Ž.
Thin Solid Films376200032᎐37
High temperature materials for thin-film thermocouples on
silicon wafers
Kenneth G.Kreider U,Greg Gillen
Chemical Science and Technology Laboratory,National Institute of Standards and Technology,Gaithersburg,MD20899-8363,USA
Received23November1999;received in revid form7July2000;accepted7July2000
Abstract
We have developed an instrumented calibration wafer for radiometric temperature measurements in rapid thermal processing Ž.
RTP tools for miconductor processing.The instrumented wafers have sputter deposited thin-film th
ermocouples to minimize物理试题
Ž.
the thermal disturbance of the wafer by the nsors.The National Institute of Standards and Technology NIST calibration wafer also employs platinum᎐palladium wire thermocouples to achieve a combined standard uncertainty of0.4ЊC in the temperature measurement of the thin-film thermocouple junction at900ЊC.The high temperatures of the wafer has required the development of new thin-film material systems.We have reported the results of our testing and characterization of sputtered platinum,palladium,rhodium,and iridium thinfilms using titanium bond coats on thermally oxidized silicon wafers.Depth profiling with condary ion mass spectrometry was ud to determine the diffusion profiles from the metalfilm to the silicon after heat treatments as high as1000ЊC.Electron microscopy and optical microscopy were ud to follow the reactions and the deterioration of the thermoelectricfilms.In addition,performance tests up to1000ЊC in the NIST RTP test bed were ud to determine the stability of the material systems.Failure mechanisms and limitations of the thin-film thermocouple materials have been discusd with data on hysteresis and drift in thermometry performance.The results of our evaluations indicated that Rh r Ir thin-film thermocouples have the best properties for wafer temperatures above900ЊC.ᮊ2000Elvier Scienc
e S.A.All rights rerved.
Keywords:Rapid thermal processing;Thinfilm thermocouples;Calibration;Rhodium;Palladium;Iridium;Platinum;Wafer
高中论语1.Introduction
Improved temperature measurement is a critical need
Ž.
in the rapid thermal processing RTP of electronic circuits.The RTP tool is normally monitored by radio-metric temperature measurements which are subject to uncertainties,due to variable emissivities and reflected radiation.The most effective way to reduce the uncer-tainties of the radiometric measurements is to cali-brate them on line in the RTP tool.Thermocouple instrumented wafers have been ud for making the
U Corresponding author.Tel.:q1-301-975-2619.
Ž.
E-mail address:kenneth.v K.G.Kreider.calibrations in RTP tools.The calibration wafers and some of the thermocouple materials have been de-
w x
scribed previously,1᎐3.At the National Institute of病毒感冒症状
吃什么祛痘Ž.
Standards and Technology NIST we are exploring the
w x u of thin-film thermocouples on the wafers2in order to minimize the thermal perturbation caud by the temperature nsor.The NIST calibration wafer includes Pt r Pd wire thermocouples which extend to the periphery of the wafer,allowing temperature mea-surement standard uncertainties of less than0.1ЊC at
w x
their measuring junction4,and we u thin-film dif-ferential thermocouples from the platinum r palladium Ž.
Pt r Pd junction to the wafer calibration location.With temperature variations of10ЊC or less across the wafer, the surface temperature of the wafer at the thin-film
0040-6090r00r$-e front matterᮊ2000Elvier Science S.A.All rights rerved.
Ž.
PII:S0040-60900001346-8
()
K.G.Kreider,G.Gillen r Thin Solid Films376200032᎐3733
thermocouple junction can be measured with a stan-
w x
dard uncertainty of0.4ЊC5.The uncertainties have been established with calibration tests of the Pt r Pd
Ž. wire thermocouples and rhodium᎐platinum Rh r Pt thin-film thermocouples on the International Te
mpera-
Ž.w x
ture Scale of1990ITS-906.
The Pt and Pd thinfilms that were evaluated previ-ously had limited stability at the higher temperatures军训过程
耻骨疼是什么原因Ž.
