
system [34]. Peregrine’s system, also known as classical Boussinesq system in the case of flat bottom
topography, was rederived in one dimension along with a whole class of asymptotically equivalent
systems known as the abcd-systems [6]. The Cauchy problem on the real line for this class of Boussinesq
systems was studied in [36, 7, 35]. In scaled and non-dimensional variables, the abcd family of systems
takes the form ηt+ux+ε(ηu)x+σ2(auxxx −bηxxt) = 0 ,
ut+ηx+εuux+σ2(cηxxx −duxxt) = 0 ,(1.1)
where xdenotes the horizontal spatial independent variable, tis the time, η=η(x, t) is the free surface
elevation above a flat bottom located at depth D0=−1, u=u(x, t) is the horizontal velocity of the
fluid measured at depth θD0with θ∈[0,1], ε, σ are small parameters characterizing the nonlinearity
and the dispersion of the waves, and a, b, c, d are parameters given by
a=1
2θ2−1
3ν, b =1
2θ2−1
3(1 −ν),
c=1
21−θ2µ, d =1
21−θ2(1 −µ),µ, ν ∈R,
so that a+b+c+d= 1/3. The parameters εand σ2in the Boussinesq regime are of the same order,
and the Stokes number is S=ε/σ2=O(1). Hence, as usual, throughout this work we take ε=σ2for
simplicity and without loss of generality.
Some important representatives of the abcd family of systems (1.1) include systems that are well-
posed and also admit classical solitary waves as special solutions. Examples of such systems are the
Bona-Smith system (a= 0, b > 0, c≤0, d > 0), which includes the coupled Benjamin-Bona-Mahony
BBM-BBM system as a special case (c= 0), the classical Boussinesq system (a=b=c= 0, d > 0), the
Nwogu system (a < 0, b=c= 0, d > 0), and the regularized Nwogu system or “reverse” Bona-Smith
system (a < 0, b > 0, c= 0, d > 0). It should be noted that the Nwogu system was derived in [33]
as an alternative to the Peregrine system with the ability to choose the coefficients a, b, c, d so that the
linear dispersion relation becomes optimal compared to the corresponding linear dispersion relation of
the Euler equations.
In practical applications and numerical simulations, systems like those mentioned above are posed
in bounded domains. This fact highlights the need for appropriately formulated initial-boundary value
problems. An important issue in this direction is the choice of appropriate boundary conditions. In
particular, in the case of both the regularized and the non-regularized Nwogu systems, it is not a priori
clear how many boundary values — and of which type — must be specified as data for a well-posed
problem on the finite interval. One of the main results of this work is the identification of appropriate
boundary conditions on the finite interval for the following generalized Nwogu system:
ηt+ux+ε(ηu)x+ε(auxxx −bηxxt)=0,
ut+ηx+εuux−εduxxt = 0 ,(x, t)∈(−L, L)×(0, T ),(1.2)
with a < 0, b≥0, c= 0, d > 0, which contains both the regularized and the original Nwogu system,
for b > 0 and b= 0 respectively.
Appropriate boundary conditions for the generalized system (1.2) are identified through solving
the linear counterpart of that system via the unified transform. This method was first introduced by
Fokas in [18] (see also the monograph [19] and the review article [14]) and has since been employed
for the analysis of linear as well as initial-boundary value problems in various settings — see, for
example, [22, 24, 17, 20, 23, 25, 26, 27, 21, 28]. Recent developments have led to the advancement of the
unified transform to systems of PDEs, in particular via the work [13]. Exploiting the framework laid out
in [13] along with recent progress noted in [29] on the linearization of the classical Boussinesq equation
on the half-line, here we employ the unified transform to derive a novel, explicit solution formula for
the linearization of the generalized Nwogu system (1.2) on a finite interval:
ηt+ux+ε(auxxx −bηxxt) = 0 ,
ut+ηx−εduxxt = 0 ,(x, t)∈(−L, L)×(0, T ).(1.3)
2