book of abstract 2002
TRANSCRIPT
-
7/30/2019 Book of Abstract 2002
1/38
1
2L
2W
nonwetting
material
weting
material
2002 ME Graduate Student Conf erenceAp ri l 13, 2002
HEAT TRANSFER AND FLUID FLOW IN AN IDEALIZED MICRO HEAT PIPE
J in ZhangM.S. Candidate
Thesis advisor: Prof. Harris Wong
ABSTRACTMicro heat pipes have been used as a heat-dissipating
device in many systems, such as micro electroniccomponents and the leading edge of hypersonic aircraft.
[1, 2]
Micro heat pipes transfer heat by evaporation, convection,and condensation, same as the conventional heat pipes.
However, the effective thermal conductivity of micro heatpipes is only 1/40 that of conventional heat pipes. Due tothe complexity of the coupled heat and mass transport, andto the complicated three-dimensional bubble geometryinside micro heat pipes, there is a lack of rigorous analysis.As a result, the relative low effective thermal conductivityremains unexplained. This work conceptualized an idealizedmicro heat pipe that eliminates the complicated geometry,but retains the essential physics. The simplified bubblegeometry allows a direct comparison between theoreticalpredictions and experimental data.
The idealized micro heat pipe is a rectangular heat pipe,the top portion of which is made of a non-wetting material,and the bottom portion a wetting material. The lower
portion is filled to the rim by a wetting liquid, and theupper portion is filled by its vapor. This configurationensures that the contact line of the liquid-vapor interface ispinned at the interception between the wetting and non-wetting materials. Pinning of the interface allows a capillarypressure gradient to drive the liquid flow. When this microheat pipe is driven at small temperature differences, theinterface should be roughly flat, allowing the analysis to begreatly simplified.
The evaporation and condensation in the idealizedmicro heat pipe is analyzed. It is found that the interface canbe separated into two regions: an inner region near the wallwhere evaporation occurs and an outer region away from thewall. The evaporation rate is solved by the method of
matched asymptotic expansions, and the leading orderevaporation rate is obtained as ln, where measures theratio of conductive heat flux in the liquid to evaporativeheat flux at the interface. The small parameter is
where kfis the liquid thermal conductivity, T is the walltemperature, c=(2RT)
-1/2with R being the universal gas
constant per unit mass of the vapor as determined by the
kinetic theory, is vapor density, hfg is liquid latent heatand W is the half width of the pipe. The Stokes flowinduced by the surface tension gradient (Marangoni stress)along the interface and by the evaporation at the interface issolved using a finite-difference method.
Fluid flowand heat transfer along the micro heat pipe
are also studied. Liquid temperature distribution along themicro heat pipe is given in Figure 2. It is found that thetemperature profile is relatively flat except the region nearthe evaporator, which is the evaporation region. The lengthscale of the region is calculated as
where Aw and A fare cross-sectional area of the wall andliquid, respectively, and kw is wall thermal conductivity.For a micro heat pipe with larger L/W the length of theevaporation region is shorter. Vapor pressure distributionalong the micro heat pipe is also given in Figure 3. It isclear that the pressure goes approximately linearly and notaffected strongly by L/W. Effective thermal conductivity keffis evaluated. A longer or wider micro heat pipe will have alarger keff. And it is found that increasing the evaporationarea at the evaporator will increase keff. It is also affected bythe latent heat of the working fluid. A fluid with largerlatent heat will produce larger keff.
FIGURES AND TABLES
Figure 1 an Idealized Micro Heat Pipe
= Awkw + A fkf2 kf - ln
,
= kfT
chfg2 W,
-
7/30/2019 Book of Abstract 2002
2/38
2
ACKNOWLEDGMENTSThis work was supported by NASA and LaSPACE.
REFERENCES
1. G. P. Peterson, Appl. Mech. Rev. 45 (1992) 175-89.2. P. Dunn & D. A . Reay, in "Heat Pipes" (Pergamon
Press, 1982) 16-18.
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
L/W = 20L/W = 50L/W = 100L/W = 200
p*
z*
p
z
0
50
100
150
200
250
0.0002 0.0004 0.0006 0.0008 0.001
L/W = 20
L/W = 50L/W = 100L/W = 200
keff
W
W
keff
Figure 4 Effective Thermal Conductivity along anIdealized Micro Heat Pipe
Figure 3 Pressure Distribution along anIdealized Micro Heat Pipe
Figure 2 Temperature Distribution along anIdealized Micro Heat Pipe
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
L/W = 20
L/W = 50
L/W = 100
L/W =200
T*
z*
T
z
-
7/30/2019 Book of Abstract 2002
3/38
3
2002 ME Graduate Student Conference
April 13, 2002
PHYSICAL PARAMETER ESTIMATION OF VIBRATING STRUCTURE FROM ITS
SPECTRAL DATA: A NEW MATHEMATICAL MODEL
Kumar Vikram Singh
Ph.D. Candidate
Faculty Advisor: Y. M. Ram
ABSTRACT
The problem of reconstructing a model with prescribed
spectral data is known as inverse eigenvalue problem.
Reconstruction of the distribution of physical parameters of
a continuous vibratory system by using its eigenvalues isaddressed here. Considering a unit length piecewise
continuous rod as shown in figure 1. The eigenvalue
problem associated with this rod is given by the following
set of differential equations
======+==+
)()()()(
0)(0)0(where,,0
where,,02
2
avauavau
Lvusvvr
quup
. (1)
Applying the boundary and matching conditions of
displacement and force leads to problem of finding the non-
trivial solution of
o=
4
3
2
sincos0
sincoscos
cossinsin
z
z
z
LL
aaa
aaa
. (2)
We named this problem the Transcendental Eigenvalue
Problem (TEP).The general form of this problem is
( ) ozA = . (3)
Frequently the classical finite element and finite difference
formulation are used in approximating such a continuous
system. The characteristic equation of the obtained
eigenvalue problem is a polynomial. In contrast, the
continuous systems are characterized by TEP [1]. By using
finite element or finite difference method, the TEP is
transformed into an algebraic eigenvalue problem. It has
been concluded by [2,3] that the solution to the discrete
problem is not a good approximation to the continuous one.
Past research associated with inverse problems of the
continuous vibratory system can be found in [4,5,6,7,8].Since the behavior of a finite dimensional polynomial is
fundamentally different from the transcendental function,
such an approach may involve inaccurate approximation of
the physical parameters, as illustrated in figure 2.
For the given continuous system in figure 1, the inverse
problem can be defined as follows:
Given the resonant frequencies21
, , anti-resonant
frequency1 and the total mass of the rod.
Determine the physical parameters21
,pp and21
,qq .
The problem now is of determining the roots of the system
of transcendental frequency equations,
( )( )( )( )( )( )
==
==
==
=
=
=
0,,det),,(
0,,det),,(
0,,det),,(
1
2
1
3
2
1
A
A
A
F
F
F
, (4)
for the given values of21
, and1 . The research aims at
developing low dimensional analytical models allowing
estimation of the physical parameters of the structures from
measured vibration test data. The main idea presented hereis to replace the continuous system with variable physical
parameters by a continuous system with piecewise uniform
properties as shown in figure 3. The boundary and matching
conditions between the various parts of the continuous
model can be expressed in the TEP form. A rapidly
converged algorithm is used for evaluation of the physical
parameters of the system. The algorithm implements the
Newtons iterative method for determining the physical
-
7/30/2019 Book of Abstract 2002
4/38
4
parameters of the system. Formulation of such mathematical
models for non-uniform axially vibrating rods and
reconstruction of their area distribution by using this
algorithm, as illustrated in figure 4, is presented. This
proposed solution of TEP can also be used to solve classical
direct problems in structural dynamics such as buckling [9]
and vibration control [10].
FIGURES AND TABLES
Fig.1. Piecewise continuous axially vibrating rod
Fig.2.Physical parameter Identification of piecewise rod
from its associated discrete model
Fig.3. New mathematical model used for the approximation
of a non-uniform rod
Fig.4. Reconstruction of the shape of the exponential rod
with model ordern=4 and n=8
ACKNOWLEDGMENTS
The research work presented here is supported by a
National Science Foundation research grant CMS-9978786.
REFERENCES
1. Singh K. V. and Ram Y. M., A mathematical model to
overcome the discrepancies between continuous
systems and their discrete approximation, ASME ETCE
2002.
2. Boley D. and Golub G.H., A Survey of matrix eigenvalue
problems, Inverse Problems, Vol. 3, pp. 595-622, 1987.
3. Paine J., A numerical method for the inverse Sturm-
Liouville problem, SIAM Journal on Scientific and
Statistical Computing, Vol. 5(1), pp. 149-156, 1984.
4. Ram Y.M. and Caldwell J., Physical parameters
reconstruction of a free-free mass-spring system from
its spectra, SIAM Journal of Applied Mathematics, Vol.
52(1), pp. 140-152, 1992.
5. Frieland S., Nocedal J. and Overton M.L., The
formulation and analysis of numerical methods for
inverse eigenvalue problems, SIAM Journal of
Numerical Analysis, Vol. 244, pp. 634-667, 1987.
6. Ram Y.M. and Elhay S., Constructing the shape of a
rod from eigenvalues, Communications in Numerical
methods in Engineering, Vol. 14, pp. 597-608, 1998.
7. Gladwell G.M.L., Inverse problems in vibration,
Applied Mechanics Review, Vol. 39, pp. 1013-1018, 1986.
8. Gladwell G.M.L., Inverse Problem in vibration, Martin
Nijhoff publishers, First Edition, 1986.9. Singh K.V. and Ram Y.M., The Transcendental
Eigenvalue Problem and Its Applications, Accepted for
publication in AIAA Journal, 2002.
10. Singh K.V. and Ram Y.M., Dynamic Absorption by
Passive and Active Control, ASME Journal of vibration
and acoustics, Vol. 122(4), pp. 429-433, 2000.
