Front. Appl. Math. Stat. Frontiers in Applied Mathematics and Statistics Front. Appl. Math. Stat. 2297-4687 Frontiers Media S.A. 10.3389/fams.2024.1466569 Applied Mathematics and Statistics Original Research Some class of nonlinear partial differential equations in the ring of copolynomials over a commutative ring Gefter Sergiy L. Piven' Aleksey L. * School of Mathematics and Computer Sciences, V. N. Karazin Kharkiv National University, Kharkiv, Ukraine

Edited by: Kateryna Buryachenko, Humboldt University of Berlin, Germany

Reviewed by: Annamaria Barbagallo, University of Naples Federico II, Italy

Boris Khesin, University of Toronto, Canada

 *Correspondence: Aleksey L. Piven' aleksei.piven@karazin.ua 22 11 2024 2024 10 1466569 18 07 2024 01 11 2024 Copyright © 2024 Gefter and Piven'. 2024 Gefter and Piven'

This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

We study the copolynomials, i.e., K-linear mappings from the ring of polynomials K[x] into the commutative ring K. With the help of the Cauchy–Stieltjes transform of a copolynomial, we introduce and examine a multiplication of copolynomials. We investigate the Cauchy problem related to the nonlinear partial differential equation ∂u∂t=aum0(∂u∂x)m1(∂2u∂x2)m2(∂3u∂x3)m3,   m0,m1,m2,m3∈ℕ0,   ∑j=03mj>0,   a∈K in the ring of copolynomials. To find a solution, we use the series of powers of the δ-function. As examples, we consider the Cauchy problem with the Euler–Hopf equation ∂u∂t+u∂u∂x=0, for a Hamilton–Jacobi type equation ∂u∂t=(∂u∂x)2, and for the Harry Dym equation ∂u∂t=u3∂3u∂x3.

 copolynomial δ-function partial differential equation Cauchy problem Cauchy-Stieltjes transform multiplication of copolynomials section-at-acceptance Dynamical Systems 1 Introduction

The first, second, and third order equations play an important role in the theory of nonlinear partial differential equations. A significant portion of classical nonlinear differential equations is dedicated to these classes (see, for example, [1–5]). In this paper, we examine a purely algebraic approach to study the special Cauchy problem with the following evolution equation:

∂u∂t=aum0(∂u∂x)m1(∂2u∂x2)m2(∂3u∂x3)m3 u(0,x)=u0δ(x).

We study this Cauchy problem in the module K[x]′ of the K-linear functionals on the ring of polynomials K[x], where K is an arbitrary commutative integral domain with identity and a, u0 ∈ K. We consider the module K[x]′ as an algebraic analog of space of distributions (see [6, 7]), where linear partial differential equations in the module K[x]′ were studied). In this paper, the elements of the module K[x]′ are called copolynomials (see Section 2). A copolynomial δ(x) is defined in the usual way: (δ, p) = p(0), p ∈ K[x]. A multiplication operation for copolynomials plays an important role for us. We define the product of copolynomials using the Cauchy–Stieltjes transform (see Section 3). We take note of several non-equivalent constructions of a multiplication that are considered in classical theories of distributions. For example, in the Colombeau theory [8, 9], the square of the δ-function is well-defined, but in some other theories it is not defined (see, for example, Antosik et al. [10]; Section 12.5).

In Section 4, we prove the existence and uniqueness theorem for the Cauchy problem (1.1), (1.2), and establish a representation of the solution in the form of the series in powers of the δ-function (Theorem 4.1). As examples, we consider the Cauchy problem for the Euler–Hopf equation ∂u∂t+u∂u∂x=0, for the Hamilton–Jacobi type equation ∂u∂t=(∂u∂x)2, and for the Harry Dym equation ∂u∂t=u3∂3u∂x3. In some of these examples, an interesting connection between classical nonlinear partial differential equations and well-known integer sequences is discovered (see examples 4.1, 4.2, and 4.4, where the Euler–Hopf equation, the Hamilton–Jacobi equation, and the Harry Dym equation are studied, respectively). Note that we restrict our consideration of equations of type (1.1) to those of the order no higher than three for two reasons. First, the representation in the proof of Theorem 4.1 generally becomes more cumbersome. Second, we are unaware of any classical examples of nonlinear equations of type (1.1) of order higher than three (see [3, 5]).

