Exact Perturbed Unsteady Boundary Layers

International Journal of Modern Research in Engineering and Technology (IJMRET)
www.ijmret.org Volume 3 Issue 2 ǁ February 2018.
w w w . i j m r e t . o r g I S S N : 2 4 5 6 - 5 6 2 8 Page 38
Exact Perturbed Unsteady Boundary Layers
S.P. Farthing
Econologica, N. Saanich BC Canada
Abstract: The time diffusive boundary layer from rest for Taylor series uniform outer flow is constructed by
integration of the complementary error function of the diffusive similarity variable . The same erfc solution is
also easily “Stokes” transformed to solve harmonic outer flow from rest.
If the outer flow vector varies weakly along the boundary, the perturbation pressure gradient accelerates
the slowest fluid nearest the wall the most to disproportionately vary the shear stress. So centripetal
acceleration causes secondary crossflow inside the boundary layer with strong wall shear towards the center of
curvature. But longitudinal acceleration also implies an inflow towards the wall which thins the outer boundary
layer with a weak further increase of the wall shear, but the strongest perturbation steady streaming.
Simple new particular solutions for these two perturbations are easily constructed in terms of products of
integrals and derivatives of any primary diffusive solution. For an outer flow as a series in the square root of
time, all homogeneous time coefficients remain just iterated error functions. Each systolic pulse in the aortic
arch was considered as a Taylor series flow from rest to calculate the wall shear vectors.
When the outer flow oscillates forever more, its primary diffusive boundary layer asymptotes to the Stokes
oscillatory exponential decay with distance from the wall. The particular perturbations are exactly evaluated
and also confined near the wall but with mean slip. The mean slip homogenous perturbations diffuse outside the
Stokes layer into steady streaming as complementary error functions with inverse time correction functions.
Extended Taylor series computations provide more detail of the perturbation transients.
Keywords: Heat Diffusion Equation, Secondary Flow; Steady Streaming,; Stokes
I. Introduction
A theoretical study [1] was undertaken of the
boundary layer in the aortic arch with a view to
possibly explaining the patchy occurrence of
atherosclerotic lesions on the arch walls, for
instance due to variation blood particle motion near
the wall[2]. Experimental study of the boundary
layer is complicated by the flexibility of the aortic
wall and the difficulty of in vivo examination,
requiring the sacrifice of large animals such as
horses.
The local diffusive boundary layer was perturbed
for pressure gradients and inflow due to the helical
curvature of the aortic arch and the stream wise
variation of that curvature‟ strength and plane. The
maximum bend is tight and it was not necessary to
resort to the inappropriate small curvature
approximation commonly used. In fact the
numerical results were quite different than the low
curvature ones, with secondary flow on the inside of
the bend much intensified by the stronger potential
vortex core flow there. This indicated the secondary
flow would focus into a jet normal to the wall at the
very inside of the bend. The jet separates the
boundary layer and injects vorticity into the flow as
a whole. This intrusion of the boundary layer into
the bulk flow is seen in the pulsed Doppler
ultrasound profiles of Peronneau [3]. Downstream
in a light coil, this core vorticity would eventually
transform the flow into the high Dean number
steady flow of greater axial velocity near the outside
of the bend.
Some correlations were found between wall shear
variations and the patterns of disease which is
actually different from that of dye uptake and other
experimental patterns. Unfortunately no absolutely
compelling matches were found and the entire
physiological localisation patterns and possible
mechanisms need closer scrutiny rather than further
flow studies.
During this work simple ways were found to
construct homogeneous solutions for the diffusion
(heat) equation and first order solutions for fluid
spatial acceleration perturbations of it. These might
also be adapted to solution and perturbation of other
linear pde‟s. Here they will be presented for the
generalised fluid dynamic problem of perturbing
starting and oscillatory boundary layers for spatial
acceleration of the coreflow.