1
1
n
n
q
p
1
1
q
p
2
2
q
p
3
3
q
p
L
L
n
n
q
p
L
L
L
L
1x2
x3
x
1nx
nxnu
1nu
2nu
3u
1u2u
h h
L
1
1
n
n
q
p
1
1
q
p
2
2
q
p
3
3
q
p
L
L
n
n
q
p
L
L
L
L
1x2
x3
x
1nx
nxnu
1nu
2nu
3u
1u2u
h h
L
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
n=4 n=8
Original shape Estimated
0 0.2 0.4 0.6 0.8 10 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 10 0.2 0.4 0.6 0.8 1
n=4 n=8
Original shape Estimated
L
222,, AE
111,, AE
a
-
7/30/2019 Book of Abstract 2002
5/38
5
2002 ME Graduate Student Conference
April 13, 2002
EFFECT OF WALL THICKNESS OF CENOSPHERES ON THE COMPRESSIVE
PROPERTIES OF SYNTACTIC FOAMS
Nikhil Gupta
Ph.D. Candidate
Faculty Advisor: Dr. Eyassu Woldesenbet
ABSTRACT
Cenospheres are incorporated in polymeric materials to
obtain composites of low density and high compressive
strength, known as syntactic foams. Some studies on themodeling and experimental behavior of such composites are
available in the literature [1-14]. However, no comprehensive
studies could be found which characterize the behavior of
syntactic foams with respect to various parameters like
cenosphere wall thickness (density) and size distribution.
This experimental work investigates the effect of wall
thickness of cenospheres on the compressive properties of
syntactic foams. As the matrix material D.E.R. 331, a di-epoxy
resin, manufactured by DOW Chemical Company was
selected. This resin is called diglycidyl ether of bisphenol A
(DGEBA). To lower the viscosity of the resin a diluent is
added to it. It is difficult to mix large volume of cenospheres
in the epoxy resin if the viscosity is very high. Adding 5% ofdiluent, C12-C14 aliphatic glycidyl ether, brings down the
viscosity of the resin from about 9000 cps at 20C to about
2000 cps at the same temperature. Average equivalent
epoxide weight (EEW) of the diluent is 285. For a 95 wt%
resin and 5 wt% diluent mixture the EEW is 177.5.
Triethylene tetramine (TETA), C6H18N4, is used as curing
agent. This chemical is commercially known as D.E.H. 24 and
manufactured by DOW Chemical Company. Molecular
weight of this hardener is 146.4 and weight per active
hydrogen is 24.4. Phr (parts per hundred parts of resin) of
amine hardener for 95-5 resin-diluent mix was calculated to
be 13.74. Stainless steel molds having inner dimensions of
990.5 in3 are used to cast the syntactic foams. Fourdifferent types of cenospheres were used for the fabrication
of syntactic foam specimens. These microballoons were
manufactured and supplied by 3M. Specimen size used for
the test was 12.712.725.4 mm3. Compression test wasconducted at constant crosshead movement rate of 0.5
mm/min. Minimums of five specimens of each type of
syntactic foam were tested. Stress-strain cures for each type
of specimens are presented here [Figs. 2-5]. Trends of peak
stress [Fig. 6] and modulus [Fig. 7] of the specimens with
respect to cenosphere density are also presented.
FIGURES AND TABLESTable 1. Particle size distribution of microballoons.
Ceno-
sphere
type
Average
Particle Size
(m)
Top Size
(m)Cenosphere
density
(g/cc)
Syntactic
foam density
(kg/m3)
S22 35 75 205 493
S32 40 80 320 545
K37 40 85 380 575
K46 40 80 460 650
Fig.1. Compressive fracture behavior of syntactic foam.
S22, ASTM D695
0
5
10
15
20
25
30
35
0 0.05 0.1 0.15Strain (mm/mm)
Stress(N
)
Fig. 2. Compression test results of S22 syntactic foam.
-
7/30/2019 Book of Abstract 2002
6/38
6
S 3 2 , A S T M D 6 9 5
0
10
20
30
40
50
0 0.05 0.1 0.15
Strain (mm/mm)
S
tress(N)
Fig. 3. Compression test results of S32 syntactic foam.
K 37, ASTM D695
0
10
20
30
40
50
60
0 0.02 0.04 0.06 0.08 0.1Strain (mm/mm)
Stress(MPa)
Fig. 4. Compression test results of K37 syntactic foam.
K 46 , ASTM D695
0
20
40
60
80
0 0.02 0.04 0.06 0.08Strain (mm/mm)
Stre
ss(N)
Fig. 5. Compression test results of K46 syntactic foam.
Change in Peak Compressive Strength
6000
7000
8000
9000
10000
11000
12000
200 300 400 500
Cenosphere Density (kg/m 3)
Peak
Stress
(M
Pa)
Fig. 6. Dependence of Peak strength of syntactic foam on
cenosphere density.
Cenosphere Density Dependence of
Modulus
1000
1200
1400
1600
1800
2000
2200
200 250 300 350 400 450 500Density of Cenospheres (Kg/mm3)
Modulus(MPa)
Fig. 7. Dependence of Modulus of syntactic foam on
cenosphere density.
REFERENCES
1. Gupta, N., Kishore, Woldesenbet, E., Sankaran, S.; J.
Mater. Sci., 36, 18 (2001) 4485-4491.
2. Gupta, N., Brar, B. S., Woldesenbet, E.; Bull. Mater Sci.,24, 2 (2001) 219-223.
3. Gupta, N., Kishore, Sankaran, S.; J. Reinf. Plast. Compo.
18, 14 (1999) 1347-1357.
4. Gupta, N., Karthikeyan, C. S., Sankaran, S., Kishore;
Mater. Charact., 43, 4 (1999) 271-277.
5. Gupta, N., Woldesenbet, E., Kishore, Sankaran, S.; J.
Sand. Str. and Mater., accepted, in press.
6. Gupta, N., Woldesenbet, E., Kishore; J. Mater. Sci.,
accepted, in press.
7. Gupta, N., Woldesenbet, E.; in proceedings of ETCE-
2002, Feb 2002, Houston, TX.
8. Gupta, N., Woldesenbet, E.; (accepted) in proceedings of
10th US-Japan Conference on Composite Materials,
September 16-18, 2002, Stanford University, CA.
9. Gupta, N., Woldesenbet, E.; in proceedings of ASC 16th
Annual Conference, Blacksburg, VA, Sept. 9-12, 2001.
10. Gupta, N., Woldesenbet, E.; in proceedings of ETCE-
2001, Feb 2001, Houston, TX.
11. Rizzi, E., E. Papa and A. Corigliano, Intl. J. Solids and
Struct. 37 (2000) 5773.
12. Corigliano, A., E. Rizzi and E. Papa, Compo. Sci. Tech. 60
(2000) 2169.
13. Ishai, O., C. Hiel and M. Luft, Composites 26, 1 (1995) 47.
14. Malloy, R. A., J. A. Hudson, in Intl. Encylop. of Compos.,
Ed. S. M. Lee (VCH, 1990).
-
7/30/2019 Book of Abstract 2002
7/38
7
2002 ME Graduate Student Conference
April 13, 2002
EFFECT OF TIP GEOMETRY ON BLADE TIP FLOW AND HEAT TRANSFER
David Kontrovitz
M.S. Candidate
Faculty Advisor: Dr. Srinath Ekkad
ABSTRACT
In an attempt to increase thrust to weight ratio and
efficiency of modern gas turbines, engine designers are
always interested in increasing turbine operating
temperatures. The benefits are attributed to the fact that
higher temperature gases yield a higher energy potential.
However, the detrimental effects on the components along
the hot gas path can offset the benefits of increasing the
operating temperature. The HPT first stage blade is one
component that is extremely vulnerable to the hot gas.
The cause for tip failures are fairly well understood and
can be explained as follows. A clearance gap between the
rotating blade tip and stationary shroud is necessary to
allow for the blades mechanical and thermal growth during
operation. Unfortunately, the gap allows for leakage flow
from the pressure side to the suction side. The gas is
accelerated as it passes through the small gap. This leads to
enhanced heat load to the blade tip region. Leakage flow, orclearance flow, also leads to undesirable aerodynamic losses
not unlike the losses associated with airplane wing tips. In
fact, one third of the losses through the turbine section can
be attributed to leakage flow. Other relevant studies by Azad
et al. [1-2], Bunker et al. [3], Bunker and Bailey [4-5]
presented detailed heat transfer results on high-pressure
turbine blade tips with different pressure ratios. The effect of
tip geometry was also considered in some of these studies.
The present study explores the effects of gap height
and squealer depth on heat transfer and flow distribution.
This investigation differs from those in the other studies
because the tip profile is for an in-service High Pressure
Turbine of an aircraft engine. Other experiments have usedthe E
3test blade or a power generation blade that have
different characteristics. The pressure ratio used was 1.2,
which is lower than the actual pressure ratio this blade sees
in service (PR = 1.7). A transient liquid crystal technique
was used to obtain the tip heat transfer distributions.
Pressure measurements were made on the blade surface and
on the shroud for different tip geometries and tip gaps.
FIGURES AND TABLES
Figure 1. Leakage Flow
Figure 1 shows the typical leakage flow behavior for a
turbine blade [1]. The plain tip is a flat tip and flow leaksthrough a constant area across the blade. The squealer tip
has a groove cut on top of the blade which increases the
area and stalls the flow thus creating back pressure and
restricts leakage flow and reduces heat transfer. In this
study, we focus on the plain tip and two different squealer
depths (shallow and deep).
-
7/30/2019 Book of Abstract 2002
8/38
8
Figure 2: Blade Pressure Distribution
Figure 2 shows the pressure distribution on the bladesurface at different span of the blade. The 100% span is on
the tip of the blade with clearance gap. The pressure
distribution changes as the blade span moves towards the
gap as expected. Figure 3 shows the pressure distributions
on the shroud for a blade with squealer tip. The reduced
static pressures are the cause of reduced leakage flows.
Figure 3: Shroud Measurements
Figure 4 presents a typical heat transfer distribution on
the blade tip with a shallow squealer. The heat transfer
distributions show the local hot spots near the leading edge
on the floor of the cavity and the reduced heat transfer
towards the trailing edge of the blade. The leakage flow is
stronger at the leading edge and weaker along the trailing
edge of the blade.