Linear functionals in the space of polynomials were extensively studied from different points of views in algebra, combinatorics, and the theory of orthogonal polynomials (cf., for example, [11–13]). In a classical case of (K = ℝ or K = ℂ), series with respect to derivatives of the δ-function are intensively studied because of their applications to differential and functional-differential equations and the theory of orthogonal polynomials [13]. Formal power series solutions of nonlinear partial differential equations were examined in a number of studies (cf., for example, [14–16]).

 2 Preliminary

Let K be an arbitrary commutative integral domain with identity, and let K[x] be a ring of polynomials with coefficients in K.

Definition 2.1. By a copolynomial over the ring K, we mean a K-linear functional defined on the ring K[x], i.e., a homomorphism occurring from the module K[x] to the ring K.

We denote the module of copolynomials over K by K[x]′. Thus, T ∈ K[x]′ if and only if T:K[x] → K and T has the property of K-linearity: T(ap + bq) = aT(p) + bT(q) for all p, q ∈ K[x] and a, b ∈ K. If T ∈ K[x]′ and p ∈ K[x], are for the value of T on p, we use the notation (T, p). We also write the copolynomial T ∈ K[x]′ in the form T(x), where x is regarded as the argument of polynomials p(x) ∈ K[x] and is subjected to the action of the K-linear mapping T. In this case, the result of action of T upon p can be represented in the form (T(x), p(x)).

Let p(x)=∑n=0manxn∈K[x]. For any x ∈ K, we consider the polynomial p(x + h) ∈ K[h]:

p(x+h)=∑n=0mpn(x)hn,

where pn(x) ∈ K. Since, in the case of a field with zero characteristic, pn(x)=p(n)(x)n!, we also assume that by definition p(n)(x)n!=pn(x), n=0,...,m is also true for any commutative ring K. For n > m, we assume that p(n)(x)n!=0.

Definition 2.2. The derivative T′ of a copolynomial T ∈ K[x]′, as in the classical case, is given in the formula

(T′,p)=-(T,p′),   p∈K[x].

By using this result, we arrive at the following expression for the nth order derivative:

(T(n),p)=(-1)n(T,p(n)),   p∈K[x].

Hence,

(T(n),p)=0,   T∈K[x]′,   p∈K[x], n>degp.

By virtue of the equality

(T(n)n!,p)=(-1)n(T,p(n)n!),   p∈K[x]

the copolynomials T(n)n! are well defined for any T ∈ K[x]′ and n ∈ ℕ .

Example 2.1. The copolynomial δ-function is given in the formula

(δ,p)=p(0),   p∈K[x].

For the copolynomial δ-function, we find its derivative of the nth order as follows:

(δ(n),p)=(-1)n(δ,p(n))=(-1)np(n)(0),   n∈ℕ.

Example 2.2. Let K = ℝ and let f : ℝ → ℝ be a Lebesgue-integrable function such that

∫-∞∞|xnf(x)|dx<+∞,   n=0,1,2,...

Then, f generates the regular copolynomial Tf:

(Tf,p)=∫-∞∞p(x)f(x)dx,   p∈ℝ[x].

Note that, in this case, unlike the classical theory, all copolynomials are regular ([13], Theorem 7.3.4), although a nonzero function f can generate the zero copolynomial {([17], Remark 1), ([18], Example 2.2)}. We present an example of a function that satisfies the property (2.2) and generates the δ-function.

It is known that for any ε > 0 there exists an even function φε(x)∈C0∞(ℝ) such that φε(x) = 1 for any x ∈ (−ε; ε) [19]. Then, φε(0) = 1 and φε(k)(0)=0, and k ∈ ℕ. The inverse Fourier transform

fε(x)=12π∫-∞∞φε(λ)eiλxdλ=12π∫-∞∞φε(λ)cosλxdλ

is an element of the Schwarz space S(ℝ). Then, φε(λ) is the Fourier transform of fε(x):

φε(λ)=∫-∞∞fε(x)e-iλxdx

and

∫-∞∞fε(x)dx=φε(0)=1,   ∫-∞∞xkfε(x)dx=ikφε(k)(0)=0,   k∈ℕ,∫-∞∞p(x)fε(x)dx=p(0),   p∈K[x],

i.e., fε(x) generates the copolynomial δ-function for any ε > 0.

