Partial Differential Equation Part C Upto 21oct
Partial Differential Equation Part C Upto 21oct
Teacher: A Ganguly
In this last Part of course, we will study three celebrated 2nd order linear PDEs
subject to different types of Boundary Conditions (BC) and Initial Conditions (IC):
డమ ట ଵ డమ ట
1. Wave Equation: (1D) = , [ ݔis position and ݐis time]
డ௫ మ మ డ௧ మ
డట డమ ట
2. Heat Equation (or Diffusion Equation): (1D) =ߙ
డ௧ డ௫ మ
ଶ ଶ
3. Laplace Equation: = ߰ 0, [ is called Laplacian Operator]
డమ డమ డమ
where in 3D Cartesian coordinates, ଶ ؠ మ
+ మ
+ .
డ௫ డ௬ డ௭ మ
We already know above three PDEs are respectively Hyperbolic, Parabolic and
Elliptic [See classification scheme discussed in Part B]. Before going into detailed
study of these PDEs, we need some conception about Fourier series.
1
We will consider only real-valued function ݂( )ݔof single real variable ݔ.
Suppose ݂( )ݔis a periodic function of period 2ߨ defined on Թ
బ
݂(= )ݔ + σஶ
ୀଵ[ܽ cos(݊ )ݔ+ ܾ sin(݊])ݔ. (1.1)
ଶ
0 ,݉ ് ݊ 0 ,݉ ് ݊ (1.2a)
గ గ
ିగ sin(݉ )ݔsin(݊ )ݔdݔ = ,ିగ cos(݉ )ݔcos(݊ )ݔdݔ =
ߨ ,݉ = ݊ ߨ ,݉ = ݊
గ
ିగ sin(݉ )ݔcos(݊ )ݔd = ݔ0, ݉ , ݊ . [݉, ݊ are non-negative integers ] (1.2b)
Assuming that the infinite series is term by term integrable, we see that if we
multiply (1.1) by sin(݉ )ݔ, cos(݉ )ݔand integrate over the interval [െߨ, ߨ], then
due to the orthogonal relations (1.2), all but one term in RHS will disappear,
thereby getting Fourier coefficients as follows:
ଵ గ ଵ గ
ܽ = ିగ ݂( )ݔcos(݊ )ݔd ݔ, ܾ = ିగ ݂( )ݔsin(݊ )ݔd ݔ, ݊ = 0,1,2, … , (1.3)
గ గ
where for the case of finding ܽ , we just integrate (1.1), and use the relation
గ గ
ିగ sin(݉ )ݔdି = ݔగ cos(݉)ݔd = ݔ0, ݉, ݊ ് 0. (1.4)
Note that in formula (1.3), I do not separately put ܽ , and do not exclude ݊ = 0, as
in most literature, because for ݊ = 0, formula gives correct expression for ܽ , and
the extra term ܾ is, in fact, identically zero by the formula. Clearly, formula (1.3)
yield coefficients if the function ݂( )ݔis integrable in the interval. Some points
worth mentioning:
¾ Instead of the interval [െߨ, ߨ], we may take any interval of length 2ߨ on Թ,
in formula (1.3), for instance, we may choose [0,2ߨ], if periodicity of ݂()ݔ
is exhibited in [0,2ߨ]. It is easy to verify that orthogonal relations hold true
for any interval of length 2ߨ.
¾ If instead of Թ, ݂( )ݔis defined only on [െߨ, ߨ], that means given ݂( )ݔis
no longer periodic. In this circumstance, we make a periodic extension of
݂( )ݔon Թ by )ݔ(ܨ, where )ݔ(݂ = )ݔ(ܨ, [ א ݔെߨ, ߨ] and = )ݔ(ܨ
ݔ(ܨ+ 2݇ߨ) outside for non-zero integer ݇, i.e we make a copy of shape of
2
݂( )ݔin all other interval of length 2ߨ. Note that )ݔ(ܨ, thus constructed, is
periodic so that above Fourier series representation is possible for it, but our
interest lies only in [െߨ, ߨ].