Figure 4: Heat Transfer Distribution
ACKNOWLEDGMENTS
This study was sponsored by the NSF through a
GOALI grant. The author would like to acknowledge Drs.Srinath Ekkad and Sumanta Acharya for their instruction and
advice. Also, thanks to Dr. Ron Bunker, at G.E. Corporate
Research and Development, for his input to this project.
Acknowledgments are also due to my colleagues in the
Turbine Blade Research Lab.
REFERENCES
1. G.S. Azad, J.C. Han, R.S. Bunker, C.P. Lee, Effect of
Squealer Geometry Arrangement on Gas Turbine Tip
Heat Transfer, in Proceedings of the ASME International
Mechanical Engineering Congress and Exposition, New
York, November 2001, HTD-243142. G.S. Azad, J.C. Han, R.J. Boyle,Heat Transfer and Flow on
the Squealer Tip of a Gas Turbine Blade, in Proceedings
of the 2000 ASME Turboexpo, Munich, Germany, May
2000, 2000-GT-1955
3. R.S. Bunker, J.C. Bailey, A.A. Ameri, Heat Transfer and
Flow on the First Stage Blade Tip of a Power
Generation Gas Turbine Part1: Experimental Results, in
Proceedings of the 1999 ASME International Gas Turbine
Conference, Indianapolis, 1999, ASME 99-GT-169.
4. R.S. Bunker, J.C. Bailey, Effect of Squealer Cavity Depth
and Oxidation on Turbine Blade Tip Heat Transfer, in
Proceedings of the 2001 ASME International Gas Turbine
Conference, New Orleans, June 2001.
5. R.S. Bunker, J.C. Bailey, An Experimental Study of
Heat Transfer and Flow on a Gas Turbine Blade Tip
With Various Tip Leakage Sealing Methods, in
Proceedings of the 4th ISHMT / ASME Heat and Mass
Transfer Conference, Pune, India, 2000, HNT-2000-055 p.
411-416.
-
7/30/2019 Book of Abstract 2002
9/38
9
2002 ME Graduate Student Conference
April 13, 2002
LARGE EDDY SIMULATIONS OF ROTATING SQUARE DUCT WITH NORMAL RIB
TURBULATORS
Mayank Tyagi
Ph.D. Candidate
Faculty Advisor: Dr. Sumanta Acharya
ABSTRACT
The goal of the present work is to develop a generalized
Large Eddy Simulations (LES) methodology for complex
geometries, and to apply this methodology to gas turbineblade cooling applications. Results from one such
application, that of internal cooling in a turbine blade, are
reported in this abstract.
The internal cooling configuration selected in this study
corresponds to the experimental study of Wagner et al.
(1992). The computations were performed at a Reynolds
number (Re) of 12,500, rotation number (Ro) of 0.12 and the
inlet coolant-to-wall density ratio (/) of 0.13. The ribheight-to-hydraulic diameter ratio (e/D) is 0.1 and the rib
pitch-to-height ratio (P/e) is 10. The ribs are square in cross-
section and are placed transverse to the flow in the duct. Of
specific interest in this study are the dynamics of the
coherent structures, and how they influence the heattransfer.
RibA
RibB
Trailin
g Wall
Leadi
ngWa
ll
BackWall
Front Wall
Inflow
Rotation
Lz
Ly
Lx
RibHeight: eRibX-section : SquareL
x= L
y= L
z= L
e/L= 0.1,Rmean
= 49.5P/L= 1.0Re
m=12500
Ro= 0.12
P/2
P /4
/= 0.13
Figure 1: Schematic of the computational domain
A novel direct method for computing the source term in
the energy equation (which arises due to unsteadiness and
uniform wall temperature in periodic geometries) is
presented. This is in contrast to the iterative approach of
Wang and Vanka, 1995, which has been used to date.
The details of flow field and the temperature fieldobtained from the LES are presented and analyzed in this
study. The LES procedure is based on a dynamic mixed
model for subgrid stresses and scalar flux.
RESULTSRib A
Rib B
Rib A
Rib B
Time-averaged VelocityVectors at Y/D= 0.5
Figure 2: Details of time-averaged vectors near the ribs
Figure 2 shows the time-averaged flow field at the spanwise
center-plane, and reveals the recirculating eddies generated
in the vicinity of the ribs. The coherent structures (Figure 3)are educed from the instantaneous flow field using the
positive levels of Laplacian of pressure field. Near the
trailing wall, the Coriolis -induced secondary flow lead to a
breakup of the flow structures along the spanwise center
plane. The roller vortices formed on the rib are connected
with the streamwise or braid vortices originating either at the
side walls or around the centerplane, thus forming a hairpin
coherent structure. The unsteady dynamics show several
-
7/30/2019 Book of Abstract 2002
10/38
10
packets of such hairpin vortices between the ribs at trailing
wall. The vortices near the leading wall are weaker than the
vortices near trailing wall. Moreover, the roller vortices
generated at the ribs are stretched and tilted into the core of
duct by the secondary flow. Hairpin packets on these ribbed
walls evolve in different fashions.
X
ZY
Leading wall
Trailing wall
X
Y
Z
Leading wall
Trailing wall
Figure 3: Coherent structures in a rotating ribbed duct
The proper orthogonal decomposition (POD) of the flow-
field using the method of snapshots is performed (Sirovich,
1987). About 99% of the turbulent energy is captured in the
first 75 POD modes (Figure 4). Clearly, the chaotic dynamics
of these coherent structures can be modeled by a low-
dimensional system.
POD mode number
EnergycontainedinthePODmod
e
50 100 150 200
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
Figure 4: Distribution of energy in POD modes based on
200 snapshots.
Heat transfer calculations show the unsteady hot streaks on
the duct walls. The distribution of the time-averaged Nusselt
number is in agreement with the experimental observations.
It is illustrated that scalar mixing is related to the scalar
dissipation of the temperature.
ACKNOWLEDGMENTS
Discussion with Dr. A.K. Saha on the preliminary resultsis gratefully acknowledged.
REFERENCES1. Sirovich, L. (1987) Turbulence and the dynamics of
coherent s tructures, Part I-III, Quarterly of Appl. Math.,
XLV(3), pp. 561-82.
2. Tyagi, M., Saha, A.K. and Acharya, S. (2001) Large eddy
simulations of rotating square duct with normal rib
turbulators, DNS/LES: Progress and Challenges, 3rd
AFOSR Intl. Conference, Eds. C. Liu, L. Sakell and T.
Beutner, pp. 807-814.
3. Wagner, J.H., Johnson, B.V., Graziani, R.A. and Yeh, F.C.
(1992) Heat transfer in rotating serpentine passages
with trips normal to the flow, J. Turbomachinery, Vol.
114, pp. 847-857.
4. Wang, G. and Vanka, S.P. (1995) Convective heat transfer
in periodic wavy passages, Int. J. Heat Mass Transfer,
Vol. 38, pp. 3219-3230.
-
7/30/2019 Book of Abstract 2002
11/38
1
2002 ME Graduate Student Conference
April 13, 2002
NODAL CONTROL OF A VIBRATING BEAM
Akshay N Singh
Ph.D. Candidate
Faculty Advisors: Dr. Y. M. Ram and Dr. Su-Seng Pang
ABSTRACT
Vibration control is an important engineering problem
and many methods for both active and passive vibration
absorption have been developed. This paper deals with
developing a method to achieve nodal control at the point of
excitation in a Bernoulli-Euler beam. Singh and Ram in [3]
have shown that under certain conditions that have been
characterized in [3] the steady state motion of a certain
degree of freedom in a harmonically excited conservative
system may be absorbed by both passive and active means.
Ram in [1] has developed a method to eliminate the steady
state motion of a prescribed location in a continuous system
like rod under the influence of a harmonic excitation. He has
presented a closed form solution for the control gain in
terms of infinite product of eigenvalues. This thesis extends
the approach in [1] to achieve nodal control for suppressing
vibration at prescribed location in beams and provides a
simpler formula for the control gain in terms ofeigenfunctions. It is established that, for a uniform
Bernoulli-Euler beam, the steady state motion at the point of
excitation can be absorbed by means of a control force
determined from displacement information at the point of
application. A closed form solution for the control gain is
presented and a criterion for implementing the control by
active and passive means is developed. The result for the
control gain is generalized for the case of a non-uniform
beam. It is also shown through some examples that the
theory can be also applied to eliminate the steady state
motion at any desired location other than the point of
excitation. Analysis is also performed to determine the
optimal control force and investigate the stability of theoverall system. Several controllability graphs are shown and
meaningful conclusions are drawn from these graphs. An
experiment is designed to validate the proposed theory and
display its practicality.
The developed theory will provide a strong foundation
for realizing realistic and convenient methodologies in
control applications in cases like surgical procedures,
drilling and turning operations etc. However, one of the
many direct applications of this method is structural
vibration control in an aircraft wing. Several measurements
such as vibrational response, air temperature, wind velocity
etc are required in order to monitor flight conditions in an
aircraft. These data also assist the pilot in flying the aircraft.
Sensors and data collection circuitry form an entire network
of the electrical wiring all in and around the airplane body.
Data acquisition devices are also located on the wing of the
airplane. Shielding of these devices from undesirable
vibration of the wing is critical in order to avoid noise in the
gathered data and prevent physical damage. Exclusion of
steady state vibration at the locations of these devices
provides the motivation for this investigation.
Consider a uniform Bernoulli-Euler beam of length L .
Suppose that the beam is excited by a harmonic force
( ) ttf cos= , as shown in Figure 1(a). The steady statemotion of a prescribed point of the beam may be vanished
by applying a concentrated control force ),( tau at ax= asshown in Figure 1(b).
( ) ( )tawtau ,, = , (1)
where is the control gain. The partial differential equation
for the controlled Bernoulli-Euler beam shown in Figure 1(b),
for Lx
-
7/30/2019 Book of Abstract 2002
12/38
2
( )0
,2
2
=
x
tLw, no moment, (5)
( )t
x
tLwEI
xcos
,2
2
=
shear force. (6)
The work here focuses on determining a closed form
solution for the control gain that absorbs the motion of
the beam at Lx = .
