Flutter of a Finite Rectangular Wing’D Mill in Pitch and Heave

S.P. Farthing, Wing’d Pump, www.econologica.org, N. Saanich B.C. V8L 5P2 Canada spfd@cantab.net Proc. IMechE A: Journal of
Power and Energy. submitted 17 /03/20 accepted 28/07/21 published 14 /06 /22: 1-14

Exact solutions of flutter are explored for a mode suitable for an oscillating 2D hydrofoil version of the summarised FlutterWing
windpump which radically surpasses the old rotary fanpump. The imaginary part G of the 1D Theodorsen wake lift factor T rises
singularly from the steady to allow 1D pure pitch flutter around the leading edge at low frequency. But its logarithmic infinite negative
pitch damping by the shed vorticity is capped by the log of the 2D span. The downwash of an oscillating trailing (tip) vortex is reduced
by a function of the reduced frequency based on spanwise distance. Its lag damps pitch and further raises the pitch unstable aspect ratio A
to above 36. All binary pitch & heave neutral flutter frequency contours in (inertia, imbalance) space still pass through a nexus of the
same total pitch inertia and imbalance as the corrected virtual mass concentrated at ¾ chord, a very high total in water. The internal
bound windwise vortex from ¼ chord to trailing edge reduces as A-2 the 1D lift slope at all frequencies. It expands the 1D T=½ double
root of high frequency elliptical contours kissing the low frequency limit line at the nexus to make them dip below the line to the right of
the nexus as hyperbolae with the nexal ray as left asymptote. Thus low A discovers a new wedge of high frequency binary flutter with
real mass moments less than the virtual, as in water.
Keywords: FlutterWing Pump, Theodorsen, pure pitch flutter, binary flutter, pitch & heave, pitch & roll
Introduction

As early aircraft increased in speed, novel self-amplifying coupled oscillations between control surfaces and bending wings or twisting
rear fins caused many structural failures and crashes[1]. For instance the inertia of an unbalanced aileron acting at a cg behind its hinge
makes it overshoot the wing at the top of a bending oscillation, so dropping the wing lift to amplify its rebound down. By 1960 flutter
instability had been eliminated by dynamic massbalancing and stiffening controls and airframes, checked by extensive linear
computation of the myriad of flutter modes exceeding all other aircraft design calculation. But the voluminous linear computational &
experimental flutter airspeed literature did not even note that there can be two roots. Proving flutter could cease in high winds was the
first step towards pursuing flutter windpower. (Sir James Lighthill FRS had stressed the need for storm protection) . In 1980 storm
stability of pitch and roll flutter had been hypothesized and proven successively with models on top of a car and then in Southampton,
CEGB Marchwood and UCL windtunnels, by computations at Gifford & Partners and later in full size prototype #3 in gales and finally
algebraically [2]. Very crudely the pitch swing to the inclination of the tailheavy wing is decreased by the rise in pitch damping as
windspeed so the pitch amplitude decreases with windpseed, avoiding high wing loadings in the windmill low drag equivalent of the
high thrust efficiency of oscillating propulsion in nature. Gust response is neutral at cutout and is more and more damped above, with
the wing feathering to the mean wind at midspan whilst inclining away from the roll torque of any shift of the upper wind. (However a
sweptback wing doesn’t cutout but basically diverges in high winds).
Duncan [1] had pedantically demonstrated that the airstream powers flutter’s unstable oscillations with his model “wing engine”, a
balanced wing articulated to pitch in set quadrature to heave (plunge). The BBC 1976 Young Scientists’ more practical pitch and roll
(Fig 1) free flutter model lacked a pump to prove better wind waterpumping than fanmill windpumps without the dangerous climbs.
(Pocklington School attributed the idea to Sir. G.I. Taylor, but his biographer G.K Batchelor FRS could not find it in GIT’s papers). In
1980 my sprung model with a cam track to depress a honed piston pumped to high psi. The power varies with amplitude squared in
linear theory, but pitch saturates at ±90 and ultimately the power must vary as large sweeping roll amplitude. In 1993 full-scale smoke
[3] found a beneficial extraordinary leading edge vortex shed at full ±90° pitch flip. Definitive confirmation of economic power far
outweighing the fatigue penalties of 40% with steel and 20% with wood came with this fullscale prototype base #7 on a test well with
instruments read by a computer
in a trailer at the BC Science Council. Altogether 8 bases and 6 wings were prototyped. Scaling shows all inertially unbalanced fluid
machines (incl. H Vawts and orthinopters) have an upper limit to design flowspeed as the square root of the strength to weight, the
endurance limit stress/ material density[2] and so the FWP niche is light wind (and also gravity-dominated large size)
Fig 1 Perspective of Floating Flo’Pump with pump down and up stroke cutaways

The rotary multiblade windpump loses all the kinetic energy of the swirl reaction to its high torque [4], which still can’t turn over its
crank to start in the best wind for its stroke. For a typical wind regime it can only usefully pump just 10% of the work of a Betz ideal
windmill [5]. Duncan's fixed cycle wing engine would likewise stall against the fixed head of a piston pump. The free amplitude of
flutter instability is inherently suited to reciprocating a pump whose stroke needs to vary to efficiently capture the changeable wind. The
rectangular Flutterwing free to pitch (360°) on top of a roll pendulum winds a spiral winch pump (Figs 1 &2) with stroke varying as
almost the cube of roll amplitude. So the pump stroke increases strongly with roll, not loading a starting swing in only 1m/s, but still
absorbing up to ½ of the Betz wind energy swept in a big roll in a good wind.  Compressing an aircylinder on the return stroke is very
simple but not so supra-linear so a flow rate of about 3 cfm to a pressure of 4 atm for 250W was the max for a total 400 peak pumping
W (eg 4L/s @10m head) in 3.5m/s by an 5.5 x1.37x.2 m 20 kg wing. Wind gradient and the non-linearities of the stiffening of the
single-acting pumping, large amplitude pitch, and crosswind roll speed exceeding windspeed all increase the power[6] and the cutout
causing a hysteresis vs. the linear restart wind. Key to curbing any overswing is that flutter pitch and roll are partly in phase so the
Flo’Pump‘s tip balance rod makes first contact with the pond to wet its absorbent material and temporarily reduce the tailheaviness that
flips the wing. A simple cusp track ensures self-starting from small amplitude in very light and steady winds but misses wing cam
contact at bigger amplitude or when excess winds blow it back[7]. The Flo’Pump prototype Fig 1 floating around its pump cylinder
with upper air chamber pumped through its underwater outlet pipe to shore and overland powerfully and reliably for 10 years with its
catamaran righting moment limiting the stresses. Cantilevering can be eliminated by putting the pendulum inside the frame with the
winch turned around and the pump and yaw axis better in front. It needs help with commercialisation or adoption in the Third World or
making very large prototypes to assist pumped hydro.