Figure 1 Illustrative Geometry
II. Perturbed Diffusive Boundary layer Equations
As a simple illustrative case consider the inviscid incompressible streamline velocity G()W(X) parallel to
and small distance Y below the centerline of the top wall of a slightly tapered square duct, as in Figure 1. Here
G() is a non-dimensional O(1) pulsatory/oscillatory time factor starting at time =0. Then the axial X gradient
of pressure p due to temporal and advective accelerations resp. is the inviscid
dX p= - (WdG + G2
WdXW ) (1)
with  the density. With  the kinematic viscosity, the 2D boundary layer equation for the centerline velocity w
at small distance Y below the top wall is
dw + wdxw + udYw = -dx p/d2
Yw (2) where dYu + dXw=0 (3)
w=0 at Y==0 and w tends to GW as Y tends to ∞.
Let t= /T, T a characteristic time such as the angular period of a harmonic G, and let y=Y/ √T, the
characteristic diffusion distance scale in time T
Then d -d2
Y becomes D/T where D= dt -d2
y is the Diffusion operator
When the WdG dominates G2
WdXW, the approximate primary diffusive 0‟th order solution near =0 is velocity
Wgo(y,t) with non-dimensional deficit or interference g= G- g0 such that
dt g-d2
y g =D[g]=0 g=G(t) at y=0 (4)
Now the pressure gradient due to the acceleration of the coreflow parallel to the wall dX p=- G2
WdXW
will perturb this boundary layer more than the correction for the acceleration in the boundary layer  wdXw so
the net “planar/pressure” 1‟st order correction is TWdXWv1 where TdXW<<1
D[v1]= G2
-go
2
=2Gg-g2
(5)
The first term dominates and the net peaks at G2
at the wall which of course restrains the response. Thus very
strong shear is to be expected from it as of interest in atherogenesis. For instance, if the duct is also curved with
centerline radius R as in Fig 1, the inviscid flow has a lateral centripetal pressure gradient dZ p=-G2
W2
/R so
that in the boundary layer a secondary lateral flow of v= T W2
v1 /R starts towards the center of curvature.
For a longitudinal acceleration the walls are converging towards the center axis flow but the normal component
u must decline to nil at the wall ie from (3) u= -dXW∫0
Y
g0 dY which increases at first as the square of the

distance from the wall where go=0 until linearly outside the boundary layer. So there is also a positive 1‟st order
correction to g0 for – udYw “advection in” or “indraft” as TWdXW w1
where D[w1]= dYgo∫0
Y
godY= dygo∫0
y
gody (6)
For a typical (starting) boundary layer profile such as go= Gerf  dgo ∫0