A closed form expression for the control gain
obtained from mathematical manipulation is expressed as
( )( ) ( )
( )
=
av
avavEIa
1
12, , (7)
where 1v and 2v are the deflections of the beams with spanax
-
7/30/2019 Book of Abstract 2002
13/38
13
2002 ME Graduate Student Conference
April 13, 2002
ON THERMALLY INDUCED SEIZURES (TIS) IN JOURNAL BEARINGS
Rajesh Krithivasan
M.S. Candidate
Faculty Advisor : Dr. Michael M Khonsari
ABSTRACT
Thermally induced seizure (TIS) in journal bearings is a
mode of failure that can occur quite suddenly and end up
with a catastrophic damage to the system. A failure, as such,
can occur quite suddenly and often the damage to the
system is catastrophic. Although it can take place in
lubricated bearings, thermally induced seizure is
predominant when a hydrodynamic bearing happens to
operate in the boundary or mixed lubrication regimes. These
conditions occur during start-up or in an event of lubricant
supply blockage. The objective of this work is to perform acomprehensive study of seizure in bearings during start-up
and arrive at a seizure time evaluation formula that is a
function of the various operating parameters. Dufrane and
Kannel1
analyzed the catastrophic seizure of bearings due to
dry friction by a simple 1D equation relating the seizure time
to the bearing operating parameters and material properties.
Hazlett and Khonsari2
performed a detailed finite element
analysis to gain insight into the nature of the contact forcesand encroachment of the mating pair leading to TIS of a dry
bearing during start up.
The finite element modeling is done using ANSYS 5.73.
First, the TIS analysis of Hazlett and Khonsari2
was
recreated. The finite element model of the present work
employs a finer mesh than the mesh used by Hazlett and
Khonsari to evaluate the contact forces with more accuracy.
The analysis of a bearing undergoing TIS during start up
was done by the following steps: 1. A 2-D static contact
analysis was performed to determine the contact forces and
the contact angle. 2. A transient heat transfer analysis was
done to model thermal effects of dry frictional heating on the
journal and the bearing. 3. A transient thermoelastic analysis
was performed to study the interactions of the journal-
bearing pair during bearing start -up. The variation of radial
clearance, contact forces and ovalization of the bearing were
studied in this analysis.
The loading applied in the thermal analysis consists of
the heat generated by the frictional contact at the shaft-
bushing interface, which is a function of the load, speed and
coefficient of friction. The heat generated is applied to the
journal and the bushing according to the areas of contact on
the shaft and the bushing2
and cooled convectively at the
areas not in contact. The external surface is also cooled by
free convection as shown in Fig 2. The loading for the non-
linear thermoelastic analysis consists of the thermal loads
applied as nodal temperatures and the radial force acting on
the journal. The time dependent thermal load is obtained
from the results of the transient thermal analysis. The static
load, W is applied to act in the negative y-direction on the
shaft. As the model utilizes half-symmetry, a load of W/2 is
applied. Symmetry boundary conditions are used to model
the one-half symmetry as shown in Figure 3. The constraint
of the bearing on its outer surface is modeled by fixing the
bearing at the node under the shaft on the outer edge of the
bearing on the symmetry plane. This constraint
approximates the boundary condition on the bottom surface
of a pillow block type of bearing as shown in Figures 1 and
3.
Due to the rise in temperature, the encroachment of theshaft on to the bushing with concomitant reduction in the
clearance continues until TIS occurs due to the increase in
frictional torque. This process is a complex, non-linear
phenomenon. Analysis shows that TIS is initiated by the
ovalization of the bearing combined with the uniform
outward expansion of the shaft yielding contact between the
top of the shaft and the inner bushing surface. This leads to
an increase in the contact forces and the formation of
multiple contact areas. Increase of contact forces raises the
frictional heat flux and sets up a positive feedback that
accelerates the loss of clearance. Analyses show that the
increase in the frictional torque is abrupt once the
ovalization of the bearing causes the shaft to encroach thebushing, as there is further loss in the operating clearance.
Seizure Criterion - Frictional torque is the torque
resisting the driving torque exerted by the motor. When the
frictional torque increases beyond the extent of the driving
torque capability, it can be concluded TIS is imminent. The
contact forces acting on the gap elements at any instant of
time determine the frictional torque at any time. The frictional
torque increased to exceedingly large values within typically
-
7/30/2019 Book of Abstract 2002
14/38
14
3 seconds after the first instance of establishment of new
areas of contact immediately after ovalization. See Figure 4.
The seizure time can be written as a function of the
speed, load, shaft radius, clearance, friction coefficient and
the bearing length. i.e.
ts = g (N, W, Rs, C, f, L).
The variation of the seizure time during the system start-upis studied when the operating parameters (variablesN, W, Rs,
C, f, L) are varied. Then a generalized equation is derived
depending on the individual relationships of the operating
parameters with the seizure time. Using the range of data
from 72 sets of simulations and applying a statistical
analysis4, we obtain the following expression for the seizure
time.
( ) ( )74.8172.126.15.14.628607 107.51105.1 +=s
LC
sRWfNeet
The empirical relationship is verified for its validity using
some of the results published by Hazlett and Khonsari2
and
Wang et.al.5. See comparitive results shown in Table 1.
FIGURES AND TABLES
REFERENCES
1. Dufrane K. and Kannel J. Thermally induced seizures of
journal bearings.ASME Journal of Tribology, 1989, 111,
288-92
2. Hazlett T.L. and Khonsari M.M. Finite element model of
journal bearing undergoing rapid thermally induced
seizure. Tribology International, 1992a, 25, No.3, 177-82
3. ANSYS 5.7 Online Users Manual, 2001,ANSYS Inc .
4. Hamrock B.J. Elastohydrodynamic lubrication of point
contacts. Ph.D Thesis, Institute of Tribology,
Department of Mechanical Engineering, The
University of Leeds, 1976, 44-66, pp. 93-102
5. Wang H., Conry C. and T.F., Cusano, Effects of
Cone/Axle Rubbing Due to Roller Bearing Seizure on
the Thermomechanical Behavior of a Railroad Axle.
ASME Journal of Tribology , 1996, Vol.118, pp.311-319
-
7/30/2019 Book of Abstract 2002
15/38
15
2002 ME Graduate Student Conference
April 13, 2002
AN EIGENVALUE CONFORMING MODEL FOR A VIBRATIING ROD
Jaeho ShimPh.D. Candidate
Faculty Advisor: Dr. Y. M. Ram
ABSTRACT
The natural frequencies determined by using finite
element or finite difference models of order n are fairly
accurate only about 3n of lower eigenvalues of the
underlying continuous system. The natural frequencies of a
uniform rod with 60=n and exact solution are demonstratedin Figure 1. The new model to improve this existing problemhas been developed. The model named as a spectral
conforming discrete model can estimate the n lowest
eigenvalues of the continuous system with uniform
accuracy. The essential ingredient in building of such a
model is the inverse eigenvalue problem of reconstructing a
chain of mass-spring system with prescribed spectral data
[1].
Consider a non-uniform axially vibrating rod of length
L , axial rigidity ( )xp , and mass per unit length ( )x ,
which is fixed at 0=x and free to oscillate at Lx = , asshown in Figure 2. The axial motion
( )txu ,of the rod at the
time t is governed by the differential equation (1) and two
boundary conditions (2).
( ) ( )2
2
t
ux
x
uxp
x =
(1)
( ) 0,0 =tu , ( ) 0, =
x
tLu , (2)
This non-uniform rod shown in Figure 3 may be
approximated as a piecewise continuous rod with runiform
parameters ip and i within the i th element. In order todetermine a higher order spectral conforming element model,
a uniform rod element of length L, axial rigidityp, and mass
per unit length in Figure 4 is considered. A matrix A
which is reconstructed from spectral data is defined as
2
1
2
1
= KMA . (3)
The matrix A is symmetric tridiagonal with positive diagonal
elementl
and negative off diagonalj
.
=
rr
rrr
1
112
211
11
OOOA (4)
From the matrixA , the stiffness matrix K and the mass
matrix M can be determined using the Lanczos method [2].
The essential novelty in the spectral conforming model
introduced here is that the dynamic response of the
continuous system is fitted to the spectrum of the discrete
estimating model.
As an example, an exponential rod of length L ,
constant Youngs modulus of elasticityE
, constant density
, and variable cross sectional areaxeA = , is presented.
The rod is fixed at 0=x and free to oscillate at Lx = , asshown in the figure 5. The eigenvalues of this system have
been approximated by using finite differences and spectral
conforming model. The finite difference model implemented
60=n elements of equal length. The spectral conformingmodel used 4=n elements of equal length, each of order
15=r . Hence, the global matrices in the three approximatingmethods used are of the same dimension 6060 . Thevarious results obtained are shown in Figure 6. As expected
the spectral conforming model yields superior overall
estimation with uniform accuracy for all eigenvalues
predicted.
Future research in this topic involves extending the
method to include tapered elements that can better capture
the geometry of a non-uniform rod. Broadening the method
over two and three-dimensional elements appears to be a
challenging problem.
-
7/30/2019 Book of Abstract 2002
16/38
16
FIGURES AND TABLES
Figure 1. Predicted natural frequencies using Finite
Difference Model, Finite Element Model and Exact Solution
for 60-digree-of-freedom model.
Figure 2. Non-uniform axially vibrating rod.
Figure 3. An r-degree-of-freedom element
Figure 4. An rth order spectral conforming element model
Figure 5. Exponential rod
Figure 6. Predicted natural frequencies using Finite
Difference Model, Spectral Conforming Model and Exact
Solution.
ACKNOWLEDGMENTS
The research work presented here is supported by a
National Science Foundation research grant CMS-9978786.
REFERENCES
1. D. Boley and G.H. Golub, A survey of matrix inverse
eigenvalue problem, Inverse Problems, Vol. 3, pp. 595-
622, 1987.
2. C. de Boor and G.H. Golub, The numerically stable
reconstruction of a Jacobi matrix from spectral data,
Linear Algebra and Its Application, Vol. 21, pp. 245-
260, 1978.
3. F.P. Gantmakher and M.G. Krein, Oscillation Matrices
and Kernels and Small Vibration of Mechanical Systems,
State Publishing House for Technical-TheoreticalLiterature, Moscow-Leningrad, 1961 (Translation: US
Atomic Energy Commission, Washington DC).