for RTP radiometer calibration900᎐1000ЊC.The work discusd in this paper was intended to clarify the caus of limitations of the Pd and Pt thinfilms and report on the evaluation of suitable thinfilms for higher temperature operations.In choosing the materi-als for thin-film nsors on silicon wafers,the key factors were the chemical and mechanical interactions between thefilms and the wafer and between the thin films and the environment.Since we needed a tempera-ture nsor which could be ud at temperatures up to 1000ЊC in a rapid thermal processing tool,the stability of the interfaces between thefilms and the wafer was most critical.At the high temperatures,thin-film thermocouples have advantages over thin-film resis-tance thermometers since they are not nsitive to mass changes which affect resistance.The key factor in
Ž
thermocouple stability and also necessary for resistor .仲大军
stability is the uniform composition of the electrically conducting pha.Stress levels and other metallurgical factors,such as grain size and defect population,have smaller effects on the Seebeck coefficient.Reactions with oxygen from the atmosphere or solid phas,com-pound formation with nearby solid constituents,or diffusion to and from nearby phas can all cau drift in the thermoelectric output and the Seebeck coeffi-cient of the thermoelements.Becau thinfilms have inherently short diffusion distances,the reactions cau instabilities at lower temperatures than the in-stabilities obrved with wire thermocouples.At1000ЊC, the u of alloy thermocouples is practically precluded becau of preferential reactions of their constituents. For example,thermocouples with thermoelements of platinum᎐rhodium alloys exhibit appreciable drift due
w x to preferential oxidation of rhodium at700᎐900ЊC7. For application in high temperature calibrations,the thin-film thermocouple is joined to a wire thermocou-ple for u up to1000ЊC.The best connection between them is a welded connection to survive a high tempera-ture rvice.Such welded junctions are more difficult with compounds such as the silicides and nitrides,which
w x have been ud for thin-film thermocouples8and metallization on silicon.The compounds form oxides on the surface which interfere with the thermoelectric circuit.Therefore,pure non-reactive elements appear to be the best choice for this application.Although gold and silver may be uful at low temperatures,their low melting points eliminate them from rious con-sideration at1000ЊC.Platinum,palladium,rhodium,and iridium appear to be logical choices due to their low reactivity and high temperature capability.The metals are also tolerant of partial pressures of10y5Ž.
atm1Pa of oxygen or water vapor which are com-monly prent in production environments.They en-able the construction of a more durable calibration wafer.We have also investigated the u of tungsten and rhenium thin-film thermocouples on the silicon wafers,but they have been found to be too nsitive to oxidation to be uful in this application.
2.Experimental procedure
Silicon wafers of200mm in diameter were obtained with a thermal oxide of300᎐320nm thickness.Some of the wafers were diced to form50=10-mm test coupons for u in thermal treatment and calibration experiments.They were cleaned with acetone and al-cohol,rind in deioni
zed water,and irradiated with 193-nm radiation for UV᎐ozone cleaning.Typically, the coupons were sputter coated in0.4Pa of99.999% Ar after a pump down to10y4Pa.Thefirst coat of material to bond to the silicon dioxide was a layer of 7᎐10nm thickness made from a titanium target of 99.995%purity.This bond coat was followed by a
Ž. coating of0.3᎐1.0-m thickness of platinum99.99%,Ž.Ž.
iridium99.9%,rhodium99.95%or palladium Ž.
99.97%.The array of thin-film thermocouples on cali-bration wafers included two measuring junctions at16 mm from the center of the wafer,11mm apart,and two measuring junctions50mm from the center,62 mm apart.The Pt r Pd wire thermocouples were welded to the thin-film weld pads10mm from the edge of the wafer.A more complete description of the calibration wafer and its u can be found in Kreider and deWitt w x2.
High temperature tests were conducted in controlled atmospheres using an alumina tube furnace,a fud
w x
silica tube calibration furnace5,and the NIST RTP w x
test bed2.Resistance measurements were made using a standard four-point resistance tester.Scanning elec-
Ž.
tron microscopy SEM was performed on a Hitachi S4000at a working distance of16mm with an Oxford 6566X-ray analyzer1.Secondary ion mass spectrometry Ž.
SIMS was performed on a CAMECA IMS4F using
O q bombardment at8.0keV impact energy,with the 2
detection of positive condary ions.Calibrations of thin-film thermocouple elements were conducted using
w x the NIST thin-film thermocouple calibration cell9. RTP cycling and calibration tests were performed in
1Identification of commercial equipment and materials in this paper does not imply recommendation or endorment by the Natio-nal Institute of Standards and Technology.
