1. Hyperbolic Equations ^{[}^{1}^{]}
A partial diﬀerential equation of second order, i.e., one of the form
Au _{x}_{x} + 2Bu _{x}_{y} is called hyperbolic if
+ Cu _{y}_{y} + Du _{x} + Eu _{y} +F
= 0
B ^{2}  4AC < 0 Example: The one dimensional wave equation:
u _{t}_{t}  c ^{2} u _{x}_{x} = 0 is an example of a hyperbolic equation.
2. Riemann Function ^{[}^{2}^{]}
The Riemann function R(x, y; ζ, η) is deﬁned as the solution of the equation R _{x}_{y}  (aR) _{x}  (bR) _{y} + cR = 0 which satisﬁes the conditions
R(ζ, y; ζ, η) = exp[
η
y
a(ζ, t) dt]
and
R(x, η; ζ, η) = exp[ x b(t, η) dt]
ζ
on the characteristics x = ζ
and y = η, where (ζ, η) is a point on the domain Ω
on which the conditions state above are deﬁned. The solution is then given by the
Riemann formula
u(x, y) = x
0
dζ y R(ζ, η; x, y)f(ζ, η) dη
1
This method of solution is called the Riemann method.
3. Green’s theorem ^{[}^{3}^{]}
Let C be a positively oriented, piecewise smooth, simple closed curve in a plane,
and let D be the region bounded by C.
If L and M are functions of (x, y) deﬁned
on an open region containing D and have continuous partial derivatives there, then
C (L dx + M dy) = D
−
∂L
∂y
) dx dy
where the path of integration along C is counterclockwise.
4. Adjoint Equation ^{[}^{4}^{]}
The adjoint equation is applicable to linear diﬀerential equations which yield a linear diﬀerential equation of a lower order. For every solution of the adjoint equation we can ﬁnd, we can reduce the order of the original equation by one.
1
2
5. Adjoint Linear Operators ^{[}^{5}^{]}
To a linear operator L, deﬁned by
L(u) = au _{x}_{x} + 2hu _{x}_{y} + bu _{y}_{y} + 2gu _{x} + 2fu _{y} + cu
where the coeﬃcients are continuously diﬀerentiable functions of x and y alone, there corresponds a unique linear operator L ^{∗} , called the adjoint of L, such that vL(u)  uL ^{∗} (v) is a divergence. If D is a domain whose boundary is a regular closed curve C, it follows by Green’s theorem that
D (vL(u) − uL ^{∗} (v)) dx dy = C (lH + mK) ds
where (l, m) are direction cosines of the outward normal to C. After doing some calculations and subsequent simpliﬁcations, we get,
where
L ^{∗} (v) = (av) _{x}_{x} + 2(hv) _{x}_{y} + (bv) _{y}_{y} − 2(gv) _{x}
− 2(fv) _{y} + cv
H = avu _{x} − u(av) _{x} + hvu _{y} − u(hv) _{y}
+ 2guv
K = hvu _{x} − (hv) _{x} + bvu _{y} − u(bv) _{y} + 2fuv
H and K are not unique; we can add to them ∂θ/∂y and ∂θ/∂x respectively; but L ^{∗} is unique and the adjoint of L ^{∗} is L. L is said to be selfadjoint if L ^{∗} = L. The condition for this is a _{x} + h _{y} = 2g, h _{x} + b _{y} = 2f so that a selfadjoint operator can be written in the form
∂
∂
∂
∂
_{∂}_{x} (au _{x} ) + _{∂}_{x} (hu _{y} ) + _{∂}_{y} (hu _{x} ) + _{∂}_{y} (bu _{y} ) + cu
3
6.
Use ^{[}^{4}^{]}
The Riemann’s method is applicable to linear hyperbolic equations of the second order in two independent variables which yields an exact solution in terms of the solution to the adjoint equation.
7.
Idea ^{[}^{4}^{]}
The solution of a noncharacteristic initial value problem in two dimensions can be found if the adjoint equation with speciﬁc boundary conditions can be solved.
8. Essentials of Riemann Method ^{[}^{6}^{]}
The solution of the problem of Cauchy given by Riemann for the partial diﬀer ential equation
(1)
u _{x}_{y}  a(u _{x} − u _{y} ) = 0, a = a(x, y)
has been extended to the general linear partial diﬀerential equation of second order in two independent variables x, y,
(2)
u _{x}_{y} + au _{x} + bu _{y} + cu = f
where a, b, c, f are given functions of x, y. For the sake of simplicity we consider the equations of the type