4. J. Paine, A numerical method for the inverse Sturm-
Liouville problem, SIAM J. Sci., Stat. Comput., Vol. 5, pp.
149-156, 1984.
5. J.W. Paine, F. de Hoog and R.S. Anderssen, On the
correction of finite difference eigenvalue approximations
for Sturm-Liouville problems, Computing, Vol. 26, pp.
123-139, 1981
0 10 20 30 40 50 600
50
100
150
200
250
k
k
Finite Difference Model
Finite Elements Model
Exact Solution
0 10 20 30 40 50 600
50
100
150
200
250
k
k
Finite Difference Model
Finite Elements Model
Exact Solution
Finite Difference Model
Finite Elements Model
Exact Solution
)(),( xpx
dx
),( txu
L
x
)(),( xpx
dx
),( txu
L
x 0 10 20 30 40 50 600
20
40
60
80
100
120
140
160
180
200
Exact Solution
Finite Difference Model
Spectral Conforming Model
i
0 10 20 30 40 50 600
20
40
60
80
100
120
140
160
180
200
Exact Solution
Finite Difference Model
Spectral Conforming Model
Exact Solution
Finite Difference Model
Spectral Conforming Model
i
)(),(),( xxAxE
L
)(),(),( xxAxE
L
1k
2k
1m
2m3k
rk
rm
( )
( )
( )e
e
e
A
E
( )e
LD
( )eL
1k
2k
1m
2m3k
rk
rm
( )
( )
( )e
e
e
A
E
( )e
LDD
( )eL
1k
2k
1m
2m3k
rkrm
p, L
1k
2k
1m
2m3k
rkrm
p, L
-
7/30/2019 Book of Abstract 2002
17/38
1
2002 ME Graduate Student Conference
April 13, 2002
BUCKLING ANALYSIS OF GRID STIFFENED COMPOSITE CYLINDERS
Samuel Kidane
M.S. Candidate
Faculty Advisor: Dr. Eyassu Woldesenbet
ABSTRACT
Due to their high stiffness to mass ratio, stiffened cylindrical
composite shells are major components of Aerospace and
Aircraft industries. These structures are employed in
fuselage and fuel tank applications, and are usually
subjected to combinations of compressive, shear or
transverse loads. Usually the failure mode associated with
these structures is buckling. This failure mode is further
subdivided into local skin and/or stiffener buckling, and
universal buckling.
In this paper buckling investigation of a grid stiffened
composite cylinder is presented using analytical model,
Finite elements model and experimentation. The cylinder
under discussion has orthotropic stiffeners integrally made
with an orthotropic shell. All the buckling analysis is based
on a uniaxial compressive load condition.
An analytical model is first developed that reduces the
grid/shell cylinder panel to an equivalent orthotropic
laminate (Fig. 1). This model makes use of the force and themoment interactions and derives the A, B and D matrix of the
equivalent laminate model. Consequently buckling loads are
calculated making use of the energy method. Using the
analytical model developed, parametric analysis is preformed
to determine optimum configuration of stiffeners and shell.
A Finite elements model is also built using ANSYS for
the same cylinder. Buckling analyses are performed on
different models built with different configurations of
stiffener and shell parameters. The effect of shell thickness
variation on buckling load and buckling mode is studied in
detail (Fig 2). Based on this the optimum skin thickness is
determined for a given stiffener configuration. In this section
correlation is made between failure mode and optimum skinthickness. In addition to skin thickness the effect of
stiffeners angle variation is also analyzed using ANSYS. The
result is plotted and conclusions are drawn on optimum
stiffener orientation.
Buckling experiment was also performed on a stiffened
composite cylinder specimen (Fig 3). The test setup and the
results obtained are discussed briefly.
Finally this paper tries to compare the different results
obtained using the three analysis methods. An attempt is
made to account for certain differences observed between
the three analysis methods. Conclusions are drawn on the
reliability of the analytical model developed and remarks
made on limitation of model.
FIGURES AND TABLES
l
x
t
Fig. 1 Unit cell.
Fig. 2 FEM analysis (local skin buckling)
-
7/30/2019 Book of Abstract 2002
18/38
18
-35
-30
-25
-20
-15
-10
-5
0T ime (s )
Load
Strain Gauge1
Strain Gauge2
h/2
1
2
Fig. 3 Experimental results.
ACKNOWLEDGMENTS
Support for this research was provided by a grant from
the NASA/Louisiana Space Consortium and the Louisiana
Board of Regents under LaSPACE (BOR12662) and LEQSF(2001-04)-RD-B-03.
REFERENCES
1. Helms JE, Li G, Smith BH. Analysis of Grid Stiffened
Cylinders. ASME/ETCE 2001.
2. Brush DO, Almroth BO. Buckling of Bars, Plates, and
Shells. McGraw-Hill Book Company, New York, NY
1975.
3. Bruhn EF. Analysis and Design of Flight Vehicle
Structures. Jacobs Publishing, Inc., Carmel, IN June
1973
4. Navin J, Norman FK, Damodar R. Formulation of AnImproved Smeared Stiffener Theory of Buckling
Analysis of Grid-Stiffened Composite Panels. NASA
technical Memorandum 110162, June 1995.
5. Ramm E. Buckling of Shells. Springer-Verlag, Berlin
1982.
6. Gerdon G, Gurdal Z. Optimal Desing of Geodesically
Stiffened Composite Cylindrical Shells. AIAA Journal,
November 1985; 23(11):1753-1761.
7. Troisky MS. Stiffened Plates, Bending, Stability and
Vibrations. Elsevier, 1976.
8. Phillips JL, Gurdal Z. Structural Analysis and Optimum
Design of Geodesically Stiffened Composite Panels.
Report NASA CCMS-90-50, (VPI-E-90-08), Grant NAG-1-643, July 1990.
9. Whitney JM. Structural Analysis of Laminated
Anisotropic Plates. Technomic, 1987.
10. Agarwal BD, Broutman LJ. Analysis and Performance of
Fiber Composites. John Wiley and Sons, 1990.
11. Dow NF, Libove C, Hubka RE. Formulas for Elastic
Constants of Plates with Integral Waffle-like Stiffening.
NACA RM L53L1 3a, August 1953.
-
7/30/2019 Book of Abstract 2002
19/38
19
2002 ME Graduate Student Conference
April 13, 2002
SYNTHESIS, PROPERTIES AND CHARACTERIZATION OF CR-DLC
NANOCOMPOSITE FILMS
Varshni Singh
Ph.D. Candidate
Faculty Advior: Dr E.I. Meletis
ABSTRACT
Diamondlike carbon (DLC) films have been extensively
studied over the past decade, due to their unique
combination of properties. One of the drawbacks with DLC
films is that they are thermally unstable beyond 350o
C [1].Above 400
oC the changes are more profound and
graphitization of the film occurs by conversion of C bonds
from sp3
to sp2, a phenomenon that is also observed during
wear at hot spots [2]. Thus for more than a decade
researchers have focused on metal-containing DLC (Me-
DLC) films in an effort to improve wear resistance, adhesion,
thermal stability and toughness. A number of studies on
synthesis and characterization of Me-DLC films have been
conducted on Si-, Ti-, Ta-, W- and Nb-DLC [3-10]. Even
though Cr is a carbide former and possesses an attractive
combination of other properties (corrosion resistance, wear
resistance, etc.) little work has been reported in this area
[6,11]. The purpose of the present work was to initiate asystematic study of the processing-structure-property
relationship in Cr-DLC films as a function of Cr content. The
objective is to develop a better understanding of this system
and identify possible compositional ranges where
tribological performance and thermal stability are
significantly improved.
Cr-DLC nanocomposite films were deposited on Si
(100) substrate, by reactive magnetron sputtering utilizing an
intensified plasma-assisted processing system. The
processing parameters (chamber pressure, bias voltage,
magnetron current, etc.) were varied to synthesize Cr-DLC
films, with Cr content ranging from ~0.1 at. % to 28 at. %.
Carbon and chromium content was determined by
wavelength dispersive spectroscopy (WDS) utilizing a JEOL
JXA 733 super electron probe microscope. X-ray diffraction
(XRD) experiments were performed, using a Rigaku Miniflex
2 diffractometer with a Cu - K source and transmissionelectron microscopy (TEM) was conducted in a JEOL JEM
2010 electron microscope. Pin-on-disc experiments were
conducted by utilizing an ISC-200 tribometer and the wear
rate was calculated by a Veeco 3D optical profilometer.
Mechanical properties, of the films were studied by
nanoindentation measurements, using a Hysitron
Triboscope instrumented nanoindentation/ nanoscratchdevice incorporated on a Digital Instrument Dimension 3100
atomic force microscope. The short-range order structure ofthe films is being studied by the x-ray absorption fine
structure (XAFS) spectra collected by using bending
magnet radiation at the double crystal monochromator 1
beamline. The thermal stability experiments were conducted
by utilizing a DSC-7 differential scanning calorimeter.
XRD patterns of DLC and Cr-DLC films, show that all
the films exhibited nearly the same XRD pattern, indicating
an amorphous structure. Electron diffraction and high-
resolution TEM studies show that the films, with ~ 9 at. %
Cr, deposited using low (-200 V) and high (-1000 V) specimen
bias during processing are composed of nanocrystalline
metallic Cr and nanocrystalline cubic chromium carbide,
respectively surrounded by an amorphous matrix. Fig. 1show the dark contrast clusters, diameter 1 5 nm,
surrounded by an amorphous matrix corresponding to
nanocrystalline Cr carbide. The XANES spectra of the Cr-
DLC films show that the Cr content (5 to 28 at. %) in the Cr-
DLC films, deposited at 1000 V bias, has little effect on its
structure and Cr atoms are incorporated in the carbon
network. This initial result is in agreement with that of the
TEM results. Furthermore, such preference for metal atoms
to be incorporated into the local carbon structure has also
been shown by our recent XAFS experiments on Si-DLC
films with ~10 at. % Si [12].