Fig 2 Schematic of the Flutterwell Pump, the Well-mounted FlutterWing Pump

The Flutterwell base in Fig 2 uses the steel well casing as a short foundation pile. A new model8 has shown
prolonging the nose down pitch in a roll overswing reduces the wing flip so the next opposite roll is safely smaller. The
arm that is manually pulled to enter a cam track on the outside of the lower pitch collar to center the wing to stop it,
could be automatically pulled by roll overswing7 so new side tracks govern the wing flip back (Just trying to center
the near peak of pitch at peak roll unfortunately assists/initiates the pitch flip). After a long calm, the high well water does not restrict
the roll of the wing but when it passes horizontal nosedown, balls roll down the nose track reducing the imbalance and the flip and so the
return swing. 8. Then the Flutterwell could pump enough easy large, but safe, strokes at lowhead to drawdown the well and a high static
watertable to a sustainable yield. The well storage drawdown transient can be minimised by inflating a ring seal between the rising
main and the well just above the strata yielding water. Some of the wind’s varying power is inevitably wasted in the varying head of
groundwater resistance. A more radical solution is to raise the tower for ground clearance at max roll of say 165. No-load model
releases from this low slow zone with the wing cam-centered do not reattain 165 for several cycles.
The novel differential area pump keeps the pumpwire in tension for downward accelerations up to its upper/lower area ratio>1 times
Newton’s g. Pulling its plunger with all the seals and valves out of the cylinder releases the water column for easy retrieval and
maintenance. Its design for low internal flow constriction Fig 3 uses Fig 1’s single part seal valve this time just above a perimeter rubber
flap valve over a grid of holes, and a ring seal inside the bottom of the plunger. A snifter, NRV and airtank have been tested at the
wellhead for overland pumping overcoming air absorption and indeed outputting some compressed air. A spinoff is a powerful well
footpump[9]. But a fluttering watermill is elusive, only demonstrated to moderate amplitude via the dominance of static gravity
imbalance in pitch and roll in water[10]

Fig 3 Detail of the actual differential well windmill and footpump

Despite being ubiquitous in early aircraft, flutter has scarcely been a problem in marine hydrofoils. The ratio of foil mass to the virtual
mass m of the circumscribing fluid cylinder is <<1 with the 700x water density whereas aircraft wing mass ratios are >>1. Pitch and
heave flutter calculations of typical hydrofoils by eminent aeroelasticians [11] backed up their empirical lower flutter limit of roughly
unit ratio. An upside down version of Fig 1 with two opposed ferrocement blades could be not be adjusted to flutter when towed through
water.
So the basic heaving fluttermill of chord c, virtual mass/unit length m, that is free to pitch  about an axis ec ahead of the ¼ chord
aerodynamic center has been analysed in a series of papers [2,12,13]. The lack of mechanical stiffness allowed a first ever algebraic
solution of binary flutter drawing all the neutrally stable frequency  contours in total mass imbalance xmc vs pitch inertia jmc2 space.
All contours pass through a nexus N=(mo2,mo) of the same total inertia and imbalance as just m mislocated (too far aft) at the ¾ chord
point o behind the pitch axis [12] where the nominal upwash U¾ and so the wake and circulation vanish for all . The apparent
mass/span of the finite (thin) wing is lm, l≈1-½A-2. Then a priori for a 2D ¾ chord line offset o to the pitch axis the flow is pure
potential implying neutrality of any flutter at N =(lmo2, lmo) because again U¾ =0 everywhere by the heave equation
Actual hydrofoil virtual and real (structural) j/x are closer to the pitch axis and so are on the left side of the ‘ray’ from (0,0) to N.
Even an understressed solid steel hydrofoil has x short of the low frequency limit line 'lfl'. With pitch axis lead e>0, this quasi-steady
“lfl” line through N extends downwards in x with j to reduce the tailheaviness x for flutter (even to negative x or noseheaviness.)
Generally the larger flutter zone of higher m, j and j/x (partially mass-balanced wings) to the right of the nexal ray and above the lfl
allows more heave/pitch for heavier-than-air Wing’d Pump power with high wind V cutout in pitch and roll.
The binary discriminant of this biquadratic in j and x is proportional to the discriminant of the quadratic in e for pure pitch
flutter[13] :

4F2 /k4 ==(2g-F-½)2-2F ={2g-F-½-√2F} [2g-F-½+√2F] = {2g -(√F+1/√2)2}[2g-(√F-1/√2)2] (1)

where T=F-iG=F-ikg Theodorersen’s famous 1933 1D wake function of reduced frequency k=c/V. Now as k ↑∞, F↓½+¼/k2 and
G↓¼/k so by extraordinary cancellation √2g >√F-1/√2 making the 1D high frequency contours ellipses, from N to 4N[13], kissing the
lfl line at N). Any 2D change from this special 1D lower limit of F of ½ will open the high frequency contours into hyperbolae
crossing the lfl line. (The back induced flow is ∫0∞d(x)/ 2(x+c/2) or by parts -(0)/c in this limit, the same as in a Wagner sudden
angle of attack so the induced fraction I=1 and the lift ½ the steady in both cases[13].)
As k↓0 a negative Ln singularity in the 1D g and so in the net pitch damping allows pure pitch flutter at kz≤.087 of {}≤0 for <0,
(with very airfoil high pitch inertia and the wake vortices fed by the drag work.) This hyperbolically repels the k≈ kz binary contours,
even bizarrely back to very large negative imbalance but no inertia at about ½kz . [13] Here 2D finite aspect ratio A is added for a vital
reality check.
2. 2D Aerodynamic model
Jones[14] fitted the Wagner impulsive start of wings down to A =3 as sums of exponential decays strengths C0 & C2 whereas in 1D the
Wagner clearly levels as inverse time with the retreat of the shed vorticity. Fourier transformed to the oscillating wings T(k) N  f - f /
2C0(1-2iC2/k) where C01+2.32/A ; C2.181+.772/A and f is the 1D measured lift slope/ the theoretical 2further reduced for 2D
by N(A)=1+Iq Because of the incorrect 1D large time asymptote, the low frequency G(k) singularity (and the linear variation of F(k)
near k=0) are missed. The literature has many complicated point 2D computations but very little insight on 2D aerodynamics or flutter.
The circulation terms alone extract power from the flow. Classic potential flow gives the exact even non-linear virtual mass terms.
But how to generalize m‘Vcq in the 1D pitch equation [13] needs study. As dissipative damping (necessarily into the wake) it
must be circulatory in origin but lacking a T factor is often wrongly taken as a virtual mass effect, e.g.[15]. Instead of it potential
theory has a large broadsiding moment in the apparent cross current Ue at the pitch axis. I [2] postulated the difference as a circulatory
“broadsiding” in the ¾ chord upwash U¾ = Ue+‘o where o=qc q=e+½, fitting with no circulation if U¾ is always nil. Andrienne
derives the linear circulatory moment as -V[m +ec2T]J where J is an integral of the vorticity over the entire wake which Theodorsen
indeed proved is U¾ for any time variation of the wake[16]. Since the potential broadsiding moment linearises to mUeV, then the 1D
net broadsiding is simply - m‘Vcq. In 2D these broadsiding m should all then just correct to lm and so m‘Vcq to just lm‘Vcq
Perhaps just taking m‘Vcq as inertial, Jones also corrected its added mass m by l for aspect ratio. Here, for the pitch balance to be
cqU‘¾ and so nil from the heave balance [13]with its U‘¾ =0 mode at the a priori no-lift nexus N=( lmq2c2, lmqc), flutter requires the
heave inertial (‘Coriolis’-like) force lm‘V must by matched by an equally l lower lm‘Vcq moment, at least at N. Assuming this
henceforth everywhere, then the general 1D flutter solutions [13] are conveniently retainned if L≈1+½A-2 multiplies j,x &T.