(3)
L(u) = u _{x}_{y}  au _{x}  bu _{y} = 0, a = a(x, y), b = b(x, y)
Actually, if a solution u _{1} of (2) and a solution u _{0} of the homogeneous equation (f = 0) are known, equation (2) can always be reduced to one of type (3) for a new unknown function u deﬁned by u = u _{1} + u _{0} u. The classical procedure associates with (3) its adjoint equation
(4)
M(v) = v _{x}_{y} + (av) _{x} + (bv) _{y} = 0
and is based on the diﬀerential identity
(5)
A _{x} + B _{y} = vL(u)  uM(v), where A = ^{(}^{v}^{u} ^{y} ^{−}^{u}^{v} ^{y} ^{)}
_{2}
 auv, and B =
^{(}^{v}^{u} ^{x} ^{−}^{u}^{v} ^{x} ^{)}
2
 buv
The identity implies that the line integral
I =
(B dx − A dy)
vanishes around closed paths lying in the interior of a domain D within which the function u, v are regular solutions of L(u) = 0 and M(v) = 0, respectively. The line integral I vanishes around closed paths leads immediately to the solution of Cauchy’s problem for (3). Let C be an arc meeting any horizontal or vertical line in at most one point. Along C the values of u _{x} , u _{y} (Cauchy’s data) for the
4
solution of (3) are given.
In order to calculate the value u(P) of the solution at the point P in the ﬁgure, the
line integral I is evaluated around the closed path P XY P formed by the horizontal
line segment P X, the arc XY of C, and the vertical line segment Y P. One ﬁnds
(6)
X
P
Y
B(x, y) dx + ^{}
X
P
(B dx − A dy)  ^{}
Y
A(x, y) dy = 0
If a solution v = v(x, y; x, y) of M(v) = 0 can be found for which
(7)
v _{x} = bv on y = y, v _{y} = av on x = x and v(x, y; x, y) = 1
hold on the characteristics x = x and y = y through P, we shall have
(8)
A(x, y) = (uv) _{y} /2 and B(x, y) = (uv) _{x} /2
and consequently (6) will yield
(9)
u(P) =
_{2} [u(X)v(X) + u(Y )v(Y )] + ^{}
X
1
Y
(B dx − A dy)
The function v is known as Riemann’s function, and, once it has been deter
mined, u(P) may be calculated from (9) from the knowledge of Cauchy data along
C.
The ﬁrst step in Riemann’s method is the formulation of the line integral I so that it vanishes around closed paths. We note that the coeﬃcients A, B of the diﬀerentials are bilinear forms in two set of variables u, u _{x} , u _{y} and v, v _{x} , v _{y} . Furthermore, the real reason for the introduction of adjoint equation (4) and its solution v is to make available the identity (5), which guarantees that I vanishes around closed paths.
9. A more general case ^{[}^{5}^{]}
Consider the following equation
L(u) = 2u _{x}_{y} + 2gu _{x} + 2fu _{y} + cu = F(x, y)
given u, u _{x} , u _{y} on some duly rectangular arc C. The data satisfy the strip condition
du = u _{x} dx + u _{y} dy on C. The functions g, f, c, F
are assumed to be continuously
diﬀerentiable functions of x and y alone. Let the characteristics y = y _{0} and x = x _{0} through P(x _{0} , y _{0} ) cut C in Q and R respectively. Let D be the domain bounded by PQ, PR and C; let Γ be its boundary. Then, if L(u) = F, L ^{∗} (v) = 0, we have
5
D vF dx dy = D (vL(u) − uL ^{∗} (v)) dx dy = D
_{[} ∂H
∂K
∂X ^{+} ∂y
] dx dy = Γ (−K dx+H dy)
where H = avu _{x}  u(av) _{x} + hvu _{y}  u(hv) _{y} + 2guv and K = hvu _{x}  (hv) _{x} + bvu _{y}  u(bv) _{y} + 2fuv by Green’s theorem, assuming that u and v have continuous second derivatives. Using the aforementioned values of H and K, the integral round Γ is equal to
− P Q [(vu) _{x} − 2u(v _{x} − fv)] dx + QR (−K du + H dy) + RP [(vu) _{y} − 2u(v _{y} − gv)] dy
= 2(uv) _{P} − (uv) _{Q} − (uv) _{R} + QR (−K dx + H dy)
provided that
that is, provided that
v _{x} − fv = 0 when y = y _{0} v _{y} − gv = 0 when x = x _{0}
and
y _{0} ) = exp x 
x 