Fig. 2 presents the variation of the coefficient of
friction () and wear rate of Cr-DLC films (deposited at 1000V) with Cr content. The results in Figs. 2 show that in
general all Cr-DLC films exhibit a low (less than 0.2) forboth alumina and 440 stainless steel pins. It is interesting to
note that remains at low levels (less than 0.15) for up to aCr/C ratio of 0.24. At a higher Cr level, the results indicate
that the coefficient of friction increases. Very similar
behavior has been observed previously for W-DLC films [5].
The wear rate was also found to be relatively independent of
-
7/30/2019 Book of Abstract 2002
20/38
20
pin material. The wear rate was low and remained at almost
the same levels in films with a Cr/C ratio less than 0.24. At
higher Cr content (Cr/C equal to 0.35), the wear rate
increased significantly (by at least an order of magnitude),
which is consistent with the relatively higher observed forthat film. The present results are in general agreement with
the previous reports however, doesnt show high wear ratesat very low Cr content.
The DSC result qualitatively suggests that the thermal
stability of the Cr-DLC films increases with increasing Cr
content, due to stabilizing effect of Cr on the DLC matrix
network, up to the point where the DLC network is
completely stabilized. Nanoindentation results suggest that
the hardness of the films reduces with increasing Cr content
to ~3 atomic % and then it stabilizes around 13 GPa. With the
exception of an initial deep, reduced modulus (E/(1- ?2))
increases with increasing Cr content and stabilizes around
118 GPa at ~11 atomic % Cr. The H/E/(1- ?2) exhibits a peak
value of ~0.17 at 0.05 atomic % Cr and then gradually
decreases and stabilizes to values around 0.11. Compared toother Me-DLC films, the present profile of H/E/(1- ?
2) for Cr-
DLC exhibits this interesting region yet to be explored
between 0.05 at. % Cr and ~2.75 at. % Cr. Presently, the
reason for this peculiar behavior is unknown.
At present, the in-depth analysis of the spectra
obtained from Cr-DLC films is underway. Study of Cr-DLC
films deposited at lower substrate bias is planned for the
next period. In addition, a couple more compositions in the
aforementioned range between 0.05 at. % Cr and 2.75 at. %
Cr are planned to explore the lower range of the Cr content.
So as to completely understand the effect of Cr content and
the substrate bias on the short-range order around Cr in the
Cr-DLC films further analysis and experiments are underway.
ACKNOWLEDGMENTS
This work was supported by the Army Research Office
grant DAAG55-98-1-0279 and Louisiana Board of Regents.
TEM was performed at MCC facility of LSU with the help of
Dr J. Jiang. Nanoindentations were performed with the help
of Ms Tracy Morris and XAFS spectroscopy was done with
the help of Dr V. Palshin and Dr R. Tittsworth in CAMD,
LSU. WDS was performed using electron microscopy
facility of the Geology Department at LSU with the help of
Dr. Xie. Assistance of Mr. Pankaj Gupta in the deposition
experiments is also acknowledged.
REFERENCES
1. Z.L. Akkerman, H. Efstathiadis, and F.W. Smith,J. Appl.Phys., 80(5), 3068-3075 (1996).
2. Y.Liu and E.I. Meletis,J. Mater. Sci., 32, 3491(1997).
3. A. Varma, V. Palshin, K. Fountzoulas and E.I. Meletis,
Surface Engineering15(4), 301-306 (1999).
4. K. Oguri and T. Arai, J. Mater. Res., 7(6), 1313 (1992).
5. C.P. Klages and R. Memming,Materials Science Forum,
52-53, 609-644 (1989).
6. Y.L. Su and W.H. Kao, J. Mater. Eng. Perf., 9(1) (2000)
38-50.
7. M. Fryda, K. Taube and C-P Klages, Vacuum, 41(4-6)
(1990) 1291-1293.
8. H. Dimigen and C-P Klages, Surf. Coat. Technol., 49
(1991) 543-547.9. W.J. Meng and B.A. Gillispe,J. Appl. Phys. , 84(8) (1998)
4314-4321.
10. K. Bewilogua, C.V. Cooper, Surf. Coat. Technol., 132
(2000) 275-283.
11. X. Fan, E.C. Dickey, S.J. Pennycook and M.K. Sunkara,
Appl. Phys. Let., 75(18) (1999) 2740-2742.
12. V.A. Palshin, R.C. Tittsworth, K. Fountzoulas and E.I.
Meletis,Journal of Materials Science 37 , (2002).
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
1E-9
1E-8
1E-7
1E-6
wear rate
WearRa
te(mm
3/N-m)
Pin Material
440 Steel
Alumina
Cr/C (atomic ratio)
0.1
0.2
0.3
0.4
0.5Load 2.5 N
Pin dia 9.5 mmSliding speed 0.1m/s
Coeffice
intoffriction
Figure 1 HRTEM image of Cr-DLC films with Cr 9
at. % deposited at a bias of 1000V.
Figure 2 Coefficient of friction and wear rate of Cr-DLC
films of Cr/C ratio.
-
7/30/2019 Book of Abstract 2002
21/38
21
2002 ME Graduate Student Conference
April 13, 2002
NUMERICAL SIMULATION OF FLOW AND HEAT TRANSFER IN MICRO HEAT
EXCHANGERS
Readul Islam
M.S. Candidate
Faculty Advisor: Dr. Sumanta Acharya
ABSTRACT
Pin fin arrays have been used for turbine blade cooling
applications in the trailing edge where they can fulfill a
structural as well as a heat transfer function. A recently
proposed concept involves covering the outer surface of
turbine blades with an array of very short pin fins andadding a coating on top that is exposed to external gases [1].
The coating completely enshrouds the turbine blade, leaving
a micrometer-sized gap bridged by the pin fin array, and
preventing contamination of the coolant by the high
temperature gases outside. Cooling air bled from the
compressor stage flows through the gap, protecting the
blade surface.
A preliminary computational study using a commercial
CFD program (Fluent) was undertaken to provide support
and future guidance to the experimental evaluation of the
concept under way at LSU. The problem was modeled as
periodic flow with conjugate heat transfer through an array
of cylindrical pin fins bounded by parallel flat plates. Theenergy input from the external gases was modeled by a
uniform heat flux perpendicular to the fin axis applied to the
top endwall, while the bottom endwall was assumed to be
adiabatic.
To validate the computational model, simulations were
run to compare previous experimental work with various
types of pin fin arrays performed by Chyu [2] and Metzger
[3]. Good agreement with published friction factor and
Nusselt number results therein inspires a level of confidence
in the current computational model.
The computational study investigated the effects of
varying Reynolds number and Prandtl number on the friction
factor and heat transfer characteristics of the basic flow
configuration. Table 1 displays that, for the range of
Reynolds numbers in the present study, the heat transfer
enhancement increases with increasing Reynolds number.
However, this increased enhancement comes with rises in
the friction factor as flow becomes more turbulent. Friction
factors and Nusselt numbers can be incorporated into a
single quantity that contains information about the pressure
penalty paid for increased heat transfer, known as the
thermal performance factor (TPF):
TPF = Nu ?
Table 1 indicates a maximum TPF that falls within the range
of Reynolds numbers used.
Simulations were also performed to evaluate the effectsof modification of the fin array geometric parameters
changing the fin height (H/D), and the streamwise (L/D) and
spanwise fin (W/D) array density. Results presented in Table
Table 1: Results of increasing Reynolds number on friction
factor and Nusselt number (normalized by corresponding flat
channel values), and associated TPF
Re /flat Nu/Nuflat TPF
675 0.3974 1.79 2.43
1350 0.2950 13.49 20.26
2025 4.0011 17.66 11.12
3038 7.5495 22.72 11.58
Table 2: Results of changing geometry parameters (base
case: L/D = W/D = 5; H/D = 2.5) on friction factor and
Nusselt number (normalized by corresponding flat channel
/flat Nu/Nuflat TPF
1.25 15.7 7.9 3.16
2.5 4.6 10.3 6.20H/D
5 3.0 16.6 11.472 5.3 12.3 7.06
5 4.6 10.3 6.20L/D
8 4.8 8.8 5.19
2 29.8 23.0 7.42
5 4.6 10.3 6.20W/D
8 1.9 6.2 4.95
-
7/30/2019 Book of Abstract 2002
22/38
22
2 indicate that from a heat transfer viewpoint, larger H/D
values are desiredthis seems to reflect the rising
effectiveness of the fluid to remove energy as the fin height,
and thus, the channel height is increased. Heat transfer canalso be increased by decreasing L/D and W/D, which means
increasing the fin array density. However, as Table 2
demonstrates, increasing fin density comes with severe
pressure penalties. Nusselt numbers are more responsive to
W/D changes than L/D, possibly because the wake of the
flow (L/D) isn't as important as the shear layers that form
around the fin region. Taken overall, the best performance
will be obtained from closely spaced arrays with large height
to diameter ratiosessentially, the current results are
predicating a move toward transverse flow across closely
spaced tube bundles. This observed trend matches well with
[4], where it was noted that the heat transfer enhancement
using pin fin arrays of intermediate H/D is lower than that for
classic tube bundles operating at the same Reynolds number
(until very high Reynolds numbers are reached).
While the pin fin arrays mentioned so far consist of
cylindrical pin fins, it is by no means certain that a circular
cross section produces the most effective heat transfer
enhancement. The computational study was extended to
investigate the effects of a variety of fin shapes at a
Reynolds number of 3000. The table associated with Figure 1
illustrates performance factors for square, diamond and
elliptical fins oriented parallel and perpendicular to the flow.
In terms of TPF, the last case offers a viable alternative to
cylindrical pin fin arrays.
The results presented allow optimal choices ofpromising geometric parameters and fin shapes to be made
for further experimental study.
ACKNOWLEDGMENTS
This study is supported by DARPA. The author wishes
to thank Dr. Acharya and Dr. Kelly for valuable insight and
guidance, and Christophe Marques for fruitful discussions.
Criticism from Dr. A. K. Saha, constructive much more often
than not, is gratefully acknowledged.
REFERENCES
1. J. C. COYNEL, MS thesis, Louisiana State University,
Baton Rouge, Louisiana, 1999.
2. M. K. CHYU, Y. C. HSING and V. NATARAJAN, J.
Turbomachinery. 120 (1998) 362-367.
3. D. E. METZGER, C. S. FAN and S. W. HALEY, J. Eng.
For Gas Turbines and Power. 106 (1984) 252-257.