.The A=4 Flutterwings are statically stable in pitch at 23.5% chord slightly ahead of the 23.7% cp from O(A-2) lifting surface theory
and experiment[17], so that 1D cp will be retainned as trailing vortices induce more the further aft of their origin at ¼ chord induction
shifting the net loading and so the static cp is forward of ¼ chord as O(A-2). Adrienne’s remaining 1D moment term -ec2TVU¾ is
just the restoring moment circulatory lift cVTU¾ about the cp where Theodorsen’s T=1/ 1+Is. Is =K0(v)/K1(v) with v=½ik and K
modified Bessel functions of the second kind. I is the fractional loss (induced=apparent-net) of angle of attack (vs the net) due to the
wake. (Fig 4), like the constant Iq =N-1 along an elliptical wing in steady lifting line theory long extended empirically for rectangular

wings to O(A-2) and further here to oscillating flow as It. Then the average trailing It and shed Is downwashes can be summed for a
combined loss, a logical symmetry missing in Jones.
Trying to explain in words the 1D Wagner and T(k) wake effects, led [12] to Is≈vevE1(v ) as induced by an oscillating wake at
midchord for a much better 1D exponential integral E1 approximation to T than Jones’, reproducing the singularity and high k limit. So
then C(v) = K0(v)/ K1(v) vevE1(v ) will be a correction factor to apply to the 2D generalisation of this Is below.
So keeping the trailing and shed parts of the vortex loops distinct, to evaluate their wash at different chordal points as in the steady
lifting vs 1D unsteady limits, as in Fig 4, is more workable and instructive. This difference from neither solving the integral equation of
wash (and vorticity) all along the chord is that the Is downwash of the vortices shed at the trailing edge is evaluated at midchord, not at
¾ chord where Weissinger[17] evaluated the downwash of vortices trailing from the ¼ chord. These trailing vortices are the drivers of
the non-uniform downwash, higher outboard; whilst the shed vortices between them aft the t.e. have slightly higher downwash centrally
for likewise different locations of mean I along the span. Only the steady has used the measured 2D static lift slope factor a0, which
depends on thickness (and section) and decreases with roughness or low Reynolds number. But these static boundary layer effects may
diminish around k the naïve scale for Stokes boundary layers. So to not wantonly affect the flutter-critical 1D T→½ limit, the lift
slope a will be tapered weakly as a=2f (a0+2kk/So the 2D lift is still c2VU¾ T with now T=f/(1+It+C(v)Is) .

Fig 4 Oscillating Wing Vortex Loops with induction distances and averages, and wind angles.

3. Induction by a Single Trailing Vortex Line

In upper Fig 4, The trailing edge divides each 'trailing' vortex into in-phase (bound) internal and lagging external segments. Their
induction is just evaluated on the lifting line in first order theory, but at ¾ chord w=0 for second-order O(A-2). y inboard a unit
internal trailing vortex from the t.e w=- ¼c to the ¼ chord w= ½c (the lifting line) induces DI
4yDI= ∫dwy/ (w2+y2)3/2 =| w/ (w2+y2)1/2 ≈ ¾c/(¼c2+y2)1/2 (2)

where the c2 term in the denominator is approximate O(A-2). Whereas each external trailing vortex progressively lags as e-ix/V with x
behind the te, for induction per unit eit at the ¾ chord point of DE= ∫0∞dxe-ix/Vy/ 4((x+¼c)2+y2)3/2 where the denominator will be
expanded out (x+¼c)2+y2=x2+y2+½cx+c2/16 with the last term to be dropped as net O(A-3), so ((x+¼c)2+y2)-3/2 =(x2+y2)-3/2–
¾cx(x2+y2)-5/2. Changing variable of integration to = x/y the angent of the enclosed angle, and letting S= y /V or the reduced

frequency based on spanwise separation y, the integrals become

4yDE≈R(S)-¼M(S)c/y if R(S)=∫0∞de-iS/(2+1)3/2 =SW-½iSL-1 M(S)=3∫0∞de-iS/(2+1)5/2=-iS2W-½SSL-3+4L-2} (3)

with WS) = K1(S)+½ iS) where K1, are Bessel, and L-1 modified Struve functions from [18], 9.6.2.5 and 2.2.3 [19]. These
unsteady factors R and M graphed in Fig 5 do not seem identified and studied before. (Note the strong decline initially with S of their
real values R(–iS) and M(–iS), for divergence at growth rate –iS.) Expanding oscillating e-iS ≈1-iS R(S) (and M) is 1-iS at small
S, so the first unsteady perturbation to DE is spatially uniform downwash -i / 4V inboard of the vortex (per unit eit trailing
vortex). The closest 90 behind phase is dominating for a phase lag, (vitally opposite to G’s phase lead of the time derivative ‘just’
shed downwash.) The negative imaginary parts of R and M peak at S1.5 at twice and three times 1D G’s peak at k= .35. Then Im{M}
declines much faster than Im{R} making the difference Im{R(S) -¼M(S)c/y} decline slower. This combination is the first two terms
of an alternating series whose imaginary part must be positive so on the imaginary side (only) extend as
4y DE = R/ (1+¼Mc/Ry). At large S integrating e-iSby parts, R→-i/S + O(iS-3) (with real part again decaying exponentially)
whereas M →-3/S2. Real 4yD is totalled as ≈ R+(3-M)c / 4(y2+ c2/4)1/2