go d peaks at about G2
/10 at  ≈1
(where 2Gg-g2
≈.3) before vanishing at infinity outside the boundary layer. So the w1 response to this forcing
can be expected to be about .1 v1„s and this is indeed borne out for the shear perturbations in both starting and
oscillating coreflows. But anticipating section IV, comparing  moments about the boundary for the same G,
the factor is about .3. Now
∫0
y
go dy = ∫0
y
(G-g)dy= {yG-∫0
y
gdy} so D[w1]= dyg {-yG+∫0
y
gdy} (7)
where the first term is the inviscid inflow dominating away from the wall, though near the wall it is cancelled to
leave the net y2
increase. So then the first order longitudinal acceleration correction terms to w are TWdXW
(v1+w1), small vs W as assumed if TdXW<< 1.
Thus there are two basic and very different perturbed diffusion equations to be solved in all first order
perturbations of a time diffusive boundary layer for spatial acceleration effects. Previously w1+v1 longitudinal or
v1 transverse have been laboriously solved for specific G(t) eg impulsive, linear and quadratic or at t= ∞
harmonic [4] without interconnection or comparative understanding. Here general (particular) solutions for the
w1 and v1 are presented and then the complete solutions found for any series in √t, and asymptotic solutions for a
simple harmonic.
III. Perturbation Solutions for General Time Variation
Since D is a linear differential operator with homogeneous solution g: D[g]=0 , then dtg or dygare further
solutions D[dtg]=0 and D[dyg]=0. Likewise the indefinite integral in t generates a solution, or the definite
integral ∫c
t
gdt if g=0 at t=c, and ∫c
y
gdy if g=dyg=0 at y=c. To generalize the inversion ery
∫c
y
e-ry
gdy of the
basic differential factor dy+r commutes with D provided e-ry
g & e-ry
dy g vanish at y=c. (This provides one way
of generating the g for harmonic G(t) starting at t=0 in section IV).
To find particular solutions of the above perturbed diffusion problems, simply note that D[gE(t)]=gdtE and that
if h is another homogeneous D[h]= 0 solution, then D[gh]=-2 dyg dyh These immediately provide a particular
solution to the simpler pressure problem v1= v1p - v1h the difference of particular and homogeneous parts such
that
D[v1h]=0 D[v1p]=2Gg-g2
v1p=2g∫0
t
Gdt + ½[∫∞
y
gdy]2
(8)
Then at y=0, v1p=v1h=2G∫0
t
Gdt+½[∫∞
o
gdy]2
(9) and dyv1p=2dyg∫0
t
Gdt -G∫0
∞
gdy (10)
For w1 (7): D[w1p]=dyg {-yG+∫0
y
gdy}
Trying -y dy g ∫0
t
Gdt gives the dominant first term plus 2d2
y g∫0
t
Gdt or 2dt g∫0
t
Gdt which like the last term dy g
∫0
y
gdy} = dy g { ∫0
∞
gdy+∫∞
y
gdy} is easily solved by the above properties to get
w1p= -y dy g∫0
t
Gdt -2dt g∫0
t
dt ∫0
t
Gdt +dy g ∫0
t
∫0
∞
gdy - ½ g ∫∞
y
∫∞
y
gdy (11)
at y=0 w1p=w1h = -2dt G∫0
t
dt ∫0
t
Gdt +dy g ∫0
t
∫0
∞
gdy - ½ G ∫∞
0
∫∞
y
gdy D[w1h]=0 (12)
dyw1p = -dy g∫0
t
Gdt -2d2
t y g ∫0
t
dt ∫0
t
Gdt +dt G ∫0
t
∫0
∞
gdy - ½ dy g ∫∞
0
∫∞
0
gdy- ½ G∫∞
0
gdy (13)
there are no common elements to v1p and w1p (because their fluid dynamic origins are distinct) so they are the
simplest basis functions to solve versus any linear combination such as v1p+ w1p
IV. Diffusive and Perturbation Solutions for Root Powers of Time
From the fundamental solution Q0=erfc  of the diffusion equation D[Q0]=0, where the diffusive similarity
variable =½y/ √t , the same dy and ∫∞
y
dy operators can construct series of solutions Qn for g=G(t) =√t n
at
y=0 as in Table 1.

TABLE 1 D[Fn]=0 solved by Qn and Pn of form Fn =√t n
fn() Fn =dy Fn+1 =½y/ √t
With q and p both solving f() such that fn=½f ’n+1 nfn fn„-½fn”=0
so ½nfn fn-1+fn-2 Define n =[n+ ½ ]
n √t n
qn() qn()=(-1)n
/n pn(0) pn()= ½ lim qn() Y-∞
4 t 2
16i4erfc() ½ ½ ½ +22
+ 24
/3
3 √t 3
-8i3erfc() -4/3√ 0 2+43
/3
2 t 4i2erfc() 1 1 2
+1
1 √t -2i1erfc() -2/√ 0 2
0 1 erfc() 1 1 1
-1 1/√t -𝒆− 𝟐/
√ -1/√𝒆− 𝟐
∫0