4. D. E. METZGER, R. A. BERRY and J. P. BRONSON, J.
Heat Transfer. 104 (1982) 700-706.
A B
C D
Figure 1. Nusselt number contours on top surface of channel for different pin shapes for L/D = W/D = 5, H/D = 2.5
arra s and Re nolds number = 3x104
and associated erformance factors
-
7/30/2019 Book of Abstract 2002
23/38
23
2002 ME Graduate Student Conference
April 13, 2002
ADVANCED TURBULENCE MODELING FOR INDUSTRIAL APPLICATIONS
Raymond M. Jones
Ph.D. Candidate
Faculty Advisor: Dr. Sumanta Acharya
ABSTRACT
Computational fluid dynamics ( CFD ) in industry is
typically performed using the Reynolds-averaged Navier-
Stokes equations ( RANS ). With RANS complex turbulent
flows at high Reynolds numbers can be solved with
reasonable accuracy and within an acceptable amount of
time. Other methods such as large eddy simulation ( LES )
and direct numerical simulations ( DNS ) can be used to
obtain more accurate results but require extensive
computational resources, even for complex turbulent flows
at moderately low Reynolds numbers.
With RANS, a turbulence model is needed to close the
momentum equations. Two-equation turbulence models
such as the k- model have been primarily used in industrybecause of their robustness. There are several different
types of two-equations turbulence models but they are
similar in that they all use damping functions to accurately
represent near wall turbulence. These damping functions
are usually derived for simple flows, such as flow in achannel, and are ill suited for most complex flows.
In the k- model, the eddy viscosity is computed as
2kCt = where C is typically 0.09. Figure 1 shows
computed values of t for 09.0=C and 075.0=C for a
turbulent channel flow. It can be seen that t is
overpredicted for both values of C . This shows that t
computed by the k- model can not accurately predict thenear wall eddy viscosity as long as C is a constant. The
damping functions mentioned above are used to replace C
with a function.Based on this observation Durbin (1991) introduced an
alternative eddy viscosity formulation defined as
kvCt2= where 2.0=C and
2v is the velocity
fluctuation normal to the wall. This formulation can also be
seen in Fig. 1. Durbin (1991) developed the kfv 2 model, which solves a scalar transport equation for k, , and
2v which are used to computed the eddy viscosity as
kvCt2= . The kfv 2 model has shown to
produce better results compared to conventional two-
equation models (Durbin (1992), Parneix et al. (1998)) .
Despite the improved accuracy, the kfv 2 model has
shown to be numerically stiff even for simple flows. In thepresent work a new model has been developed which uses
the general framework of the original kfv 2 model but
has shown to be more robust. In the new model, the
equation in the kfv 2 model has been replaced by the
equation of Wilcox and is called the kfv 2 model. A newrelationship between and is introduced based ondimensional arguments, and several of the closure constants
calibrated with respect to DNS data.
The kfv 2 model has been tested for predicting the
channel flow of Moser (1999) for Re .=395. The kfv 2
model was compared with the k- model and the kfv 2 model and the results for the streamwise velocity predictions
can be seen in Fig. 2. All of the models accurately predict
the correct velocity profile. Figure 3 shows kinetic energy
predictions. It can be seen that the kfv 2 model showsimproved predictions compared to both the k- model and
the kfv 2 model. The kfv 2 model was tested forpredicting the backward facing step of Kasagi (1993). Figure
5 shows that the kfv 2
model and the kfv 2
modelproduce better near wall velocity predictions compared to
the k- model.
-
7/30/2019 Book of Abstract 2002
24/38
24
FIGURES AND TABLES
+
y+
0 0.02 0.04 0.06 0.08 0.10
50
100
150
200DNS
0.09k2/
0.075k2/
0.2kv2/
Fig. 1 Exact eddy viscosity compared with the k- model
(blue and orange) and the kfv 2 model (red line)
y+
u+
100
101
1020
4
8
12
16
20
24 Re = 39 5
v2f-k
k-
v2f-k
Fig. 2 Mean Velocity
k+
y+
0 1 2 3 4 5 60
80
160
240
320
400 Re = 395
v2f-k
v2f-k
k-
Fig. 3 Kinetic Energy
y+
diss
prod
0 25 50 75 100-0.3
-0.2
-0.1
0
0.1
0.2
0.3Re = 395
v2f-k
k-v
2f-k
Fig. 4 Production and Dissipation
x/H
y/H
0 1 2 3 4 5 6 7 8 9 10 110
0.5
1
1.5
2
2.5
3
Fig. 5 Streamwise velocity for the k- model (solid line),
kfv 2 model (dashed), kfv 2 model (dotted)
x/H
y/H
0 1 2 3 4 5 6 7 8 9 10 110
0.5
1
1.5
2
2.5
3
Fig. 6 Kinetic Energy for the k- model (solid line),
kfv 2 model (dashed), kfv 2 model (dotted)
ACKNOWLEDGMENTS
I would like to acknowledge The Dow Chemical for their
support on this project.
REFERENCES
1. P. A. DURBIN,AIAA Journal. 33 (1991) 659-664.
2. P.A. DURBIN,Annual Research Briefs, (1992) 3-16.
3. N.KASAGI, A. MATSUNAGA, and S. KAMARA, J.
Wind Eng. Ind. Aero. 46 (1993) 821-829.
4. R.D.MOSER, J.KIM, N.MANSOUR, Phys. Fluids. 11
(1999) 943-945.
5. S.PARNEIX, M.BEHNIA, P.DURBIN, Annual Research
Briefs, (1998) 149-164.
-
7/30/2019 Book of Abstract 2002
25/38
25
2002 ME Graduate Student Conference
April 13, 2002
TRIBOLOGICAL BEHAVIOR OF NANOSTRUCTURED NICKEL
Dean J. Guidry
M.S. Candidate
Faculty Advisor: Dr. Efstathios I. Meletis
ABSTRACT
The present study reports the effects of electroplating
parameters on the microstructure, and thus the mechanical
and tribological properties, of nanostructured nickel.
Electroplating was conducted in a Watts type bath at
current densities of 30 mA/cm2 and 15 mA/cm2 in
electroplating bath temperatures of 30C and 50C. The PHof the bath was maintained at 3.0 using sulfuric acid. The
electroplating was carried out using a direct current in
galvanostatic mode with a nickel anode contained in a
titanium wire basket. Average grain sizes and uniformity of
grains were determined from TEM and SEM micrographs.
Tribological tests were carried out on a pin-on-disc type
tribometer. The same tests were conducted on Ni-200 for the
purpose of comparison. Wear rates were calculated for the
nickel surfaces using optical profilometry and for the
alumina pins using optical microscopy. Nano-indention
techniques provided the nanohardness, stiffness, and
reduced modulus values for all samples. Microhardnessreadings were also recorded to further study the surface
properties. Result s show how grain size, mechanical
properties, and wear properties change with the variations in
plating parameters. Grain size reduction shows surface
hardness increases and improved tribological properties.
Plating bath temperature increases showed a decrease in
grain uniformity.
ACKNOWLEDGMENTS
The Author acknowledges Dr. Efstathios I. Meletis, Dr.
Kun Lian, Dr. Jie Chao Jiang and Varshni Singh for their help
and advice throughout this research project.
REFERENCES
1. G. Robinson, Electronic Engineering Times, 958 (1997)
33.
2. T. E. Buchheit, T. R. Christenson, D. T. Schmale, D. A.
Lavan,Mater. Res. Soc. Sym., 546 (1998) 126.
3. A. M. El-Sherik, U. Erb,J. Mater. Sci., 30 (1995) 5743.
4. V. Provenzano, R. Valiev, D. G. Rickerby, G. Valdre,
NanoStructured Mater., 12 (1999) 1103.
5. F. H. Froes and C. Suryanarayana, J. Mater. Sci., June
(1989) 12.
6. H. Gleiter,Acta Materialia, 48 (2000) 1.
7. H. Gleiter,NanoStructured Materials, 6 (1995) 3.
8. C. Meneau et. al., Surface and Coatings Technology,
100 (1998) 12.
9. F. Ebrahimi, G. R. Bourne, M. S. Kelly, T. E. Matthews,
NanoStructured Mater., 11 (1999) 343.
10. Z. N. Farhat, Y. Ding, D. O. Northwood, A. T. Alpas,
Mater. Sci. Eng. A, 206 (1996) 302.
11. S. W. Banovic, K. Barmak, A. R. Marder, J. Mater. Sci.,
33 (1998) 639.
12. E. O. Hall,Proc. Phys. Soc., B64 (1951) 747.
13. N. J. Petch,J. Iron Steel Inst., 174 (1953) 25.
14. L. S. Stephens, K. W. Kelly, E. I. Meletis, S. Simhadri,
Mat. Res. Symp. Proc., 518 (1998) 1.
15. A. H. Chokshi, A. Rosen, J. Karch and H. Gleiter, Scripta
Metall. Mater., 23 (1989) 1679.
16. K. Lu, W. D. Wei and J. T. Wang, Scripta Metall.Mater., 24 (1990) 2319.
17. G. E. Fougere, J. R. Weertman and R. W. Siegel, Scripta
Metall. Mater., 26 (1992) 1879.
18. G. Palumbo, U. Erb and K. T. Aust, Scripta Metall.
Mater., 24 (1990) 2347.
19. N. Wang, Z. Wang, K. T. Aust, U. Erb, Acta Metall.
Mater., 43 (1995) 519.
20. R. Z. Valiev, N. A. Krasilnikov, N. K. Tsenev, Mater. Sce.
Eng. A, 137 (1991) 35.
21. S. M. Myers, J. A. Knapp, D. M. Follstaedt, M. T.
Dugger,J. Appl. Phys., 83 (1998) 1256.
22. Y. Ichida, K. Kishi, Trans. of the ASME, 119 (1997) 110.
23. D. A. Rigney,Mat. Res. Innovat., 1 (1998) 231.24. V. Singh, X. Nie, P. Gupta, E.I. Meletis, 6
thNat. Cong.
Mech. Proc., 2 (2001) 1.
25. K. Laul, M. Dorfman, Adv. Mater. and Proc., 158 (2000)
46.