Fig 5 Trailing vortex induction factors vs reduced frequency S (growth rate–iS) based on spanwise y
The first order steady lifting line theory ( See pg 8) has principal value integrals of 4D=1/ y just workable by Fourier span
decomposition of the trailing vortex variation but their normal Fourier solution by integration would involve the PolyGamma function.
Instead Glauert introduced solving the steady circulation series equation (6) numerically at discrete y points. Even the simple analytic
solution for an elliptic wing is lost with the above uniform i downwash of R. Also the corresponding shed induction must be
included, but a windwise integral similar to R &M of a strip of perpendicular shed vorticity steps dy long is elusive analytically.
Numerical solution of a second order unsteady lifting line integro-differential complex equation would need singularity-free
formulation and was too big a computing diversion for analytical and physical understanding of flutter for waterpower. Instead
advantage was taken of the rectangular wing geometry to study a simplified discrete trailing vortex model [20] vs. Jones.
Rectangular wings (semispan s) concentrate their steady trailing vorticity near their tips at high A=2s/c. So ignore the widening of
the trailing vorticity but just move these tip vortices inboard to ±ts, cross linked by shed line vortices, and evaluate the 2 now unsteady
inductions spanwise where each is mean, as in lower Fig 4. The unsteady shed induction should have a parabolic decrease in y for
symmetry about the centerline so its average will be taken as at y=s√⅓ =.58s from centerline. The induction there of each 2ts long shed
vortex line is integrated next (alas not its windwise Is integral over wake distance x except at the k limits.)

4. Induction by Shed Vortex Lines

 So our shed vortices are simply from – ts to ts of strength i e-i x/Vdx/V where =½acV the bound vorticity varies like as
ei t. Their backwash fraction I of is induced at midchord Is=½aik∫0∞dxe-iwx/V(x+½c) ∫ dy / 4((x+½c)2+y2)3/2 (4)
The inner ∫ in dy along the length of the shed vortex y gives |Y|/ 4((x+½c)2+Y2)1/2(x+½c) with abs value |Y|= sz; z=t±√⅓
to have the not another confusing  Then Is=½aike½ikJ/4 where J=e-½ik ∫∞0dxe-iwx/V |Y|/ (x+½c)((x+½c)2+Y2)1/2 Non-d
wrt to semichord b, v=|Y|/b, and use u semichords from midchord bu=x+b . ThenJ(v)=∫∞1 du e-½iku v / u(u2+v 2)1/2
So for AY, &v →∞, the 1D J=E1(v) Is=fvevE1(v ), when v=½ik→0, E1(v)=∫∞1du e-vu /u→-Ln(.89k) >0

5.1.11 [19] and .89 ½ Ln of Euler’s gamma (.577). [12] reproduced this limit by taking cos as 1 up to .89ku=1 and 0 thereafter so
likewise J → 1/.89k ∫1 v du / u(u2+ v 2)1/2 →1/.89k |1 -Sinh-1[v /u] →Sinh-1[v]-Sinh-1[.89 k v]
kv is the anticipated ratio [13] of span versus the wake wavelength . If this kv >O(1) by very large v »O(1) then 1D J→-Ln[.89k].
Conversely at kv <O(1) or span width << the wavelength by the same Sinh-1 x=Ln(x+√1+x2), J→Ln[2v]-.89kv. With v =(t±√⅓)A,
the shed backwash fraction is at small k, Is=½aikeik/2J / 4 ≈¼ikf Ln{4(t2⅓)A2} (5)
So the 1D logarithmic singularity Im{Is/k} as -Ln k ↑∞ as k ↓0 is limited by Ln A in 2D.

General Is=f Q/2, Q=vev∫1 ∞du e-vu / u(u2A-2z-2+1)1/2 if z= v/A =y/s=t±√⅓. (remembering the factor of ½ in v) which checks down to
Is=fvevE1(v ) when A →∞, but is otherwise not known in general. Or Q=zvev ∫1 ∞du e-vu d(u) where d(u) =u-1(u2A-2+z2)-1/2 . Integrating
the exponential by parts first reverses the original time differentiation that got the shed vorticity and gives a smaller integrand
converging better numerically at large u for large k↑∞

Is=½f Q, Q= zd(1) + zev ∫1 ∞du e-vu d’(u) Q =z(A-2+z2)-1/2 +zev∫∞1du e-vu(2u2A-2+z2)/ u2(u2A-2+z2)3/2 (6)

Also continuing by parts Q= z (-1)n dn(1)/vn-1, where d n(1) is the n’th u derivative of d evaluated at u=1. When (u2A-2+z2)-1/2 is
differentiated it adds a factor A-2 so working to that A-2 accuracy, only it and its first derivative need be evaluated, yielding the simple
first few terms Q=1+2i/k+(2i/k-1)/ 2A2z2 -8/k2-48i/k3 with A-2 terms cancelling in k-2 & k-3. Then correcting to the expansion[13] of
K0(v)/K1(v)=1+i/k-3/2k2-3i/k3 gives CIs/f→1+i/k+(i/k-1)(t2⅓)/ (t2⅓)2A2 -3/2k2-3i/k3 (7)
always assuming v =|y|/b=Az >>1. Im{Is} is dominated by the vorticity shed at the trailing edge over that shed  phase earlier /k
whose counter induction is reduced much more by the finite length of the shed vortex; so Gs actually increases with
A-2. Whereas the inductions of vorticity shed / 2k and the opposite 3/2k ago, are more equally reduced by A-2 so Re{Is } induction is
reduced, increasing Fs.