𝒆 𝐳 𝟐
dz p-1
-2 t -1
𝒆− 𝟐
/√ 0 1 1- p-1
-3 t -3/2
(½-
𝒆− 𝟐
/√ ½ /√ 0 (2
-½) p-1 -
-4 t -2
(
3/2𝒆− 𝟐
/√ 0 -1 (-
3/2 p-1+ (2
-1)
Notes: The multiplier of -𝒆− 𝟐
/√ is the same as the multiplier of p-1= 2DawsonF() The negative index
P-n= dn-1
y (2DawsonF()/√t) so p-n= (½)n-2
d
n-1
DawsonF() which alternates in non-zero value at y==0
with the q-n to allow solution for G(t) as asymptotic series in inverse powers of large time. The pn series
breaks between ∫∞
y
p-1dy divergent and dyp0=0, and for positive index the pn diverge at infinity.
_________________________________________________________________________________
These solutions of the general form Fn=√t n
fn() so dy =d / 2√t dy=2√t d .
D= t ½ t -1
 - ¼t -1
2
 so D[Fj] =½√t n-2
(nfn fn„-½ fn”) =0
a second order ode with two independent set of solutions being the Qn=√t n
qn() and Pn=√t n
pn()
tabulated above. Since all fn() are series of  derivatives, this o.d.e gives the recursion relation……
½nfn fn-1+fn-2 useful for computing.
For non-negative n, qn =(-2)n
inerfc()=(-2)n
in the iterated complementary error functions as in Figure 2
where d in= -in-1 . All qn or in vanish at infinite 
So for G(t) =  an√t n
(with ∫0
t
Gdt=  2an√t n+2
/ n+2) g=  an n (2√t)n
in where the n= (½n+1)
normalize the 2n
in as in Figure 2. Note n/ n-2 = ½n.
Remember in+1=∫
∞
d in so the primary shear is dyg= - ann (2√t)n-1
in-1 . Thus with a pulsatile flow for
example a parabolic pulse, the shear reverses before the coreflow. Then from (8) and (9)
v1= anas(2√t)n+s+2
[nin/2s
(s+2)+ ½ nin+1sis+1 -n+s+2 in+s+2{4/(s+2)+ ½ ns/ n+1s+1 }] (14)
As anticipated the v1p or first two terms in the bracket are positive with negative shear reflecting the increase
of the forcing towards the boundary. So the positive shear of the negative {}homogeneous solution must
prevail, since the overall shear should be positive. Physically the strong slip of the particular solution at the
boundary from the forcing generates strong homogeneous shear, dominating the weak reverse shear from the
decrease of the particular solution away from the boundary. So from (10) and (14)
at y=0 dyv1=anas(√t)n+s+1
[-4n/ n-1(s+2) - n/n+1+n+s+2/ n+s+1{4/(s+2)+½ns/ n+1s+1}] (15)

Comparing the two particular wall shears for n=s, 4n+1/ n-1 =2(n+1) vs. (n+2) so the first particular term
dominates with increasing n from equality at n=0. Regarding Figure 2 as relative velocity profiles it is clear that
the shear of the homongeneous normalised i2n+2 dominates the first particular and so the sum even though it is
less than the shear of the square of in+1 in the second terms.
Evaluating the w1 perturbation requires further
∫ y
∞
gdy=ann (2√t)n+1
in+1 d g=-ann(2√t)n
in-1 , dt G= ½ nan√t n-1
,
∫0
t
∫0
t
Gdt=4an√t n+4
/ (n+2)(n+4), ∫0
t
g=∫∞
y
∫∞
y
gdy= ann (2√t)n+2
in +2, dt g=d2
y g=ann (2√t)n-2
in-2 ,
d2
ty g = d3
y g=-ann (2√t)n-3
in-3, ∫0
t
∫0
∞
gdy=ann (√t)n+3
/n+1/ n+3 ,
∫∞
y
∫∞
y
gdy=ann (2√t)n+2
in+2 ∫∞
0
∫∞
0
gdy=an (2√t)n+2
/ n+2
so that
w1p=anas(√t)n+s+2
[2n+1
nin-1/(s+2)-2n+1
nin-2/(s+2)(s+4)-2n
nin-1s/ s+1(s+3)-2n+s+1
sisnin+2] (16)
The first term dominant at average  vanishes at the boundary leaving the negative corrections terms with
positive shear. But the positive shear of the dominant term means the negative shear of the positive
homogeneous terms is overcome. w 1= w1p - w1h Here the wall shear is dominated by the particular solution
increasing with distance from the boundary like the forcing just overcoming the weak homogeneous shear from
the weak slip of the particular solution , the opposite situation from v1.
w1p= w1h=  anas√t n+s+2
[ -4n/(s+2)(s+4) -2n s/ n-1s (s+3) -½n/n+2] at y=0 (17)
w1h= - anas(2√t)n+s+2
[4n/(s+2)(s+4)+2n s/ n-1s+1(s+3)+1 / n+2]n+s+2 i n+s+2 (18)
At y=0 dyw1h =  anas(2√t)n+s+1
[ “ “ ] n+s+2/ n+s+2 (19)
dyw1p =anas√t n+s+1
[2n/ n-1(s+2)+8n/n-3/(s+2)(s+4)+ns/s+1/(s+3)+n/ n-1/(s+2)+½n/n+1] (20)