26. M. Legros et. al.,Phil. Mag. A, 80 (1999) 1017.
27. A. B. Witney, P. G. Sanders, J. R. Weertman, Scripta
Metall. Mater., 33 (1995) 2025.
-
7/30/2019 Book of Abstract 2002
26/38
26
28. X. J. Wu et. al., NanoStructured Materials, 12 (1999)
221.
29. T. R. Christenson, T. E. Buchheit, D. T. Schmale, R. J.
Bourchier,Mat. Res. Soc. Symp. Proc., 518 (1998) 185.
30. E. Bonetti et. al.,NanoStructured Mater., 11 (1999) 709.
31. S. Greek, F Ericson,Mat. Res. Symp. Proc., 518 (1998) 51.
32. A. Cziraki et. al., Thin Solid Films, 318 (1998) 239.33. G. Auner, et. al., Thin Solid Films, 107 (1983) 191.
34. R. V. Nandedkar et. al.,Phys. Status Solidi, 72 (1982) 89.
35. J. C. Pivin et. al.,J. Mater. Sci., 22 (1987) 1087.
36. P. S. Barlow, R. A. Collins, G. Dearnaley,J. Phys. D: Appl
Phys., 22 (1989) 1510.
37. D. M. Mattox,J. Electrochem. Soc, 115 (1968) 1255.
38. K. Bouslykhane, J. P. Villain, P. Moine, Tribology
International, 29 (1996) 169.
39. L. Jun, W. Yiyong, W. Dianlong, H. Xinguo, J. Mater.
Sci., 35 (2000) 1751.
40. F. Ebrahimi, A. J. Liscano, Mater. Sci. Eng. A, 301 (2001)
23.
41. R. I. Pratt, G. C. Johnson, Mat. Res. Soc. Symp. Proc.,518 (1998) 15.
42. T. Bieger, U. Wallrabe,Microsystem Tech., 2 (1996) 63.
43. I. M. Hutchings,Mater. Sci. Eng. A, 184 (1994) 185.
44. J. Ferrante,Phys. World, July (1991) 46.
45. S. Jahanmir, in N. P. Suh and N. Saka (eds),
Fundamentals of Tribology, MIT Press, Cambridge,
(1980) 455.
46. J. F. Archard,J. Appl. Phys., 24 (1953) 981.
47. D. J. Tillack, E.B. Fernsler, ASM Handbook Ninth
Edit ion, 10 (1996) 754.
48. LECO Corporation, Metallography Principles and
Procedures, (1994) 38.
-
7/30/2019 Book of Abstract 2002
27/38
27
2002 ME Graduate Student Conference
April 13, 2002
INTRINSIC STRESS DEVELOPMENT IN TI-C:H CERAMIC NANOCOMPOSITE
COATINGS
B. Shi
Ph. D. Candidate
Faculty Advisor: Dr. W. J. Meng
ABSTRACT
The development of intrinsic stresses within titanium-
containing hydrocarbon (Ti-C:H) coatings was monitored by
in-situ substrate curvature measurements using a multi-beam
optical sensing (MBOS) technique. Stress as a function of
the Ti-C:H layer thickness was monitored in a wide range ofspecimens, from nearly pure amorphous hydrocarbon (a-
C:H) to nearly pure titanium carbide (TiC). The intrinsic
stress within Ti-C:H was found to vary significantly in
magnitude and depend systematically on the Ti
composition.
Ti-containing hydrocarbon (Ti-C:H) coatings,
consisting of a nm-scale mixture of crystalline titanium
carbide (TiC) and amorphous hydrocarbon (a-C:H)1, form a
prototype of pseudo-binary ceramic nanocomposites. Ti-C:H
coatings possess mechanical properties and tribological
characteristics which depend systematically on coating
composition2, demonstrating the potential of engineering
ceramic nanocomposite coatings for specific applications.The dependence of tribological characteristics of Ti-C:H
coatings on the Ti composition has been related to a
percolation type transition3. In addition to the influence of
plasma characteristics, the coating composition may
therefore exert a significant influence on intrinsic stress
development within ceramic nanocomposite coatings.
This paper addresses the dependence of intrinsic
stresses within Ti-C:H coatings on the Ti composition. Ti-
C:H deposition was accomplished using a high-density
plasma assisted hybrid chemical vapor deposition
(CVD)/physical vapor deposition (PVD) process. Under
nominally identical plasma conditions, a series of Ti-C:H
coatings, ranging from nearly pure a-C:H to nearly pure TiC,
was deposited onto Si(100) substrates. The development of
intrinsic stresses was monitored by in-situ measurements of
substrate curvature change. Our results show that the
intrinsic stress within Ti-C:H coatings depends
systematically on the Ti composition. A significant increase
in stress was observed as the Ti composition increases
beyond 30 at. % and suggested to be related to the
percolation type transition in the coating mi crostructure.
Details of experimental setup and procedures will be
described in the presentation.
Figure 1 shows a typical substrate temperature time
history during a complete Ti-C:H deposition run. The Ti
cathode current was 1.0A during Ti-C:H deposition. During
the etch and cool stages, the substrate temperature rosefrom ~ 25 oC to ~ 225 oC and fell to ~ 150 oC. It stayed ~ 150oC during Ti interlayer deposition, and rose to ~ 225
oC
during Ti-C:H deposition. During the entire deposition run,
the temp erature difference between front and back beam
surfaces was 5 K. Such a temperature difference across a300 m thick Si wafer would induce a substrate curvature
change K of ~ 1/21 m-1. Figure 1 shows that, during Tiinterlayer and Ti-C:H deposition, change in relative reflected
laser spot spacing D/D0, which is measured experimentallyand linearly related to the curvature change, induced by
temperature gradient across the Si substrate is substantially
smaller than 10 %.
A multitude of Ti-C:H/Ti/Si(100) specimens weredeposited. During each deposition, the curvature change of
the Si(100) beam substrate was monitored by MBOS. Figure
2 shows the average composition of the Ti-C:H layers as a
function of the Ti cathode current obtained by combining
RBS and ERD measurements. The Ti and hydrogen
compositions respectively increase and decrease in a
monotonic fashion with increasing Ti cathode current. The
observed trend is consistent with our previous results and
supports the fact that Ti-C:H coatings are pseudo-binary
TiC/a-C:H nanocomposites, in which hydrogen inclusion
occurs only through incorporation into the a-C:H phase4.
Figure 3 shows measured D/D0 during Ti-C:H deposition asa function of time. The time origin coincides with the
beginning of Ti-C:H deposition. Only the curvature change
due to Ti-C:H deposition is taken into account, as D/D0was set to zero at time zero. In the early stage of growth, 0
400 sec, D/D0 increases approximately linearly to ~ 30%,independent of the Ti composition. In the late stage, 400
2500 sec, D/D0 continues to increase linearly with time, butin most cases with a distinctly different slope as compared
to the early stage.
-
7/30/2019 Book of Abstract 2002
28/38
28
Temperature measurements, such as the one shown in
Figure 1, showed that during the early stage of Ti-C:H
growth, the substrate temperature rose ~ 70 K. In the late
stage, the substrate temperature change was 25 K at all Ticathode currents. This temperature change T would inducea substrate curvature change KT due to the difference in
thermal expansion between Si and the Ti-C:H coating s-c,
where Ys, ts, Yc, and tc are respectively the biaxial modulus
and thickness of the substrate and the coating. s-c wastaken to be ~ 410-6 K-1, according to measurements on W-C:H coatings
5. For the present measurements, a
conservative estimate yielded D/D0 +5% forT = +100 K6. It is thus concluded that thermal contribution to the
present measurements can be neglected, and that in all cases
the measured D/D0 reflects intrinsic stress development.The linear dependence of D/D0 on time during late stagegrowth indicates a constant level of incremental intrinsic
stress as the Ti-C:H layer thickens.
In summary, a detailed experimental study of the
dependence of intrinsic stress within Ti-C:H coatings on the
Ti composition was performed by measuring in-situ
substrate curvature change. The intrinsic stress within Ti-
C:H layers was found to vary significantly in magnitude and
depend systematically on the Ti composition. The observed
stress dependence on coating composition is suggested to
correlate with a percolation type transition in the coating
microstructure, and may be one generic characteristic of
ceramic nanocomposite coating systems.
ACKNOWLEDGMENTSWe gratefully acknowledge partial project support from
NIST through ATP program 70NANBH0H3048, NSF through
grant DMI-0124441, and Louisiana Board of Regents
through contracts LEQSF(2000-03)-RD-B-03 and
LEQSF(2001-04)-RD-A-07. The ion beam analysis work at the
Argonne National Laboratory was performed by L. E. Rehn
and P. M. Baldo and supported by the DOE Office of
Science, Basic Energy Sciences, under contract #W-31-109-
ENG-38.
FIGURES AND TABLES
REFERENCES
1. W. J. Meng, R. C. Tittsworth, J. C. Jiang, B. Feng, D. M.Cao, K. Winkler, V. Palshin, J. Appl. Phys., 88, 2415
(2000).
2. W. J. Meng, R. C. Tittsworth, L. E. Rehn, Thin Solid
Films 377/378, 222 (2000).
3. B. FENG, D. M. CAO, W. J. MENG, L. E. REHN, P. M. BA
4. LDO, G. L. DOLL, THIN SOLID FILMS 398/399, 210
(2001).
5. D. M. Cao, B. Feng, W. J. Meng, L. E. Rehn, P. M.
Baldo, M. M. Khonsari, Appl. Phys. Lett. 79, 329 (2001).
6. J. S. Wang, Y. Sugimura, A. G. Evans, W. K. Tredway,
Thin Solid Films 325, 163 (1998).
7. For the present Si(100) beam substrates, Ys = 180 GPa
and ts = 300 m. tc is taken to be 800 nm, the entire
thickness of the Ti-C:H layer. The maximum biaxial
modulus for Ti-C:H coatings is ~ 180 GPa (ref. 8), thus
Yc is taken to be 180 GPa.
Time (sec)
0 500 1000 1500 2000 2500
Temperature(C)
0
50
100
150
200
250
Ti cathode 1.0A, front surface
Ti cathode 1.0A, b