5. Net finite wing influence at high A as k↓0

So at low k, It+ CIs =I0-½ifk +¼ikCfLn{4(t2⅓)A2}. Physically the Im part of It is from vorticity that trailed from the tip ¼ cycle
ago whereas Im{Is} is from the vorticity currently shed into the wake as the circulation changes. It offsets the shed Is negative pitch
damping by A< 5 but there is also the non-T damping not proportional to e. Check the 2D flutter discriminant (1) for G/k=g=½F+L/ 4+
√ LF/2 to see if a kc still exists for zero net damping making binary flutter hyperbolic >0 and allowing unitary flutter at lesser k at
some e >0 [13]. Now
F-iG= f (N-ifk{¼CLn[4(t2⅓)A2]- ½}) / (N2 +f2k2 {…}2) In the limit kc ↓0 and high A gJ =f/ 4C0C2N , C↓1.04
F= f/N , g=¼(CLn[4(t2⅓)A2]-2)f2/N 2 (8)
Figure 6 g and Pure Pitch Discriminant F factor vs Aspect ratio A at k=0 & default f=.95 N 2.13fA-1 +1.11f2A-2 21]clately.tif

Note the N2 reduction of g (vs the N reduction of gJ which incorrectly asymptotes in A lower than ½FJ +L/4 +√LFJ /2 missing 1D
unitary flutter entirely!) Fig 6 also shows a root of large A =36 at Theodorsen’s f=1 increased to A=42 at f=6/2vs glider wings
highest A <35, motorglider A <50. Dropping the trailing damping 2f2 from R≈1-iS in It more than halves this A to 19, so this new
damping of the trailing vortices is as important as the Ln A in g from the expected capping of the singular 1D g by the length of the
shed vortices. So the net A effect on [13] is non trivial and was best left to this new paper to find no g flutter in practical 2D, and
therefore none of its bizarre distortion of binary flutter contours at its very large j (and –x)

6. Approximate Net Influence of the Trailing vortices

In steady 1st order lifting line theory, the circulation is K(y)=½ac0VBwhere By) n sin nand y/s=-cosEach Fourier component
mode
sin nof K has a downwash Vtimesthe factor ½ac0n 4ssinnpsin  where p=ac0/8s, so the circulation equation is ½aV 
times
coB= consinn c {nnsinnp /sin} (9)
with the simple n=1 elliptical c=cosin  solution p Iq ) with Iq= p uniform downwash. The trailing vorticity varies

as dB/dy centered at ts=0∫s ydB / ∫dB =∫Bdy/ B0) so t=∫ Bsind/ B0)= ¼ B1 /-…) where the denominator is
the peak lift factor B0) at the center y=0. For a rectangular wing cco solving (9) at 4 evenly spaced values of  gave t as cubic
rational function in p= a/ 4A=f/2A, which corrected (in its smallest p factor offset) for t=1 for p=0 , and then fit/economized to (Fig
7) is t= (1+¼5.4p)/(1+5.4p) = A+6.36/ A+8.1 at a=6 so, first t=1-1.16 p and ultimately t is the elliptic ¼at low A.
Physically as p increases the influence reach of the p=0 tip vortex, its circulation loss is spread inboard.
The back induction ∫d/y is average at some rs inboard of the center ts of the trailing vortices. Each steady “tip” vortex has the
same strength ½ acVas the lifting line for mean back induction fraction Iq=½ac(1+½c/y)y) where y/s=t±r are the spanwise
y separations of the average evaluation point at rs from each tip vortex. Iq= p{1+ 1/ (t±r) A}/ (t±r)
so = Iq/ 2p= t/(t2-r2)+ (t2+r2)/A(t2-r2)2 for the negative root of r2 by the quadratic: r4+r2(t-2t2-1/A)+t4-t3-t2/A=0. The actual
rectangular wing lift fits[21] Iq  1.36p +.45p 2 (about 2A-1 +A-2 @ a=6) to give the r in Fig 7. Now It =½acD
It /p= R±+(3-M±)/ 2A(t±r)}/ (t±r) (10)
Then at small S±=y/V=½kA(t±r), and large A, R≈1-iS It – Iq ≈ ½ pA-i k (t±r)/(t±r)= –½ifk that uniform lagged downwash
independent of A. At big S=½kAy/s R±→ -i/S±→-2i/ kA(t±r) & negligible M±→-3/S2± , It →(3-4i/k)f(t2+r2/ 2A2(t2-r2)2
(11)
Thus the internal trailing vortices give a positive asymptote to It as O(A-2) So in the sum It + CIs the real fA-2 term is 3(t2+r2(t2-
r2)2/2 -(t2⅓)/ (t2⅓)2 >0 for small p so It overcomes the O(A-2) reduction of Is. in the computations below. Likewise It reduces the
dying G more than the increase from smaller Is in (6). Expanding Jones with It  2A-1 +A-2 gives FJ →½ +0.16 A-1 -3.5 A-2 +… with a
similar negative dominant A-2 term. Since the correction should be monotonic , the first A-1 coefficient should ≤0, probably 0 as
here.

Fig 7 Rectangular wing lifting line fit for t and r to fit empirical lift slope. (cps.tif)

The R & M reduction and phase change on the trailing influence can be used to factor an effective first order aspect ratio (actually
p) to refine t and r.This pe for the same magnitude of average N=1+ It from the dominant near tip vortex is pe=|(t-r)+[R+(1-
M)/{1+A2(t-r)2}1/2]p|-(t-r) where the 2nd order coefficient has been adjusted to ensure pe=p for small S and k, and to numerically
match (below) Jones asymptotic F. NB t and r are weak functions of pe however decided.
7. Aerodynamic Computations
At each A from the k=0 pe=p and t-r, R and M were calculated to get one refined new pe , r and t at each k from which It (7) is
evaluated and Is numerically integrated (4) or expanded (5) to calculate F(k) and G(k), and then move on to the next k Figure 8. pe
drops with k as expected and t and r rise but limited by the persistent induction of the internal trailing vortices. Comparing with the
Theodorsen, 1D GJ not curling vertical at k=0 extends to a slow rise to a late, narrow, ¼ too high peak and then it overdecays. 1D FJ is
flat (quadratic instead of linear) at small k but then drops too far at intermediate k≈1, where Jones’ phase G J /FJ thus errs high. Whereas
F and G here are exact to 1D at very high A. The shape differences persist strongly down to the smallest A =3 suggesting systematic
Jones error throughout beyond the arguably academic 1D limit and its unitary flutter. The Jones FJ and GJ certainly remain in error to
at least A <20&12 resp. where they are improbably above the 1D values for the same ideal f=1.
The lift slope correction f=.95 reduces F, FJ and GJ equally, but doubly lowers small G more as the square of f/N in eqns (8). Finite
A =30 causes a 1/15 increase in N likewise dropping F, FJ and GJ proportionately (but G here as the square) These differences in drop
lower the phase lag angle of G/F from 1D, to which flutter will prove sensitive.
By A =15 the late GJ peak now matches the 1D peak in height but then still overdeclines. G nearly matches the 1D decline but is ¼
lower at the peak. The G peak lowers and shifts up in k due to the dominant 1/ kA 2 reduction of Im{It} in eqn (11).
 Fig 8 Circulatory Lift wake factor T(k)=F-iG vs k (log scale) for different aspect ratios (& f). (czero.tif)
At A =9 G peaks at about ⅔ 1D with the GJ peak in between with stronger asymptotic decline. F and FJ asymptote equally and
visibly below the 1D limit of ½. Around k≈1 the necking down and unlikely proximity of FJ curves between A=30 and A=9 is due to
the small erroneous positive A-1 term seen in the FJ(∞) limit.
At A =6 the GJ peak is about ⅔ 1D and later whereas G peaks less than ½ 1D. Now F and FJ asymptote to .46. By the Flutterwing
A =4 the GJ peak is now about ½ 1D and later whilst G peaks at about ⅓. Fig 6 implies it now declines at first at k very near zero.
At A =3 t and r rise only slightly whilst pe almost inflects; the GJ peak is now about ⅓ 1D and G peaks at about ¼. -Im{It} has
closed on Im{Is} so G is very small especially at large k. Physically G and phase angle reversal is inevitable as the increase of the Im[It]
trailing downwash as A must overwhelm the decreasing Im[Is] upwash.
A priori sensitivities of the exact flutter solution to such differences in aerodynamics are clearly to the asymptote of F below ½
opening the high k contours, and to the static lift slope decreasing the lfl k=0 flutter boundary slope…