This has a very high degree of reduction by the homogenous shear so the net positive shear is as expected about
1/10th
of v1 ‟s declining similarly with n. So in the combined axial acceleration problem v1 will totally dominate
the shear, as anticipated due to the w1 forcing peaking low in the middle of the primary boundary layer.
The aortic systolic flow pulse was represented by a quartic polynomial in t by and the above diffusion
approach was extended to higher order and to the aorta entry by beginning diffusion at the t≥0 when the Oseen
.35 times the inviscid flow left this leading edge of the wall. The 2 first order perturbation and the linear  terms
then agreed to +/- 20% with perturbations of the Blasius solution. [1]
But the sharp curvature in the aortic arch makes the v1 and higher secondary flow perturbations very strong,
indeed too strong at the inside of the strongest bend where the secondary flows converge to a stagnation point,
and so thicken the boundary layer that it separates. Actually this begins just downstream of the maximum bend
where the axial coreflow is also decelerating longitudinally on the inside of the bend. The separation of the
boundary layer in the computations was seen in ultrasound measurements in humans by Perroneau[3].
V. Stokes Transformation, Layer, and Steady Streaming
The Stokes „layer‟ S= ert+y√r
, dy S=S√r D[S]=0 solves the diffusion equation with complex √r
If any solution at all D[Q]=0 is expressed in terms of t and  ie as Q(t,) and „ again denotes differentiation wrt
the second (similarity) variable =½y/ √t
tQ½Q’/t¼Q’’/t=0 again. Now if =√ rt then dt=-/t½√r/t=-/2t√r/t so
D[SQ(t,)]=D[S]Q+SD[Q]-2dySdyQ=S{tQ-½Q’/t-¼Q’’/t) Q’√r/t }-2rSQ’dySQ’√r/t-2SQ’½√r/t
So “Stokes transforming‟‟ any solution Q(t,) may give one other solution Q√r=S Q(t,√rt). Transforming
again at -√r reverts to the original Q(t,) or at √r just transforms once at 2√r . Since (√rt)2
= 
y√r +rt ,
Sexp -(√rt)2
= exp-2
, the Qj of negative index aren‟t changed 1
but for all the Qn of positive index
Qn,√r=ert+y√r
(2√t)n
inerfc(√rt) are distinct solutions. A less general, more arduous construction is
Qn,√r=ey√r
∫∞
y
e-y√r
Qn-1,√r dy
Note erfc(z) + erfc(z) =2 so the solution for G(t)= ert
complex exponential starting at t=0is 2

g=½(Q0,√r +Q0,-√r)= ½ert
{ey√r
erfc(√rt)+ e-y√r
erfc(√rt)}. (21)
y integrations of Q0, √r=ey√r
∫∞
y
dy e-y√r
𝒆− 𝟐
/ √tfor the particular solutions are possible by parts.
Eg. ∫∞
y
gdy= ½(Q0,√r -Q0,-√r)/√r , ∫∞
y
∫∞
y
gdy=½(Q0,√r +Q0,-√r)/r+ Q0 /√r
So ∫∞
0
gdy= ert
erf (√rt)/ √r as in Fig 3 but upon squaring this has defeated exact homogeneous solution v1h .
But asymptotic ones will be considered after clarifying the asymptotics of g and how real g relates to the
complex. For large│rt│>>or t >> ½y/│r│ erfc(z)≈ 𝒆−𝒛 𝟐
/ z√ (1-½/z2
+..) for large z [5] . Then
g≈ ert-√r y
- Q-2r - Q-4r2
+…(22)
where the second term as t -1
will asymptotically dominate for all r with a negative real part. As Q-2=dt Q0 =
dterfc it represents the languishing as t -1
at diffusing distance of some effective net g from small t. For
positive real part of r2
or exponential growth g ≈ S≈ ert-y√r
the Stokes layer has an oscillatory component in y
if G has one in t. For pure oscillatory imaginary r at k‟th harmonic r=ik. Choosing and henceforth implying
the Real part by a cos t (impulsive) start avoids the imaginary Q-2r so the correction is - Q-4r2
as t -2
which
doesn‟t complicate the study of the asymptote to steady streaming. Then at large t
g≈eit-y√i
+Q-4≈e-y/√2
cos(t-y/√2)+Q-4dyg=-e-y/√2
cos(t-y/√2+¼)+Q-5, ∫∞
0
gdy≈-cos(t-¼)+Q-3
1
though transforming Dawson‟s integral solves the unbounded G(t)= i t –½
erf i√rt .
2
Recursion gives i1erfc(z)= -zerfc(z)+ Q-1 so even in z g= (Q1,-r- Q1,r)/ 4√r solves G(t)= tert
from t=0