We now consider the issue of convergence in the space K[x]′. In the ring K, we consider the discrete topology. Further, in the module of copolynomials K[x]′, we consider the topology of pointwise convergence. The convergence of a sequence {Tn}n=1∞ to T in K[x]′ means that for every polynomial p ∈ K[x], there exists a number n0 ∈ ℕ such that

(Tn,p)=(T,p),   n=n0,n0+1,n0+2,....

By the definition of convergence in the module K[x]′, we arrive at the following statement [6].

Theorem 2.1. Let {an}n=0∞ be a sequence of elements from K and let T ∈ K[x]′. Then, the series ∑n=0∞anT(n)n! converges in K[x]′.

The following assertion [6] shows the possibility of an expansion of an arbitrary formal generalized function in a series in the system {δ(n)n!}n=0∞ {see also ([12], Proposition 2.3) in the case K = ℂ}.

Lemma 2.1. Let T ∈ K[x]′. Then,

T=∑n=0∞(-1)n(T,xn)δ(n)n!. 3 Multiplication of copolynomials 3.1 The Cauchy–Stieltjes transform

Let K[[z,1z]] be the module of formal Laurent series with coefficients in K. For g∈K[[z,1z]] and g(z)=∑k=-∞∞gkzk, we naturally define the formal residue:

Res(g(z))=g-1.

Definition 3.1. Let T ∈ K[x]′. Consider the following formal Laurent series from the ring 1sK[[1s]]:

C(T)(s)=∑k=0∞(T,xk)sk+1.

The Laurent series C(T)(s) will be called the Cauchy–Stieltjes transform of a copolynomial T.

We may write informally as follows: C(T)(s)=(T,1s-x). Obviously, that the mapping C:K[x]′→1sK[[1s]] is an isomorphism of K-modules.

Proposition 3.1. (The inversion formula). Let T ∈ K[x]′ and p ∈ K[x]. Then,

(T,p)=Res(C(T)(s)p(s)).

Proof. It is sufficient to consider the case p(x) = xn for some n ∈ ℕ0. We have

C(T)(s)sn=∑k=0∞(T,xk)snsk+1.

Therefore, Res(C(T)(s)sn) = (T, xn).

Example 3.1. For the copolynomial δ-function, we have

C(δ)(s)=1s.

The following proposition shows that in some sense the differentiating commutes with the Cauchy–Stieltjes transform.

Proposition 3.2. For any T ∈ K[x]′, the equality

C(T(n))=C(T)(n),   n∈ℕ

holds valid.

Proof. It is sufficient to consider the case n = 1, so that

C(T′)(s)=∑k=0∞(T′,xk)sk+1==-∑k=1∞k(T,xk-1)sk+1=-∑k=0∞(k+1)(T,xk)sk+2=C(T)′(s). 3.2 Multiplication of copolynomials and its properties

The Cauchy–Stieltjes transform and Proposition 3.2 allow to introduce the multiplication operation on the module of copolynomials such that this operation is consistent with the differentiation.

Definition 3.2. Let T1,T2∈K[x]′, i.e., T1, T2 are copolynomials. Define their product by the following equality:

C(T1T2)=C(T1)C(T2),

i.e.,

T1T2=C-1(C(T1)C(T2)),

where C:K[x]′→1sK[[1s]] is a Cauchy–Stieltjes transform.

In the following lemma, the action of the product of copolynomials on monomials is expressed through the action of multipliers on monomials.

Lemma 3.1. Let T1,T2∈K[x]′ and n ∈ ℕ0. Then,

(T1T2,xn)={∑k=0n-1(T1,xk)(T2,xn-1-k),   n∈ℕ,0,   n=0.