v(x, 

0 

y 

v(x _{0} , y) = exp y 0 
f(ξ, y _{0} ) dξ
g(x _{0} , η) dη
where we have taken v(x _{0} , y _{0} ) to be unity, as we may without the loss of generality. Hence v is given by a characteristic boundary value problem for L ^{∗} (v) = 0. This problem has a solution, the RiemannGreen function, which we denote by
v(x, y; x _{0} , y _{0} ).
If Q is (x _{1} , y _{0} ), and R is (x _{0} , y _{1} ), it follows that
u(x _{0} , y _{0} ) =
1
_{2} 1 u(x _{1} , y _{0} )v(x _{1} , y _{0} ; x _{0} , y _{0} ) + _{2} u(x _{0} , y _{1} )v(x _{0} , y _{1} ; x _{0} , y _{0} )
1
2
QR
(K dx − H dy) +
1
2
D
v(x, y; x _{0} , y _{0} )F(x, y) dx dy
+
This is the required solution, since H = vu _{y}  uv _{y} + 2guv and K = uv _{x}  vu _{x} + 2fuv are given on QR by the Cauchy data.
10. Determination of the RiemannGreen function ^{[}^{5}^{]}
The diﬃculty in Riemann’s solution is the determination of the RiemannGreen function. The method replaces a Cauchy problem by a characteristic boundary value problem. It suﬃces to consider the case when independent variables are characteristic variables, so that the linear operator is
L(u) = 2u _{x}_{y} + 2gu _{x} + 2fu _{y} + c
6
If we make the change of dependent variable u = φU, and divide through by φ, we get a linear operator
M(U) = 2U _{x}_{y} + 2GU _{x} + 2FU _{y} + CU
where 