As in Fig 3 the transient between ∫∞
0
gdy= Re{-eit
erf√it)/√i} and -cos(t-¼) can alternatively be uniformly
approximated by the sum of two exponential decays .18e-.4t
+.5e-4t
. The square of this accurate ∫∞
0
gdy
would give v1h at y=0 in complex exponentials again, which can be solved
Figure 3 Approximations to the y integral of g for G= cos t from t=0
by the above with complex r. Including these decays would give small corrections to the following coefficients
of Q-2 or t -1
asymptotes which arise entirely in the homogeneous problems. With the asymptotic g, the
solution v1 is straightforward to determine
v1≈2 e-y/√2
cos(t-y/√2) sint + ½ e-y√2
cos2
(t-y/√2- ¼) -5e-y
sin(2t-y)/8 - ¼ erfc-5Q-2/8 (23)
The square of the y integral of the primary Stokes oscillation in y and t thus creates a mean particular flow at the
wall whose homogeneous reaction diffuses to infinity as a Q0=erfc. The second harmonic homogeneous terms
carries a Q-2 or t -1
correction from (21). Because the particular solutions asymptote faster as t -2
there is no
point in using the exact ones, but those do show the asymptotic particular solutions below do not need to be
corrected for initial values. From (10) and (23)
dyv1≈¼√2+(5/4√2-3/2)sin(2t+¼¼Q-1-⅝Q-3≈.353+.267sin(2t+¼¼/ √t-5/ 16√t 3
(24)
So the mean wall shear decays as 1 /√t to an ultimate positive wall shear of mean ¼√2 with second harmonic.
Thus the forcing in the Stokes layer is met by a particular solution confined to the Stokes layer but slipping at
the wall. That generates vorticity / a homogeneous solution to prevent slip whose mean diffuses and spreads all
the way out to the outer flow in time.
The mean shear is just the time average of the y integral of the forcing, whilst as eqn 5.13.19 [4] the streaming
mean velocity is the time average of the y moment of the forcing. Now the v1 forcing is strong and near the wall
whilst the w1 forcing is weak but further from the wall so its streaming mean velocity can be much more
significant than its shear.
The constant values in (11) at t=0 multiply homogeneous solutions, so these are exactly cancelled by their
counterparts in w1h, and all that briefly survives are the t -1
asymptotes due to these initial values. (This did not
arise in the power of √t basis as all t integrations are automatically zero at t=0.)
The highest error term in w1p comes from the highest integrated term ∫0
t
∫0
∞
gdy of Q-1 multiplying dy g to
give a term at the wall as cos(t- ¼)/√t whose homogeneous solution is unfortunately unknown. How the
running mean of this boundary value decays should be how its mean diffused value would decay. The running