Proof. By Equation 3.2, we have

C(T1T2)(s)=C(T1)(s)C(T2)(s)=∑k=0∞∑j=0∞(T1,xk)(T2,xj)sk+j+2 =∑n=1∞∑k=0n-1(T1,xk)(T2,xn-1-k)1sn+1.

Applying the inversion formula to the both part of this equality (see Proposition 3.1), we obtain (3.3).

Remark 3.1. Definition 3.2 means that the module of copolynomials K[x]′ with the introduced product is a associative commutative ring, which isomorphic to the ring of formal Laurent series 1sK[[1s]] with a natural product operation. In particular, the ring of copolynomials is an integral domain and this is a ring without identity.

Example 3.2. Let n = 1. With the help of Proposition 3.2, we find the square of δ-function:

C(δ2)(s)=(C(δ))2(s)=1s2=(-1s)′=(-C(δ))′=C(-δ′),

i.e.,

δ2=-δ′.

Moreover, by Equations 2.1, 3.1, we have

C(δ(n)n!)(s)=∑k=0∞(δ(n)n!,xk)1sk+1=∑k=0∞(δ,1n!dnxkdxn)(-1)nsk+1==(-1)nsn+1=(-1)n(C(δ))n+1,

so that

(-1)nδ(n)n!=δn+1,   n=0,1,2,...,

and therefore,

(δn)′=-nδn+1,   n∈ℕ.

Hence, by Theorem 2.1 and (3.4), the series

∑k=0∞ukδk+1=∑k=0∞(-1)kδ(k)k!uk

converges for any uk ∈ K.

Remark 3.2. By Lemma 2.1 and (3.4) for any copolynomial T ∈ K[x]′, the expansion in powers of the δ-function holds:

T=∑k=0∞(T,xk)δk+1.

Remark 3.3. The equalities (3.1) and (3.4) show that in a certain sense δ(x) and 1s are related (see also [1], p. 79).

 4 Main results and examples 4.1 Formal power series over the ring of copolynomials

The ring of formal power series in the form u(t,x)=∑k=0∞uk(x)tk with coefficients uk(x)∈K[x]′ will be denoted by K[x]′[[t]]. In this subsection, we remind several notations from Gefter and Piven' [6].

The partial derivative with respect to t of the series u(t, x) ∈ K[x]′[[t]] is defined by the formula

∂u∂t=∑k=1∞kuk(x)tk-1.

The partial derivative ∂u∂x of the series u(t, x) ∈ K[x]′[[t]] is defined as follows:

∂u∂x=∑k=0∞uk′(x)tk.

By (u(t, x), p(x)), we denote the action of u(t, x) ∈ K[x]′[[t]] on p(x) ∈ K[x], which is defined coefficient-wise.

(u(t,x),p(x))=∑k=0∞(uk(x),p(x))tk.

Thus, (u(t, x), p(x)) ∈ K[[t]].

 4.2 Existence and uniqueness theorem

Let a, u0 ∈ K and let mj ∈ ℕ0 (j = 0, 1, 2, 3), ∑j=03mj>0. Consider the Cauchy problem (1.1), (1.2) in the ring K[x]′[[t]]. We prove the following existence and uniqueness theorem for this Cauchy problem.

Theorem 4.1. Let K ⊃ ℚ. Then, the Cauchy problem (1.1), (1.2) has a unique solution in K[x]′[[t]]. This solution is in the form

u(t,x)=∑k=0∞ukδnk+1tk,

where uk ∈ K and n=∑j=03(j+1)mj-1. Moreover, for every t ∈ K, this series converges in the topology of K[x]′.

Proof. We will find the solution of the Cauchy problem (1.1), (1.2) in the form (4.1). Differentiating (4.1) on x and t and taking into account (3.5), we have

∂u∂t=∑k=0∞(k+1)uk+1δnk+n+1tk, ∂u∂x=-∑k=0∞(nk+1)ukδnk+2tk,∂2u∂x2=∑k=0∞(nk+1)(nk+2)ukδnk+3tk,∂3u∂x3=-∑k=0∞(nk+1)(nk+2)(nk+3)ukδnk+4tk.