8. Binary Flutter Solution

Contours will still be drawn in constant j,x space where the 2D nexus thus shifts smaller as A is dropped. The low k limit line just
depends upon LF, as ej+(1/ 8LF–e2)x=qy/2L ; y=e+1/ 4LF passing through the nexus N Lj =q2, Lx=q still horizontal at e=0.
Otherwise its slope dx/ dj the inverse of e-N/8feL will be less (negative) by domination of N =1+Iq.
Here the dominance of the internal trailing vortex (over the finite lengths of the shed vortex and apparent mass) all O(A-2) drops the
LF asymptote below ½ with A so with G still going to zero the contours are always hyperbolic at large enough k by (1). For A>2.4,
with G>0, LF passes ½ at some finite k of elliptic contour.
8.1 Flutter for Aerodynamic Balance e=0
Jones 1D high k contours don’t close until about 5.6N, vs exact 4N. due to inaccuracy in his T asymptotes. At moderate k his high G J
/FJ displaces the contours equally from both sides towards the nexus ray to err on the side of stability, as much as the closer spacing of
the contours on the right.
f=.95 shifts our middle 1D contours slightly to the right, but never more than ⅛ the closer contour spacing on the right for less
stability there. Then smaller A = 9 increases the shift biggest on the right middle to at most ¾ a contour spacing Fig 9. The contours
close first at 5.7N at k=6.
At A =6 our contours are shifted up to 1¼ spacing away from the nexus ray for higher critical k. Flutter pitch always leads heave
by <90 [12], so the Theodeorsen lag was further misaligning the lift force with heave velocity for less heave power to balance the pitch
damping loss[12]. So the G/F lag reduction destabilises on the right with its near 90 phase shift [13]. But there is net stabilisation on the
low phase shift left except upper crossovers around x =2. The contours close above 6N at k=6 and larger k elipses are thinner but a bit
longer to close later with some contour crossings.

Fig 9 k neutral stability contours in the inertia/imbalance plane vs inertia at e=0 A =9 (cff6.tif)

By the Flutterwing’s A =4 the shift on the right middle is about 1½ contour more unstable and the contours start closing at 7N for
k=4, as roughly corresponds to LF=½. For A =3 the contours only begin closing at 10N and k=3 and much later for higher k and
ultimately not at all, further decreasing the right hand stability. The right shift peaks about 2 contours more unstable from 1D. Small
looping below the lfl is noticeable around the nexus. In Fig 10, A=2.5 raises the left crossovers again up to k=5 but on the right the shift
vs 1D is more than 2 contours down. See the second order scaling down of the nexus and lfl x below ½. With Jones’ stability bias the
left is evenly stabilised ½ a spacing vs 1D and his right hand stability error halves the net spacing change to one contour more unstable.
With his higher g, the contours first close at an early 5N at k=6 but at greater multiples of N for higher k until ∞. But graphing out at
1/10th scale shows no sign of any of our contours closing. .
The amplified looping seen near the nexus below the nominal lfl stability boundary Lx= ½ is explained by the G≈0 distance above
the lfl line ∝ 2ej+(y-2eq)x-qy/L = ½k2(j-qx)(j-yx)/ F [12] In 1D the right hand contours all crossover at the nexus from outside the
nexus ray j=qx above to inside it below. Then according to their k and so F they reascend and cross their ‘midray’ j= yx at C:
x=2Fy=2Fe+½ with k=0 endpoint E y=n=e+¼, x=2n j=nx. The mode is y=0 feathering in the apparent wind at y behind the pitch
axis. As k→∞ 1D F→½ y↑q explaining the high k contours kissing the lfl line at this N 1D double root.
Here N>1 for LF(0)<1 has made k=0 y=e+lN/ 4f >n to move E closer to the nexus N. Now LF drops through ½ as k rises so then
the y>q contours pass down through the nexus and loop up to the right through P with mode pure pitch aft of the ¾ chord. The k=∞
limiting contour no longer collapses to the nexal ray j=qx above N but hyperbolically left asymptotes to it with right asymptote j=yx
for
y =1/ 4LF∞ with a wedge to the origin in between. Below the lfl NP the neutral k is now the minimum for instability, whereas above
it was the maximum. So the wedge unstable k are too high and the heave at 90 max pitch unfortunately too low for good watermill
power. (But the added mass suffices for a positive spring rate [12].) y/q is highest for e=0 and Jones’ lower LF(∞) for a bigger dip.

Fig 10 Max k neutral stability contours in the inertia/imbalance plane vs inertia at e=0 A =2.5 with unstable wedge (cf7.tif)

The “shortest” distance in (j,x) space to get from the pure virtual mass (hydro)foil to the nexal ray is a m /160 counter 2 chords in
front of the axis. Or for lower k≥31, balance m /6 at the te and at ¼ chord in front of the l.e. for ⅓ more inertia. But the Fig 10 line of
increasing uniform foil skin weight crosses lfl x= ½ at j=.24 < nexal ¼ ; with the solid hydrofoil line more to the left and short of inertia.
Note a net perturbation which instead raises LF(∞) above ½ would put the wedge on the high side of the nexus ray closer to normal
hydrofoils.
Thus the reduction of the LF limit below ½ and of LG with aspect ratio affects both ends of the high k 1D shallow (elliptical)
contours, dipping them even to j=x=0 and then to the right side of the reduced nexus as well as stopping them closing at high j and x.
The moderate contours are further distorted by the reduction in G/F up to 2.5 spacings to the right.