total is ∫t
cos(t+¼)/√t = (FresnelS-FresnelC)√2√t which tends to 0 so there is no mean transient 1/t response
to add to those from the homogeneous pure harmonics (that do not start as cos‟s). So then
w1p≈ye-y/√2
cos(t-y/√2+¼)sint-2e-y/√2
sin(t-y/√2)cost-½e-y√2
sin(2t-y/√2)+½e-y√2
cos(y/√2)-¼e-y√2
sin(2t-√2y)
w1h= -7e-y
sin(2t-y)/4+ ½ erfc+13Q-1/8 (26)
Here the mean of w1p’s dy g ∫0
t
∫0
∞
gdy’ is twice as strong at y=0 as the mean of v1„s half square ½ [∫∞
y
gdy]2
.
Thus each wall mean just comes from one in- phase “square” term , and w1p’s is clearly if suprisingly twice as
strong as v1p’s. Equation (13) is more direct than (26) for some of the shear terms.
dyw1≈(5/2-7√2/4) sin(2t+¼–½Q-1-13Q-3/8 ≈+.025sin(2t+¼½/ √t-13/ 16√t3
(27)
The second harmonic component is about 1/10th
of v1‟s , as expected from section I. Three terms in w1p, have
ultimate mean shears but they cancel. Thus the ultimate mean w1 flux into the y=0 sink vanishes so there is no
net mean w1p forcing integrated over y. The higher steady w1 streaming comes from more of the mean w1
produced further from the wall sink than v1 diffusing to infinity. The w1 mean forcing from -dy g ∫0gdy}
decays as e-y√i
whereas the v1 mean forcing from -g2
decays as e-y√2
so the former has twice the y moment and
so twice the steady streaming. 3
dy(v1 +w1) ≈ . 353+.292sin(2t+¼¾/ √t -9/8√t3
(28)
Figure 4 Perturbation wall shears computed for G=sin t from t=0 by Taylor series. NB different shear scales.
3
[4] twice attributes steady streaming under longitudinal acceleration to just the w1 momentum transfer by mean
udYw but calculates the net from v1+ w1. Though a 50% enhanced w1 is indeed solely responsible for the phase
steady streaming in -¾WT(dXW+WdX) under a wave coreflow WRe{eit+i
}in Batchelor‟s (5.13.20).

The combined shear is practically dyv1 except the transient shear from the outward diffusion of the steady
streaming is trebled. In Taylor series calculations for G= sin t from t=0 convergence was accelerated by the
Pade method[6,7] . Fig 4 shows the high steady streaming transient shear of w1 declining slowly as t–½
 to the
low ultimate second harmonic shear. The transient first harmonics in w1 decline as t –2
. The large ultimate
mean and second harmonic dominate the small steady streaming transient in dyv1.
The support of a Commonwealth Scholarship is gratefully acknowledged.
VI. Conclusions
The perturbation of a time-diffusive boundary layer for longitudinal acceleration is the sum of a
pressure gradient perturbation large near the wall and a „indraft‟ perturbation with a modest peak in the middle
of the boundary layer. The first gives an order of magnitude greater shear, but the second has twice the steady
streaming flow which spreads outwards as the complementary error function of the diffusive similarity variable.
This gives twice the transient shear declining as the inverse square root of time. The next order transient
outward diffusion is an inverse time function of the diffusive similarity variable. For moderate time, a Taylor
series in time numerically converges for almost 5 cycles of a harmonic. For any such (root) time series, the
general particular solutions found are easily matched with exact homogenous ones to complete an explicit exact
perturbation solution.
References
[1] Farthing S., Flow in the thoracic aorta and its relation to Atherogenesis Ph.D. thesis 1977 University of
Cambridge, Cambridge England
[2] Cunningham K.S., Gotlieb A.I. The role of shear stress in the pathogenesis of atherosclerosis. Lab
Invest. 2005 85(1):9-23
[3] Perroneau , P. Flow in the thoracic aorta. Cardiovascular Research 1979 13 (11) 607-620
[4] Batchelor, G.K. (1970) An Introduction to Fluid Dynamics Cambridge Univ. Press p.358
[5] Abramowitz, M.; Stegun, I. A., eds. Handbook of Mathematical Functions with Formulas, Graphs, and
Mathematical Tables, New York: Dover Publications1972,
[6] Shanks, D. "Non-linear transformation of divergent and slowly convergent sequences", Journal of
Mathematics and Physics 1955 34: 1–42
[7] Wynn, P. "On a device for computing the em(Sn) transformation". Mathematical Tables and Aids of
Computation 1956 10: 91–96.

Exact Perturbed Unsteady Boundary Layers

More Related Content

What's hot (20)

Similar to Exact Perturbed Unsteady Boundary Layers (20)

More from International Journal of Modern Research in Engineering and Technology (20)

Recently uploaded (20)

Exact Perturbed Unsteady Boundary Layers