Then,

um0=∑τ0=0∞∑|α|=τ0uα1⋯uαm0δnτ0+m0tτ0,(∂u∂x)m1=(-1)m1∑τ1=0∞∑|β|=τ1(nβ1+1)⋯(nβm1+1)uβ1⋯uβm1δnτ1+2m1tτ1,(∂2u∂x2)m2=∑τ2=0∞∑|γ|=τ2(nγ1+1)⋯(nγm2+1)(nγ1+2)⋯(nγm2+2)uγ1⋯uγm2δnτ2+3m2tτ2,(∂3u∂x3)m3=(-1)m3∑τ3=0∞∑|σ|=τ3(nσ1+1)⋯(nσm3+1)(nσ1+2)⋯(nσm3+2)··(nσ1+3)⋯(nσm3+3)uσ1⋯uσm3δnτ3+4m3tτ3,

where α, β, γ, σ are multi-indexes, α = (α1, ..., αm0), β = (β1, ..., βm1), γ = (γ1, ..., γm2), σ = (σ1, ..., σm3). Therefore,

aum0(∂u∂x)m1(∂2u∂x2)m2(∂3u∂x3)m3=(-1)m1+m3a∑k=0∞∑|τ|=k∑|α|=τ0uα1⋯uαm0··∑|β|=τ1(nβ1+1)⋯(nβm1+1)uβ1⋯uβm1··∑|γ|=τ2(nγ1+1)⋯(nγm2+1)(nγ1+2)⋯(nγm2+2)uγ1⋯uγm2··∑|σ|=τ3(nσ1+1)⋯(nσm3+1)(nσ1+2)⋯(nσm3+2)(nσ1+3)⋯(nσm3+3)uσ1⋯uσm3δnk+n+1tk,        

where τ = (τ0, τ1, τ2, τ3). Equating coefficients at δnk+n+1tk in right-hand sides of (4.2) and (4.3), we obtain

(k+1)uk+1=(-1)m1+m3a∑|τ|=k∑|α|=τ0uα1⋯uαm0··∑|β|=τ1(nβ1+1)⋯(nβm1+1)uβ1⋯uβm1··∑|γ|=τ2(nγ1+1)⋯(nγm2+1)(nγ1+2)⋯(nγm2+2)uγ1⋯uγm2··∑|σ|=τ3(nσ1+1)⋯(nσm3+1)(nσ1+2)⋯(nσm3+2)(nσ1+3)⋯(nσm3+3)uσ1⋯uσm3.

Since K ⊃ ℚ, we obtain that for any k ∈ ℕ0 the element uk+1 is uniquely expressed through u0, ..., uk. Now, if t ∈ K, then by Equation 3.4

u(t,x)=∑k=0∞ukδnk+1tk=∑k=0∞(-1)nkδ(nk)(nk)!uktk

so that the convergence of the series (4.1) follows from Theorem 2.1. Now, we prove the uniqueness of the solution of the Cauchy problem (1.1), (1.2) in the ring K[x]′[[t]]. We will find a solution of the Cauchy problem (1.1), (1.2) in the form

u(t,x)=∑k=0∞vk(x)tk,

where vk(x)∈K[x]′. Then, by the initial condition (1.2), we have v0(x) = u0δ(x). Substitute u(t, x) into Equation 1.1 and equate coefficients of tk. Then, there exist polynomials pk ∈ K[z1, ..., z4(k+1)] (k = 0, 1, 2, ...) such that

(k+1)vk+1(x)=pk(v0(x),∂v0∂x,∂2v0∂x2,∂3v0∂x3,...,                                 vk(x),∂vk∂x,∂2vk∂x2,∂3vk∂x3).

Since the ring K contains the field of rational numbers, from this we uniquely find uk(x), k ∈ ℕ:

vk(x)=k−1pk−1(v0(x),∂v0∂x,∂2v0∂x2,∂3v0∂x3,...,                      vk−1(x),∂vk−1∂x,∂2vk−1∂x2,∂3vk−1∂x3).

The proof is complete.

 4.3 Examples

We consider some examples of classical equations that illustrate Theorem 4.1. In what follows, we suppose that K is of characteristic 0 ([20], Section 1.43). We denote by F the quotient field of K. Obviously, K ⊃ ℤ and F ⊃ ℚ.

