3.2 Review of Established Solution Techniques
3.2.3 Approximate LG Solution for a Nonuniform Cochlea
dissipative systems [112]. Our approach will be (1) to nd an expression for the energy ow as a function of amplitudeA, (2) to nd an expression for the group velocity as a function of k, and (3) to combine these expressions to determineAas a function ofk. We shall develop the condition for the undamped case, which is quite simple; the validity of the result in the damped case has been shown by de Boer and Viergever [22]. An alternate approach, which appeals to Hamilton's principle and Lagrangian mechanics, was used by Steele and Taber [112], based on the general treatment by Whitham [126]. We now proceed with the simpler rst-principles energy approach.
The energy of the wave consists of three components: the kinetic energy of the membrane massKm, the potential energy of the membrane stinessVm, and the kinetic energy of the uidKf. The time-averaged kinetic energy of the membrane mass per unit area is
Km = 12
Z 2
0 1
2Mv2yd= 14M!2A2: (3:21) The time-averaged potential energy of the membrane stiness per unit area is
Vm= 12
Z 2
0 1
2S2d = 14SA2: (3:22)
To computeKf, we need the velocity potential in terms of the amplitude factor A. Com- bining Equations 3.7 and 3.19 yields
= i!A
ksinh(kh) cosh(ky)expi(!t kx): The time-averaged kinetic energy of the uid per unit area is
Kf = Z0h 1 2
Z 2
0 1
2(2)(v2x+v2y)ddy (3.23)
= !2A2
2ktanh(kh) (3.24)
= 12!2A2Q(k); (3.25)
where the 2 term accounts for the uid in both chambers, and the important function
Q(k) is dened as
Q(k) = 1 ktanh(kh): The energy balance then assumes the form
Vm =Kf +Km;
which is identical to the dispersion relation when Equations 3.21, 3.22, and 3.25 are substi- tuted [22]. Finally, the total energy density E is given by
E = Vm+Kf +Km (3.26)
= 2Vm (3.27)
= 12SA2: (3.28)
We now have an expression for the energy density E as a function of membrane dis- placement amplitude A. The second step in the derivation is to nd an expression for the group velocityU as a function ofk. The lossless dispersion relation can be written in terms of the function Q(k):
Q(k) = 1
ktanh(kh) = S M!2 2!2 : Dierentiating with respect tok yields
@Q@k = @Q
@!@!
@k (3.29)
= U @Q@! ; (3.30)
which leads to
U = @Q
@k !3 S
!
: (3:31)
Energy ows at the group velocity [126]. For a constant rate of energy ow, we must have
EU = const: (3:32)
A(k) = C @Q
@k
= iCktanhkh
ptanhkh+khsech2kh;
where C is a constant of dimension (length)2. This important equation is called the transport equation, since it relates to the transport of energy.
Substituting this result into Equation 3.20 yields the full equation for the membrane displacement:
(x;t) = iCktanhkh
ptanhkh+khsech2khexpi(!t Z0xk(u)du); (3:33) where k is the local root of the dispersion relation. Combining this result with Equation 3.7 yields the expression for the velocity potential:
(x;y;t) = C!cosh(ky)
cosh(kh)ptanhkh+khsech2khexpi(!t Z0xk(u)du); (3:34) We can derive similar expressions for the membrane velocity vy(x;t), and uid pressure p(x;y;t), using the dening relations 3.1 and 3.6.
Rhode's data are expressed in the form of a ratio of basilar-membrane displacement to malleus displacement, which we assume is proportional to stapes displacement [91]. We must now compute the stapes displacement.
Recall that the horizontal uid velocity
v
x(x;y) at any point (x;y) in the uid isv
x(x;y) = @@x:
The horizontal uid displacement
d
x(x;y) is the time integral of the horizontal uid velocity.For a sinusoidal disturbance with angular frequency!, we have
d
x(x;y) = i!@
@x: (3:35)
Dierentiating Equation 3.34 with respect to x yields
@@x = ( ik+O[k0(x)])(x;y): (3:36) As a rst approximation, we shall ignore the terms involvingk0(x). Combining Equations 3.36 and 3.35 yields
d
x(x;y) = k!(x;y): (3:37)
Following Steele and Taber [112], the stapes displacement dst is the value of the hori- zontal uid displacement at x= 0, averaged over the height of the ducth:
dst = 1h
Z h
0
d
x(0;y)dy (3.38)= k0
!h
Z h
0 (0;y)dy;
wherek0 is the value of the wavenumber kevaluated atx= 0. Substituting Equation 3.34 and performing the integration yields
dst= Ctanh(k0h)exp(i!t)
hqtanh(k0h) +k0hsech2(k0h): (3:39) Combining Equations 3.33 and 3.39 yields the ratioDof membrane to stapes displace- ment:
D(x;!) =
dst =ikhtanh(kh) tanh(k0h)
stanh(k0h) +k0hsech2(k0h)
tanh(kh) +khsech2(kh) exp iZ0xk(u)du: (3:40) Note that the above expression is only a rst approximation, since the terms involvingk0(x) indst have been neglected.
The general LG solution for the velocity potential given in Equation 3.34 degenerates to the following simple form at y=h under the long-wave approximation:
(x;t) = constk 1=2 expi(!t Z0xk(u)du); (3:41) where kis given by the simple long-wave expression of Equation 3.17. This form was rst applied to cochlear mechanics problems by Zweig, Lipes, and Pierce [134]. The validity
k12dk dx1:
Zweig and colleages also showed that it is possible to evaluate the integral in closed form, under the scaling assumption:
S(x) = S0exp( 2x=d); (x) = 0exp( x=d); M(x) = const:
Alternate derivations of the long-wave LG result are given by Lighthill [60, p. 189{90] and by Viergever [121, p. 103{106].
In the short-wave region, the general LG solution for the velocity potential given in Equation 3.34 degenerates to the following simple form aty =h:
(x;t) = constexpi(!t Z0xk(u)du);
where k is given by the simple short-wave expression of Equation 3.18. This form was rst applied to cochlear mechanics problems by Siebert [103]. Under the same scaling assumptions used by Zweig, Lipes and Pierce, for the long-wave case, we can evaluate the short-wave integral in closed form.
A Mathematica implementation of the two-dimensional LG algorithm is given in Ap- pendix A.
Comparison of LG and Finite-Dierence Results
Steele and Taber compared their LG results to the nite-dierence results of Neely for a number of dierent frequencies, using identical parameters for the two models [112].
Their results have been recomputed, and are presented here for comparison with the mode- coupling LG solution presented in section 3.2. The physical parameters used for their comparison were
S(x) = 1:0107 e x=d g s 2 mm 2; (x) = 2 g s 1 mm 2;
M = 1:510 3 g mm 2; d = 5 mm;
h = 1 mm; L = 35 mm;
= 1:010 3 g mm 3:
Note that neither Neely nor Steele and Taber regarded these parameters as physically realistic. In particular, the membrane mass M is unrealistically large.
The amplitude and phase of the basilar-membrane displacement ratios are shown in Figure 3.5 for the two methods. Clearly, the LG solution captures the general behavior of a gentle increase in amplitude toward a peak, followed by a sharp cut-o. However, the quantitative agreement is poor for the lowest frequencies, and in the cut-o region, the amplitude of the LG solution decreases much too fast, and the phase behavior is incorrect.
Both of these problems are addressed in Section 3.3.