()K.G.Kreider,G.Gillen r Thin Solid Films 376200032᎐37
34w x the NIST RTP test bed 2.Adhesion tests were made using a Sebastian tab pull tester.3.Results
The thin-film thermocouple materials were evaluated Ž.using both long duration 2᎐20h exposure in a con-trolled atmosphere tube furnace and in the NIST RTP test bed which permits rapid thermal cycling.The tests were ud to determine whether any problems could be expected during applications as a calibration wafer for RTP.We expected problems such as the oxidation or reaction of the thin film with its atmo-sphere,reaction between the thin film and the silicon wafer,reactions with the Ti-bond coat,which can lead to loss of adhesion,and coalescence of the films due to lf-diffusion.One of the simplest ways of detecting changes in the film’s thermoelectric properties was to measure its resistance.A change in resistance could result from a change in resistivity,which would indicate a change in its thermoelectric properties or a loss of material to vaporization,or reactions which would not affect the thermoelectric properties.Changes such as recrystallization,oxidation,inclusion precipitation,porosity,and delamination are well detected by optical and electron microscopy.Interpha diffusion is best detected by depth profiling with SIMS and SEM X-ray compositional analysis.Since the adhesion of the films in this application is critical,we cho to test this property directly.We have also reported on lifetimes in the RTP tool and t
he stability of thermoelectric properties in a thin-film thermocouple calibration test.3.1.Palladium films
We have previously reported on the limitation of Pd w x thin-films 3.The most nsitive test related to the drift in emf at 880ЊC of 9%in 24h for a 0.7-m thick film,and a more rious drift at higher temperatures.Although RTP calibration applications would rarely require exposures that long,the practical limit for the Pd films is probably clo to 850ЊC for films less than l m thick.Fig.1shows an SEM image of a 0.7-m Pd Žfilm annealed at 900ЊC for 2h.This film shows large 3.m pores and excessive grain growth with faceting after recrystallization.SIMS depth profiling on the as-deposited 0.7-m thick Pd film clearly showed the boundaries Pd r Ti and Ti r SiO ,with enhancement of 2the signal where oxygen was prent as expected.After annealing for 2h at 700or at 800ЊC,the Ti signal was broadened significantly,but it remained centered around the interface.Ti has very little solubility in Pd w x Ž10and the broadening of all three signals Ti,Pd,and .Si may be related to pore formation,and surface
roughening.
Fig.1.Pd film on Si wafer,0.3-m thick,annealed for 2h at 900ЊC.Notice the pore growth and large facetted grains.Bar is 6m.
3.2.Platinum
Ž.Platinum has a higher melting point 1773ЊC than Ž.Pd 1550ЊC and does not oxidize in air above 450ЊC.It w x is ud in commercial wire thermocouples and Pt-675is the reference standard maintained by NIST.The Pt r Pd wire thermocouple has a Seebeck coefficient of approximately 18.2V r K at 900ЊC.The calibration of a sputtered 0.5-m Pt film on a Si wafer up to 950ЊC vs.a Pt wire,and its unexpected results of 1᎐2V r K emf w x has been reported previously 3.This test indicated Ž.very low hysteresis -1%on rapid cooling compared to the output on heating.The Pt films are also subject to coalescence.In addition,at 900ЊC and above,Ti w x becomes more soluble 10.Fig.2shows the SIMS depth profile of a 0.7-m thick Pt film with a Ti-bond coat on a Si wafer with 310nm of thermal oxide after a 2-h exposure in N at 900ЊC.The migration of Ti 2Ž.through the Pt film to the expod surface t s 0was evident.We noted the enhancement of both the Ti
and
Fig.2.SIMS depth profile of a 0.6-m thick Pt film with a Ti-bond coat after 2h at 900ЊC.Ti has migrated to the surface and oxidized.
()
K.G.Kreider,G.Gillen r Thin Solid Films376200032᎐37
35
Fig.3.SEM image of a0.4-m thick Ptfilm after1h at1000ЊC. Small pores are0.1᎐0.3m in diameter and thefilm has lost adhesion.Bar is100m.
Pt signals near the surface due to oxygen enhancement,
which was not evident in the as-deposited sample.This oxygen prence was probably due to TiO forming on
2
the exterior surface.The migration of Ti was also evident to a lesr extent on a sample annealed at 630ЊC.The diffusion of Ti to the surface also led to a
weakening of the bond between the Ptfilm and SiO.
2 The loss of Ti at the interface and the increa of the diffusion coefficients of Pt above900ЊC led to a coales-cence problem in Ptfilms.Fig.3shows an SEM image of a0.4-m thick Ptfilm held at1000ЊC in N for2h,
2 indicating vere pore growth,loss of adhesion,and surface roughening.The circular pores in Fig.3were 1᎐3m in diameter.We saw similar pores optically at 2000=in a0.3-m thick Ptfilm on a calibration wafer, which was thermally cycled44times for calibration runs.The last cycle reached1011ЊC and it caud the failure of the thin-film thermocouple.