Example 4.1. Let u0 ∈ K. In K[x]′[[t]], consider the following Cauchy problem for the Euler–Hopf equation:

∂u∂t+u∂u∂x=0, u(0,x)=u0δ(x).

By Theorem 4.1, the Cauchy problem (4.4), (4.5) has a unique solution in F[x]′[[t]] and this solution can be represented in the form (4.1) of n = 2:

u(t,x)=∑k=0∞ukδ2k+1tk,

where uk ∈ F. Substituting (4.6) into (4.4), we obtain (see Proof of Theorem 4.1):

∑k=0∞(k+1)uk+1δ2k+3tk=∑k=0∞∑j=0k(2j+1)ujuk-jδ2k+3tk.

Equating coefficients at δ2k+3tk in (4.7), we have

(k+1)uk+1=∑j=0k(2j+1)ujuk-j,   k∈ℕ0.

Since

∑j=0k(2j+1)ujuk-j=(k+1)∑j=0kujuk-j,

the equality (4.8) implies

(k+1)uk+1=(k+1)∑j=0kujuk-j,   k∈ℕ0.

Since K is of characteristic 0, the equality (4.9) is reduced to the following recurrence equation:

uk+1=∑j=0kujuk-j,   k∈ℕ0.

If u0 = 1, then the solution of (4.10) is uk = Ck, where Ck=(k+1)-1(2kk) (k∈ℕ0) is the sequence of the Catalan numbers ([21], Section 7.5). Generally, the solution of (4.10) is in the form uk=Cku0k+1 (k∈ℕ0), so that

u(t,x)=∑k=0∞Ckδ2k+1u0k+1tk=∑k=0∞Ckδ(2k)(x)(2k)!u0k+1tk

(see Equation 3.4). Since u(t, x) ∈ K[x]′[[t]], it is a unique solution of the Cauchy problem (4.4), (4.5) in the ring K[x]′[[t]].

Remark 4.1. Note that for any t ∈ K, the series (4.11) converges in the topology of K[x]′. The Cauchy–Stieltjes transform of (4.11) is the following Laurent series ∑k=0∞Cku0k+1tkx2k+1. If K = ℝ, then this series is an expansion of the function w(t,x)=x-x2-4u0t2t in the domain D={(t,x)∈ℝ2:x>0,x2-4u0t>0}. The function w(t, x) is a classical solution of the Euler–Hopf equation (4.4) in the domain D.

Example 4.2. Let u0 ∈ K. In K[x]′[[t]], consider the following Cauchy problem for a Hamilton–Jacobi type equation ([5], Section 24.1.6):

∂u∂t=(∂u∂x)2, u(0,x)=u0δ(x).

By Theorem 4.1, the Cauchy problem (4.12), (4.13) has a unique solution in F[x]′[[t]] and this solution can be represented in the form (4.1) for n = 3:

u(t,x)=∑k=0∞ukδ3k+1tk,

where uk ∈ F. Substituting (4.14) into (4.4), we obtain (see Proof of Theorem 4.1):

∑k=0∞(k+1)uk+1δ3k+4tk=∑k=0∞∑j=0k(3j+1)(3(k-j)+1)ujuk-jδ3k+4tk.

Equating coefficients at δ3k+4tk in Equation 4.15, we have

(k+1)uk+1=∑j=0k(3j+1)(3(k-j)+1)ujuk-j,   k∈ℕ0.

We prove that yk=2kCk(3)k+1 is a solution of the recurrence Equation 4.16 with the initial condition u0 = 1, where Ck(3)=(3k+1)-1(3k+1k)=(3k)!k!(2k+1)! (k∈ℕ0) are the Fuss–Catalan numbers {[21], Section 7.5, Formula (7.67)}.

Consider the following combinatorial identity that was proved in Gould [22]:

43k+4(3k+4k)=∑j=0k23j+2(3j+2j)23(k-j)+2(3(k-j)+2k-j),   k∈ℕ0.