8.2 Flutter with pitch axis leading ¼ chord

In 1D the nexus N=(q2,q) is raised by pitch axis lead e increasing q the axis distance to the ¾ chord point and n=e+¼ its distance to
midchord, so the k=0 endpost E (2n2,2n) is raised e more. Thus the k=0 lfl line in the ⅛ between them is tilted downwards by about -
8e. The k=∞ virtual ¾ pitch line j=qx is rotated by only -.8e, so the contours are rotated away more and more as k decreases (say
proportionately to their angle at the nexus) and so expanded.
The G=0 midray crossing is j=xy & x=2Fy or F=x2/ 2j i.e. y =e+j/ 2x2 =j/x so the G=0 contours cross their midrays j=yx on the
hyperbolic locus (x-½)j=ex2 sagging between N and E At the nexus N the locus’ slope is -e/q2≈-4e or ½ of the lfl slope, so the 1D
G=0 low k≤.45 contours that cross between the locus and the lfl line at the nexus recross the locus at their midrays j=yx from above
and can continue below the lfl line well beyond E to j=0 . [12].
At e= ⅛ Jones’ 1D high k contours still don’t close until about 5.6N. Again at moderate k his contours are displaced equally from
both sides towards the nexus ray to err on the side of stability, but only ⅓ of the closer spacing of the contours on the right, whilst at
k=.4 no change and k=.2, ½ of the spacing more unstable.
Now f=.95 has the big effect of reducing the lfl line slope and proportionately the contours above it until at least k=.6 above
which the slight e=0 change (destabilizing rightwards shift of the contours ) faintly re-emerges.
At A=9 Fig 11 the lfl line is raised much more to old k=.16 but without any net change at k=.5 and again at A =6, where the lfl
line is now ontop of the 1D k=.2. k=.5 remains the same as 1D but the contours close sooner than e=0 at 4.5N at k=6.
At A=4 the contours first close at k=5 at 7N as at e=0. Above the immutable k=.5 the max contour change from 1D is at k=.7 one
spacing more unstable on the right and one less on the left but this seems to be the limit with A. Emerging is the high k contour dip to
the right of the nexus which shifts the closing contour centerline weakly to the right. Whereas the left zone below the lfl is thinning,
because the lfl has risen but the locus is fixed so this lowers the k of contours that can pass between them at the nexus and then cross
the locus from above, to none at A=2.5 when LF(0)=½. But nevertheless the leveling of the lfl line and the nexus mean to the left the
absolute stability boundary is lowered.
Fig 11 Maximum k neutral stability contours in the inertia/imbalance plane vs inertia at e=⅛ A =9 (1D dashed) cff8.tif
Fig 12 Maximum k neutral stability contours in the inertia/imbalance plane vs inertia at e=⅛ A =3 (1D dashed) (cff9a.tif)

At A =3 k=.5 is still unchanged on the right, but more stable on left. Now with the nexus lowered and the lfl flatter, the lfl line
intersects the 1D for k=0 unchanged at j=1.27 q2. The loss of lfl. steady slope is being slowed by the L=1+½A-2 inertia factor. Fig 12
shows the contours only begin closing at 11 N now at k=3 and much later for higher k and ultimately not at all. Notice the intersect of
k=1.4 with k=10.
At A=2.5 the contours close at earlier k=2.2 but much further 18N so the central hole of unconditional instability at any k is being
stretched out towards a hyperbola opening more and more the further out along the nexal ray. Naturally on the right this displaces the
high k contours away slightly countering the upwards push from the lfl line flattening. At large j the previously unchanged k=.4 contour
begins to straighten under this influence and drop to the unstable side of 1D.
For Jones with his higher G and lower F, the first contours close much earlier at k=1.8 at 6.5N. Thus there is less displacement of the
contours away from the nexus ray so the upper right destabilising is reduced. Below k=1.4 the contours match as both methods are
dominated by the lfl shift .
9. Conclusions

The preceeding 1D paper [13] begged the question of how real world finite span would affect its low k l.e. pitch flutter and associated
“pathological” hyperbolic binary flutters at very high j or –x. As anticipated the wake wavelength 2/k exceeding A limits the spike of
Is /k from infinite -iLnk , to finite iLnA as k↓0 in 2D. Also the windwise vorticity that trailed off the tips ¼ of a cycle ago has a
dominant lagged and so damping downwash. So all told l.e. pitch flutter needs unrealistic A >36.
The exact factor for the oscillation of a trailing vortex depends just on S, the frequency reduced by spanwise distance, and vanishes
at infinite S. But the A-2 induction of the internal trailing (bound) vortex remains and dominates the A-2 reduction of the shed induction
(and even of apparent mass) for a net increase, and so a decrease in the high k lift slope F∞ (and even LF∞), agreeing with Jones.
Generally low A has lower F but much lower G and so lower G/F lift lag behind pitch which always itself lags heave velocity in
flutter. The better alignment gives more flutter with more power to the right of the reduced nexus, eg for the Flutterwing.
 Aerodynamic overbalance had tilted the 1D q.s f=1 line down for more instability with more inertia. Both f<1 and finite A
creating Iq reduce the tilt. All told the contour of no change vs 1D is about k=.4 at e=⅛.
The high frequency flutter ellipses from N to 4N are changed to hyperbolae from N to ∞N with right slope lowered by LF∞. Thus
decreasing A opens a flutter wedge below j=qx and the lfl of very high k flutter even at the added imbalance if the small added inertia is
increased ⅓ by real inertia. So the naive assumption that there is no flutter below the lfl line is still not true in 2D, even if the 1D l.e.
flutter transgressions far below the lfl [13] at large j and -x are gone. Jones’ TJ deviations in academic 1D persist at finite A and
underpredict flutter except for a wider more unstable wedge. But this second order low A wedge mode of very high k virtual pitch
behind the ¾ point is unlikely as a flutter watermill, despite its feasible j and x.
But pitch and roll are also coupled by gravity static imbalance which is dominant over the above dynamic imbalance at low stream
speeds. A very long pitch counterbalance arm of 3.5c above water provided sufficient coupling and pitch inertia for a c=15cm foil to
flutter in water[10]. At full scale[2] enough coupling might come from a 1c arm that could stay clear of the water until max roll.
Perhaps, the low phase shift and amplitude ratio linear instability at low inertia to the left of the e=0 nexus would distort non-linearly to
higher amplitude ratio and power. Otherwise above water a pitch gear up to a flywheel [12] could increase the pitch inertia above N’s
¼ lmc2. A simple first test would be below a low steel bridge over a small slow river. Instead of the winch, an immersed rhombus could
invert the increase in distance with swing into non-linear compression on its diagonal for pumping.
In all, this paper has given the state of the art of power flutter, from the proven Flo’Pump to the Flutterwell pump which needs testing
on real wells to final proof of the impracticality of pure pitch flutter and finite aspect ratio extension of the search for hydrofoil variants.