3.3.Rhodium
Ž.Ž. Rhodium Rh has a higher melting point1966ЊC than Pt,and correspondingly,has lower diffusion co-efficients at1000ЊC.Rh is not like Pt or Pd which have a yield stress of near30MPa at room temperature,but like Ir,it has a yield stress more than10times that of
Ž9. Pt.The Young’s modulus of Rh f300=10Pa is
Ž
also twice that of Pt.Therefore,Rh and Ir which has
.
even higher moduli and yield stress will be stresd primarily in the elastic region on thermal cycling in contrast to Pd and Pt which deform plastically at the
temperatures.Rh O is stable in air up to approxi-
23
mately900ЊC and,therefore,we performed all testing Ž.
in an N O s2Pa atmosphere.
卤牛肉怎么炒好吃2
2Fig.4.SEM image of0.4-m thick Rhfilm after2h at1000ЊC. Recrystallized grains are facetted and have2-m wide pores.Bar is4m.
SEM analysis at10000=displayed a recrystalliza-tion and coarning of the0.4-m thick Rhfilm after2
Ž.
h at1000ЊC Fig.4.The recrystallization was also highly facetted and2-m wide pores were visible in Fig.4at the triple grain boundary interctions.X-Ray compositional analysis of thefilm after the1000ЊC
Ž
exposure revealed a small Si signal3%of the peak
.
height of the Rh signal from the pores,but no Ti was obrved.
SIMS depth profiling of the Rh thinfilm after2h at Ž.
1000ЊC Fig.5displayed a spreading of the Ti,Si,and Rh signals compared with the as-deposited signals.This indicated roughening of the surface and probably some interpha diffusion,but it had less movement of
Ti
Fig.5.SIMS depth profile of0.4-m thick Rhfilm after2h at 1000ЊC.
()K.G.Kreider,G.Gillen r Thin Solid Films 376200032᎐37
36Fig.6.EMF of Rh thin film vs.Pt as a function of temperature.
than with the Pt sample.Similar patterns,but to a lesr degree,were obrved for Rh films after heat treatments at 750and 900ЊC.
A calibration of the thermoelectric emf of a Rh film 0.4-m thick film vs.Pt is given in Fig.6,showing both heating and cooling curves which have less than 0.6%hysteresis between 875and 975ЊC.The indications are that the Rh films are very stable at the temperatures.Very little damage was apparent on the Rh films of the calibration wafer heated to 1011ЊC.The Rh films also Ž.had excellent adhesion )30MPa after annealing in N at 1000ЊC for 1h.The resistivity of Rh films was 2Ž.found to be 5⍀cm u s 1⍀cm,k s l after a 900ЊC anneal in N ,compared to the bulk value of 4.6⍀cm 2w x 11and the as-deposited value of 6⍀cm.3.4.Iridium
Iridium thin films bonded to the thermally oxidized Si wafer with Ti were very stable at 1000ЊC.Ir,as previously mentioned,has one of the highest elastic moduli,a very high yield stress,a high melting point Ž.2440ЊC ,and low lf-diffusion coefficients at 1000ЊC.With a thermal expansion coefficient mismatch of 3=10y 6K y 1compared to Si,the expected residual stress on cooling from 1000ЊC would be elastic ᎐compressive,as is true with Rh.Oxidation rates are slow in air up to near 1100ЊC where the vapor pha of IrO forms.We 3Ž.heat-treated the Ir films in N O s 2Pa and SIMS 22depth profiling indicated some Ti migration to the outer surface,but most of the Ti remained near the Ir r SiO interface after 2h at 1000ЊC.An SEM image 2of a 0.3-m thick Ir film is prented in Fig.7.The Ir recrystallized after 2h at 1000ЊC and had deep grain boundary grooving and facetting on crystal planes.The black spots,which were approximately 0.25m in
diameter,were the result of the early stages of lf-dif-fusion and coalescence.No Si or Ti signal was visible on the X-ray analysis,leading to the conclusion that the pits in Fig.7did not reach the Si,or that they were so small that the X-ray signal could not escape through the 0.3-m thick Ir film.The resistivity of the Ir films Ž.was 6⍀cm u s 1.5V r K,k s 1after a 900ЊC anneal in N compared to the as-deposited 11⍀cm 2w x and a handbook value for bulk Ir of 5⍀cm 11.The Ir thin film vs.Pt calibration up to 980ЊC,Fig.8,illustrates the excellent hysteresis of less than 0.1%in heating and cooling from 850to 980ЊC.The Seebeck coefficient of the film vs.Pt in this range averaged
14.5
Fig.7.SEM image of 0.3-m thick Ir film on wafer with 0.25-m diameter pits.Bar is 10m.