Since

23j+2(3j+2j)=2(3j+1)!j!(2j+1)!(2j+2)=1j+1(3j+1j),j∈ℕ0,

the equality (4.17) can be written in the form

43k+4(3k+4k)=∑j=0k1j+1(3j+1j)1k-j+1(3(k-j)+1k-j),   k∈ℕ0.

Since

43k+4(3k+4k)=4(3k+4)!k!(2k+4)!(3k+4)==2(3k+4)!(k+1)(2k+3)!k!(k+1)(k+2)(3k+4)=2(k+1)(k+2)(3k+4)(3k+42k+3)==2(k+1)(k+2)(3k+4)(3(k+1)+1k+1)=2(k+1)k+2Ck+1(3)=(k+1)yk+12k,

after the multiplication (4.18) by 2k, we have

(k+1)yk+1=∑j=0k2jj+1(3j+1j)2k-jk-j+1(3(k-j)+1k-j)==∑j=0k(3j+1)(3(k-j)+1)yjyk-j,   k∈ℕ0,

i.e., yk satisfy (4.16). Since yk=2k(3k)!(k+1)!(2k+1)! is the number of inequivalent rooted maps of some vertices {[23], p.409, Section 5 and Formula (5.7)}, we have yk ∈ ℤ (see also the integer sequence A000309 in Sloane [24]). Therefore, if u0 = 1, then uk = yk ∈ ℤ.

Now, we consider an arbitrary u0 ∈ K. Multiplying the equality

(k+1)yk+1=∑j=0k(3j+1)(3(k-j)+1)yjyk-j,   k∈ℕ0

by u0k+2, we obtain

u0k+2yk+1=1k+1∑j=0k(3j+1)(3(k-j)+1)u0j+1yju0k-j+1yk-j,k∈ℕ0.

Therefore, for any u0 ∈ K, the sequence uk=u0k+1yk∈K satisfies Equation 4.16. Hence, Equation 4.14 defines the unique solution to the Cauchy problem (4.12), (4.13) in K[x]′[[t]].

Example 4.3. Let b, u0 ∈ K. Consider the following Cauchy problem for the heat equation in K[x]′[[t]]

∂u∂t=b∂2u∂x2, u(0,x)=u0δ(x).

By Theorem 4.1, the Cauchy problem (4.19), (4.20) has a unique solution in F[x]′[[t]] and this solution can be represented in the form (4.1) for n = 2:

u(t,x)=∑k=0∞ukδ2k+1tk,

where uk ∈ F. Substituting (4.21) into (4.19), we obtain (see Proof of Theorem 4.1):

∑k=0∞(k+1)uk+1δ2k+3tk=b∑k=0∞(2k+1)(2k+2)ukδ2k+3tk.

Equating coefficients at δ3k+4tk in Equation 4.22, we have

(k+1)uk+1=b(2k+1)(2k+2)uk,   k∈ℕ0

Since K is of characteristic 0, this implies the following difference equation

uk+1=2b(2k+1)uk,   k∈ℕ0,

which, for any given u0 ∈ K, has the unique solution uk=(2b)k(2k-1)!!u0,   k∈ℕ0, where (−1)!! = 1. Therefore, the unique solution of the Cauchy problem (4.19, 4.20) is in the form

u(t,x)=∑k=0∞(2b)k(2k-1)!!u0δ2k+1tk=∑k=0∞bku0δ(2k)(x)k!tk

(see also Equation 3.4). Since u(t, x) ∈ K[x]′[[t]], it is a unique solution of the Cauchy problem (4.19, 4.20) in the ring K[x]′[[t]].

Now let K = ℝ, b > 0 and t > 0. Taking into account the equality (3.14) [6] from Equation 4.23, we arrive

(∑k=0∞(2b)k(2k-1)!!δ2k+1tk,xj)=14πbt∫-∞∞xje-x24btdx,   j∈ℕ0,

i.e.,

∑k=0∞(2b)k(2k-1)!!δ2k+1tk=14πbte-x24btinℝ[x]′.

Example 4.4. Let K ⊃ ℚ and u0 ∈ K. Consider the following Cauchy problem for the Harry Dym equation in the ring K[x]′[[t]] ([5], Section 13.1.4)

∂u∂t=u3∂3u∂x3 u(0,x)=u0δ(x).