¾ If given ݂( )ݔis even or odd function, then Fourier series (1.1) reduces to
Fourier-cosine or Fourier-sine series as follows:
݂(െ)ݔ(݂ = )ݔ: ܾ = 0 ฺ ݂( = )ݔబ + σஶ ୀଵ ܽ cos(݊)ݔ, (1.5a)
ଶ
݂(െ = )ݔെ݂()ݔ: ܽ = 0 ฺ ݂( = )ݔσஶ ୀଵ ܾ sin(݊)ݔ. (1.5b)
¾ If ݂( )ݔis defined on [0, ߨ], then there are two ways of periodic extension:
Even extension: Define ݂(െ)ݔ(݂ = )ݔ, [ א ݔെߨ, 0], and then make
identical copies in all intervals of length 2ߨ, by constructing )ݔ(ܨ,
discussed above for second instance. In this case we have half-range
Fourier-cosine series (1.5a), wherein ܽ s are given by
ଶ గ
ܽ = ݂( )ݔcos(݊ )ݔdݔ. (1.6a)
గ
Odd Extension: Define ݂(െ = )ݔെ݂()ݔ, [ א ݔെߨ, 0], and then make
identical copies in all intervals of length 2ߨ, by constructing )ݔ(ܨ,
discussed above for second instance. In this case we have half-range
Fourier-sine series (1.5b), wherein ܾ s are given by
ଶ గ
ܾ = ݂( )ݔsin(݊ )ݔd ݔ. (1.6b)
గ
Note that all of the arguments for above instances apply for period 2ܮ, so I’m not
repeating those.
Fourier series converges to ݂( )ݔat all points in the interval except for
discontinuities. At point of discontinuity and at end-points it converges to average
of right-hand and left-hand limits of ݂()ݔ, e.g. At the point = ݔെܮ, ܮ, value of
ଵ
Fourier series converges to [ = ܣlim௫՜ିା ݂( )ݔ+ lim௫՜ି ݂(])ݔ.
ଶ
3
Generalized Fourier Series
We have seen that Fourier Series is an expansion of ݂( )ݔin terms of elements
of set S of periodic functions {1, sin(݊ )ݔ, cos(݊})ݔஶ
ୀଵ . Now property of the set
S is that it is a complete* set of orthogonal functions. There exists many other
such set, e.g. {ܲ (})ݔஶ
ୀ is also a complete set of orthogonal functions, where
ܲ ( )ݔis Legendre polynomial, defined on [െ1,1], and is given by
(ଶିଶ)!
ܲ ( = )ݔσே
ୀ(െ1)
ݔିଶ , ܰ = ݊/2 or (݊ െ 1)/2 (1.9a)
ଶ ! (ି)! (ିଶ)!
according as ݊ is even or odd. First three members are ܲ ( = )ݔ1, ܲଵ (ݔ = )ݔ,
ܲଶ (( = )ݔ3 ݔଶ െ 1)/2. Orthogonality relation is
ଵ ଶ
ିଵ ܲ (ܲ)ݔ ()ݔd = ݔଶାଵ ߜ . (1.9b)
4
ends at = ݔ0 and ݈ = ݔ. The transverse vibration of string will start from a
disturbance given at equilibrium, and ߰(ݔ, )ݐdenotes the profile of this transverse
vibration. IC means that at = ݐ0, the profile ߰ of wave and the velocity ߰௧ of
wave are prescribed by ݂( )ݔand ݃()ݔ. The constant ܿ is the speed of wave
generated by vibration.
We will use Method of separation. Its concept is to assume solution in the
following separable form:
߰(ݔ, = )ݐX(x)T(t). (2.2)
We are to find X(x) and T(t). Substituting into Wave equation (2.1a),
ଵ ᇲᇲ ் ᇲᇲ
ܺ ᇱᇱ ܶ = మ
ܶ ᇱᇱ ܺ ฺ = = െ݇ ଶ . (2.3)
మ்
ŏConception of eigenvalue equation for matrix and differential operator are identical.
