  polish  time   versus  polish  pad  life  relationship

for  tables  1  and  2   for  a  CMP  two -table  process

at  one  of  Texas  Instruments Fabrication  Facilities .

I  then  used  JMP  scripting  language ,

the  built -in  integration  function to  calculate  the  area

under  the  polish  time   versus  pad  life  curve  for  both  tables .

Then ,  finally ,   I  used  the  system  of  equations

to  solve  for  a  Table  1 polish  time  starting  point

that  enabled  equal  polish  time   versus  polish  pad  life  integrated  areas

for  tables  1  and  2 .

This  ensured  that  we  had  achieved a  maximized  tool  output

for  the  CMP  process .

Conclusions  are  that  we  can  use  JMP

to  develop ,  first  of  all ,  polish  time   versus  polish  pad  life  curves .

Then ,  second ,  we  can  use  JMP   scripting  languages ,  built -in  integrator .

We  can  use  this  combination

as  an  efficient  way  to  optimize   CMP  two  table  processes  for  throughput .

Let 's  go  to  the  next  slide .

This  curve  that  we 're  looking  at  is what  we  get  when  we  plot  the  polish  time

on  the  first  table  versus the  polish  pad  life  of  that  first  table .

As  we  can  see , the  polish  time  as  the  pad  wears

and  we  polish  more  and  more  wafers   on  this  polish  pad ,

we  start  to  see  polish  times   increase  quite  dramatically .

Using  JMP ,  we  can  fit  a  line  to  this  data

and  develop  a  simple  Y  equals  Mx  plus  b   slope  intercept  function  to  fit  this  data .

I 'm  using  a  first  order  here ,

but  if  the  data  justified  it , we  could  use  a  second  or  third  order .

But  in  this  case , a  first  order  is  sufficient .

We  developed  the  equation for  polish  time  versus  pad  count ,

and  then  we  use  the  JSL  integrator to  integrate  the  area  under  this  line .

That 's  for  the  Polish  Table  1 .

Then  I 'll  talk  a  little  bit  more  about the  equations  needed  to  solve

for  the  optimal  settings .

But  first  of  all , we 'll  jump  to  Polish  Table  2 .

Polish  Table  2  is  the  final , it 's  the  finishing  table

where  we  finish  removing   whatever  the  bulk  material  is

we 're  using  to  isolate  or form  these  circuits ,  electrical  circuits .

The  same  thing  for  Polish  Table  2 ,

we  developed  the  curve   from  historical  data

of  polish  time  in  seconds   versus  the  polish  pad  life .

Typically ,  for  the  polishing  plat ,   for  the  final  plat  or  final  table ,

the  trend  with  polish  pad  life   is  not  as  pronounced  as  it  is

for  the  bulk  for  Table  1 .

Before  discussing  a  little  bit  more about  these  equations ,

just  to  make  sure  everyone 's   on  the  same  page ,

I  wanted  to  show  what  physically this  looks  like  on  the  wafer  surface

for  doing  this  two -table  polish  process .

Before  we  start  to  polish  anything ,

we  have  a  wafer   with  pattern  circuits  on  it .

Then ,  on  top  of  that  wafer ,

we 've  deposited  a  large  amount , relatively  speaking ,

of  whether  it 's  isolator  or electrically  conductive  material ,

whatever  is  needed  to  form the  circuits  that  we 've  etched  here .

After  Polish  Table  1 , also  known  as  the  bulk  polish  table ,

we 've  removed  a  significant  amount   of  this  bulk  material

and  we 've  planarized  it , made  it  very  flat .

Then  after  Polish  Table  2 ,

which  is  also  called   the  clear  or  the  finish  table ,

we 've  removed  all  of  this  bulk  material

except  where  it 's  actually  needed   in  the  pattern  circuits .

Physically ,  that 's  what 's  happening in  this  two  polish  table  process .

Let 's  go  back  quickly  to  the  equations .

We  integrate  the  area  under  Table  1 ,   which  is  on  the  preceding  slide .

That 's  the  bulk  polish  time   versus  Table  1  polish  pad  count .

Then  we  also  integrate   the  area  under  this  curve ,

which  is  the  final  or  finish  table .

Same  thing ,  integrate  the  area  under the  polish  time  versus  pad  count  line .

Then  we  solve  the  linear  system of  equations  to  find  which  Y -intercept ,

in  this  case ,  it 's  b1 , and  that 's  for  the  Table  1 .

We  solve  for  that  b1  that  will  create an  equal  integrated  area

under  the  Table  1  and  the  Table  2  curves .

Just  as  a  reminder ,

the  JMP  scripting  language has  a  handy  function

for  enabling  quick  integration

and  determining  the  areas   under  these  two  curves .

Then  based  on  a  few  other  properties

that  we  know  based  on  historical   or  baseline  analysis ,

then  we 're  able  to  set  up   a  system  of  equations

with  four  unknowns ,  four  equations ,

and  then  we  solve  for  that  intercept on  the  first  curve

that  enables  maximum  throughput in  equal  area  under  these  curves .

With  that ,

almost  the  entire  process   can  be  done  in  JMP ,

with  the  exception  of  solving a  system  of  equations

that  can  indirectly  be  done  in  JMP ,   or  you  can  write  a  script .

But  I 've  found  it  easier  just  to  solve the  system  of  equations  manually ,

but  it  could  easily  be  set  up   to  be  done  in  JMP  as  well .

When  you  finish  this  process ,

then  you 're  able  to  optimize   this  two -table  CMP  system

and  you  can  maximize  the  throughput

and  so  minimize  the  number   of  polish  tools  required ,

which  saves  a  lot  of  money   and  saves  a  lot  of  time .

That  is  the  end  of  my  presentation .

Thank  you .