G = g + ^{φ} φ ^{y} 

and 

F = f + ^{φ} φ ^{x} 

and 

C = c + 2g ^{φ} φ ^{x} 
+ 2f ^{φ} φ ^{y} 
_{+} φ xy φ 
The RiemannGreen function for L(u), v(x, y; x _{0} , y _{0} ), satisﬁes the adjoint equation
L ^{∗} (u) = 0 and the conditions v _{x} = fv on y = y _{0} , v _{x} = gv on x = x _{0}
and
v(x _{0} , y _{0} ; x _{0} , y _{0} ) = 1
It follows that the RiemannGreen function for M(U) = 0 is
V (x, y; x _{0} , y _{0} ) =
φ(x, y)
_{φ}_{(}_{x} _{0} _{,} _{y} 0 _{)} v(x, y; x _{0} , y _{0} )
For example, if we put u = (x + y)U in u _{x}_{y} = 0, for which the RiemannGreen function is constant, we get
_{U} xy _{+} U _{x} + U _{y}
x + y
= 0
for which the RiemannGreen function is therefore
V (x, y; x _{0} , y _{0} ) =
x + y
x _{0} + y _{0}
The change of dependent variable is useful if we can choose φ so that M(U) = 0 is selfadjoint. This occurs if
f =  _{∂}_{x} logφ, g =  _{∂}_{y} logφ
∂
∂
such a transformation is thus possible when g _{x} = f _{y} . The RiemannGreen function is the solution of a characteristic boundary value problem, and does not depend in any way on the arc carrying the Cauchy data. If it is possible to solve by some other method the problem of Cauchy with a simple curve carrying the data, a comparison of this solution and the Riemann solution should give the RiemannGreen function. In the case of the two equations discussed by Riemann, it was possible to do this; he solved the problem of Cauchy with data on a straight line by using Fourier cosine transforms.
11. An Example ^{[}^{4}^{]}
7
Consider the following partial diﬀerential equation (E1)
α ^{2} w _{β}_{β} − β ^{2} w _{α}_{α} = 0
w(α, 1) = f(α)
w _{β} (α, 1) = g(α)
where ∞ < α < ∞, 1 < β < ∞, and f(α) and g(α) are given functions.
If we
change variables in equation (E1) from {w, α, β} to {u, x, y} by
u(x, y) = w(α, β)
x = αβ and y =
^{β}
α
Now, equation (E1) becomes upon doing the transformations stated above
(10)
u xy
 _{2}_{x} 1 u _{y} = 0
The boundary conditions in equation (E1) transform to
(11)
u(s, ^{1} _{s} ) = f(s) and su _{x} (s, ^{1} _{s} ) + ^{1} _{s} u _{y} (s, ^{1} _{s} ) = g(s)
where −∞ < s < ∞. By manipulations of equation (11), we can derive the following
set of equations (E2),
u(s,
^{1}
_{s} ) = f(s)
u _{x} (s, ^{1} _{s} ) = ^{1}
_{2}
[f ^{} (s)
+
^{1}
_{s} g(s)]
u _{y} (s, ^{1} _{s} ) = ^{1}
_{2} [sg(s) − s ^{2} f(s)]
The domain in which equations (10) and (E2) are to be solved is shown in this ﬁgure.
To solve equations (10) and (E2), we use Riemann’s method.
Comparing equa
tion (10) to L[u] = u _{x}_{y}
+ a(x, y)u _{x} + b(x, y)u _{y} + c(x, y)u = f(x, y), we determine
a = 0, b = −1/2x, c = 0, f = 0
The Riemann’s method gives solution of the form
u(ζ, η) =
Q
_{2} 1 R(P; ζ, η)u(P) + _{2} R(Q; ζ, η)u(Q) −
1
P
B[u(x, y), R(x, y; ζ, η)] + D f(x, y)R(x, y; ζ, η) dx dy
This formula is valid when we wish to ﬁnd u(S) = u(ζ, η), where S represents an
arbitrary point as shown in the ﬁgure below. If we assume that the initial curve
Γ is monotonically decreasing then we can write the solution in the above stated
8
form. Here, P and Q are as shown in the ﬁgure below.
Coming back to our problem at hand (10), we can write the solution (E3) as follows
u(ζ, η) =
Q
_{2} R(P; ζ, η)u(P) + _{2} R(Q; ζ, η)u(Q) −
1
1
P
[( _{2} Ru _{y} − ^{1}
1
1
_{2} R _{y} u) dy − (− _{2}_{x} Ru +
1
_{2}
Ru _{x} − ^{1}
_{2} R _{x} u) dx]
All that remains to ﬁnd the Riemann’s function. Using equations from section 2,
we say that R(x, y; ζ, η) satisﬁes (E4)
R xy +
1
_{2}_{x} R _{y} = 0
R(ζ, y; ζ, η) = 1
R(x, y; ζ, η) =
ζ
x
R(ζ, η; ζ, η) = 1
Because the ﬁrst equation of (E4) can be integrated directly with respect to x and
then with respect to y the general solution to equation (E4) is easily seen to be of
the form
(12)
R(x, y; ζ, η) = M(x; ζ, η) + ^{K}^{(}^{y}^{;}^{ζ}^{,}^{η}^{)}
^{√} x
for some M(x; ζ, η) and some K(y; ζ, η). Using equation (12) in the boundary
conditions in the set of equations (E4), the solution is found to be
(13)
R(x, y; ζ, η) =
ζ
x
Using equation (13) in equation (E3), we can ﬁnd u(ζ, η) and hence, w(α, β) for
any values of α and β.
12. Riemann Zeta Function ^{[}^{7}^{]}
The Riemann zeta function, ζ(s), is a function of a complex variable s that
analytically continues the sum of inﬁnite series
∞
n=1
1
n ^{s}
which converges when the real part of s is greater that 1. The Riemann zeta func
tion plays a pivotal role in analytic number theory and has applications in physics,
probability theory, and applied statistics.
Deﬁnition: The Riemann zeta function ζ(s) is a function of a complex variable s
= σ + ιt.
9
The following inﬁnite series converges for all complex numbers s with real part
greater than 1, and deﬁnes ζ(s) in this case:
where
σ = R(s) > 1
ζ(s) =
∞
n=1
1
n ^{s}
It can also be deﬁned by the integral
ζ(s) =
s−1
Γ(s) ∞ e ^{x} − 1
1
0
_{x}
dx
The Riemann zeta function is deﬁned as the analytic continuation of the function
deﬁned for σ > 1 by the sum of preceding series.
10
References
[1] Wolfram Mathworld: Hyperbolic Partial Diﬀerential Equations [2] Wolfram Mathworld: Riemann Function [3] Wikipedia: Green’s Theorem [4] Zwillinger D. (1997), Handbook of Diﬀerential Equations 3rd edition, Academic Press [5] Copson E.T., Partial Diﬀerential Equations, Cambridge University Press 1975 [6] Martin M.H., Riemann’s Method and the Problem of Cauchy, University of Maryland [7] Wikipedia: Riemann Zeta Function