5
dependence on ݇.
Using BC in (2.1b)
గ
ܺ (0) = ܺ (݈) = 0 ฺ ܥଵ, = 0, sin(݈݇) = 0 ฺ ݇ = ݇ = , ݊ = 1,2,3 … (2.6)
Equation (2.6) gives all eigenvalues of both equations of (2.4). We see now that the
separation constant ݇ is real number, and takes values for all natural numbers.
Substituting for ܺ()ݔ, ܶ( )ݐinto assumed solution (2.2), we get
గ௫ గ గ
߰(ݔ, = )ݐ(ܶ)ݔ(ܺ = )ݐσஶ
ୀଵ sin ቀ ቁ ቂܣ cos ቀ ݐቁ + ܤ sin ቀ ݐቁቃ, (2.7)
where for convenience we have identified ܣ = ܥଶ, ܦଵ, |ୀ , ܤ = ܥଶ, ܦଶ, |ୀ ,
so that we are to determine two families of arbitrary constants {ܣ , ܤ }.
Using 1st IC in (2.1b),
గ௫
߰(ݔ, 0) = ݂( = )ݔσஶ
ୀଵ ܣ sin ቀ ቁ. (2.8)
Equation (2.8) is Fourier-sine series [see Sec 1]. Thus we get the constants {ܣ }:
ଶ గ௫
ܣ = ݂( )ݔsin ቀ ቁ d ݔ, ݊ = 1,2,3, ….. (2.9)
Hence, final solution of IBVP of (2.1) of 1D Wave Equation for finite string is
߰(ݔ, = )ݐ
ଶ గ௫ గ గ௫
σஶ
ୀଵ sin ቀ ቁ ቂcos ቀ ݐቁ ݂( )ݔsin ቀ ቁ dݔ +
గ గ௫
sin ቀ ݐቁ ݃( )ݔsin ቀ ቁ d ݔቃ. (2.12b)
గ
6
Example: Suppose a string is tightly tied at both ends at = ݔ0,2. If the string is
released from rest with a initial profile ݂( = )ݔsinଷ (ߨݔ/2), find the solution
߰(ݔ, )ݐof transverse vibration at any point (ݔ0 ݔ ݈) at any instant (ݐ 0).
Assume 2 m/s is the speed of wave, and all quantities are in ms unit.
Solution: Transverse vibration of a finite string is described by Wave equation:
߲ଶ߰ 1 ߲ଶ߰
= , [ א ݔ0,2], ݐ 0
߲ ݔଶ 4 ߲ ݐଶ
Note that ߰௧ (ݔ, )ݐdenotes velocity of wave at any position ݔ. In question, string is
released from rest, so that ߰௧ (ݔ, 0) = 0. Also, ߰(ݔ, )ݐdenotes profile of wave at
any position ݔat any instant ݐ, so that according to given data about initial profile,
߰(ݔ, 0) = sinଷ (ߨݔ/2). Thus IC are
߰(ݔ, 0) = sinଷ (ߨݔ/2), ߰௧ (ݔ, 0) = 0
Regarding BC, since string is tied at both ends, we have, as discussed above,
߰(0, (߰ = )ݐ2, = )ݐ0 , ݐ 0
Hence, to solve this IBVP, we will use Method of Separation of Variables. I’ll not
repeat the steps [see discussion above]. Final solution is
ஶ ଶ
݊ߨݔ ߨݔ ݊ߨݔ
߰(ݔ, = )ݐ ܣ sin ቀ ቁ cos(݊ߨ )ݐ, ܣ = න sinଷ ቀ ቁ sin ቀ ቁ dݔ
2 2 2
ୀଵ
3 ߨݔ 1 3ߨݔ
߰(ݔ, = )ݐsin ቀ ቁ cos(ߨ )ݐെ sin ൬ ൰ cos(3ߨ)ݐ
4 2 4 2