References
1. Duncan, J.W. The fundamentals of flutter (1948) , Ae.Res.Co. R&M No. 2417, November, 1-36
2. Farthing, SP. Binary Flutter as an Oscillating Windmill :Scaling and Linear Analysis Wind Engineering 2013 37 (5) : 483-99
3. Farthing, SP. FWP Leading edge smoke url:youtube.com/watch?v=16kB6p-kcC0 7sec uploaded 1993
4. Farthing, SP. Visualisation and analysis of Torque Reacting Flow through Starting FanMill Wind Engineering 2010Vol 35 (5).: 625-
634
5. Dixon, JC. Load Matching Effects on Wind Energy Convertor Performance p 418-421 in Future Energy Concepts , IEE Conference
Publication London 30Jan-Feb 1 1979
6. Farthing, SP. The Flutterwing Pumps: Design, NonLinearities, & Measurements Wind Engineering 2014 38 (2): 217-31
7. Farthing, SP. Wing’d Pumps : CC scripted video url:youtube.com/watch?v=6zIj7LCtX0U 10min uploaded 2012
8. Farthing, SP. Flutterwell Governor url: https://youtu.be/Swzd7t26YjA 27sec uploaded 2023
9. Farthing, SP. Foot Well Pump url: http://econologica.org/yardpump.html uploaded 2005
10. Farthing, SP. Binary Flutter in Water url: youtube.com/watch?v=NDm78DOcEOM 12sec uploaded 2009
11 Ashley, HA. & Dugundji, J. Aeroelastic Stability of Lift Surfaces in High-Density Fluids J. Ship Research 1959 3(1) : 10-28
12. Farthing, SP. Binary Flutter Solution for Fluid Power Journal of Aerospace Engineering,  (ASCE) 2018 31(3): 1-11
13. Farthing, SP. Exact Pitch and Heave Flutter for Complex Theodorsen function Aero. Journal 2019 123(1265), p.1053-1074
14. Jones, RT. The Unsteady Lift of a Finite Wing 1939 NACA-TN-682 url: https://ntrs.nasa.gov/search.jsp?R=19930081468
15. Boutet, J & Dimitriadis, G. Unsteady Lifting Line Theory Using the Wagner Function for the Aerodynamic and Aeroelastic
Modeling of 3D Wings Aerospace, url: https://orbi.uliege.be/bitstream/2268/227423/3/51%20aerospace-05-00092.pdf uploaded 2018
16. Andrianne, T. Theodorsen Aerodynamics url:http://www.ltas-aea.ulg.ac.be/cms/uploads/AERO0032-1_2015-16_Lecture3.pdf
uploaded 2016
17. Schlichting, Hermann. & Truckenbrodt, Erich.  (1979) Fig 3-43  Aerodynamics of the airplane.  New York :  McGraw-Hill
18. Wolfram Research, Inc., Mathematica, Version 7.0, Champaign, IL (2008).
19. Abramowitz, M.; Stegun, I. A., eds. (1972), Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables,
New York: Dover Press, ISBN 978-0-486-61272-0 p 378 (130)and p 375 (127)
20. Jumper, EJ &. Hugo, RJ. Finite wings undergoing rapid unsteady motions -Theory 2012 Journal of Aircraft 31, (3) p495-502
21. Hoerner, SF & Borst H. 1975. Fluid-dynamic lift: practical information on aerodynamic and hydrodynamic lift Hoerner Pubs.
Bricktown N.J.

Symbol Table
subscripts ± evaluated at y/s= t±r
a = 2f 1D lift slope, static a0
b semichord c chord of foil, c0 at root of ref elliptic planform
ec trail of quarter chord behind pitch axis in chord
f lift slope correction factor for Re
g=G/k negative pitch damping rate of G
h heave of pitch axis
h3/4 heave of the ¾ chord point
i square root of -1
j pitch inertia/mc2, j pitch inertia/mc2
k chordal reduced frequency c/V
l≈1-½A-2 apparent mass reduction factor
L≈1+½A-2 effective increase factor on J,x,T
m 1D added mass of foil/unit length
n= e+¼ chords from pitch axis to midchord.
o=qc distance from pitch axis to ¾ chord q=e+½
p= a/ 4A parameter in It
r fraction of semispan where trailing induction is average
s wing semi-span
t center of trailing vorticity as fraction of semispan
t time
x pitch imbalance/ mc ; also distance of vortex element from trailing edge line
w distance from the ¾ chord line
y =e+¼/LF chordal distance parameter
A or AR Aspect ratio 2s/c
B(y) steady lift spanwise function
n steady lift Fourier coefficient of sin n
C 1D Correction factor for Is
D induction by wake vortex strand of unit strength ,
I(nternal) E(xternal)

E1(u) =Exponential Integral function E1(u) = ∫1dt e /t

∞ -ut

I back-induced factor of nominal upwash (on average)

It from trailing vorticity , evaluated at ¾ chord, steady Iq
Is from shed vorticity, evauated at midchord
In Back induction of nth Fourier spanwise mode of steady lifting line

M(S) A -2 unsteadiness factor in induction of external trailing vortex

N =1+ Iq is the steady lift divisor due to A N 2.13fA-1 +1.11f2A-2
N Nexus point (j,x)=(q,q2)
R(S) primary unsteadiness factor in induction of external trailing vortex
S= y /V reduced frequency based on spanwise separation y,
T =F-iG=F-ikg complex Theodorsen T ≈f/(1+It+C(v)Is) FJ(ones). =Tl
U upwash ¾ evaluated at ¾ chord point or e at pitch axis.
V flowspeed
Fluid density
 pitch angle
apparent ¾ chord angle of attack
net ¾ chord angle of attack
Vorticity distance x behind airfoil section
Discriminant of the neutral stability quadratic eqn in j,x =4F2k4
circular frequency in radians of phase/unit time