By Theorem 4.1, the Cauchy problem (4.12, 4.13) has a unique solution in K[x]′[[t]] and this solution can be represented in the form (4.1) for n = 6:

u(t,x)=∑k=0∞ukδ6k+1tk,

where uk ∈ K. As in the proof of Theorem 4.1, we have

∂u∂t=∑k=0∞(k+1)uk+1δ6k+7tk, ∂3u∂x3=-∑k=0∞(6k+1)(6k+2)(6k+3)ukδ6k+4tk, u3=∑k=0∞∑|α|=kuα1uα2uα3δ6k+3tk,

where α = (α1, α2, α3). Substituting (4.27–4.29) into (4.24), we obtain

∑k=0∞(k+1)uk+1δ6k+7tk=-∑k=0∞∑|τ|=k(6τ4+1)(6τ4+2)(6τ4+3)uτ1uτ2uτ3uτ4δ6k+7tk,

where τ = (τ1, τ2, τ3, τ4). Equating coefficients at δ6k+7tk in the right-hand side of (4.30), we obtain

uk+1=-(k+1)-1∑|τ|=k(6τ4+1)(6τ4+2)(6τ4+3)uτ1uτ2uτ3uτ4.

Computer experiments demonstrate that the first 200 terms of the sequence uk are integers. Although this sequence is not found in the online encyclopedia of integer sequences [24], we formulate the conjecture that uk ∈ ℤ for all k ∈ ℕ0.

The following example shows that the condition K ⊃ ℚ is essential for the assertion of Theorem 4.1.

Example 4.5. Let K ⊃ ℚ. Consider the following Cauchy problem in K[x]′[[t]]:

∂u∂t=u(∂u∂x)2, u(0,x)=δ(x).

By Theorem 4.1, the Cauchy problem (4.31, 4.32) has a unique solution in K[x]′[[t]] and this solution can be represented in the form (4.1) for n = 4:

u(t,x)=∑k=0∞ukδ4k+1tk,

where u0 = 1. Substituting (4.33) into (4.31), we obtain

∑k=0∞(k+1)uk+1δ4k+5tk=∑k=0∞∑|τ|=k(4τ1+1)(4τ2+1)uτ1uτ2uτ3δ4k+5tk,

where τ = (τ1, τ2, τ3).

Equating coefficients at δ4k+5tk in the right-hand side of Equation 4.34, we obtain

uk+1=(k+1)-1∑|τ|=k(4τ1+1)(4τ2+1)uτ1uτ2uτ3,   k∈ℕ0.

This implies that u1 = 1 and u2=112∉ℤ. Therefore, the Cauchy problem (4.31), (4.32) in ℤ[x]′[[t]] has no solutions.

 5 Conclusion

We investigated the Cauchy problem of the nonlinear partial differential equation

∂u∂t=aum0(∂u∂x)m1(∂2u∂x2)m2(∂3u∂x3)m3,m0,m1,m2,m3∈ℕ0,   ∑j=03mj>0,   a∈K

in the ring of copolynomials. We have found a solution to this Cauchy problem, as the series in powers of the δ-function. We considered the Cauchy problem for the Euler–Hopf equation ∂u∂t+u∂u∂x=0, for a Hamilton–Jacobi type equation ∂u∂t=(∂u∂x)2 and for the Harry Dym equation ∂u∂t=u3∂3u∂x3. In the first two examples, an interesting connection between classical nonlinear partial differential equations and well-known integer sequences is revealed. The conjecture were formulated that all the coefficients of an expanding in powers of the δ-function of the solution of the Cauchy problem for the Harry Dym equation are integers.

 Data availability statement

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

 Author contributions

AP: Writing – original draft, Writing – review & editing. SG: Writing – original draft, Writing – review & editing.

 Funding

The author(s) declare financial support was received for the research, authorship, and/or publication of this article. This research was partially supported by Akhiezer Foundation grant, 2024.

The authors express their gratitude to Eugene Karolinsky and Sergey Poslavsky for their insightful discussions regarding the findings of the paper.

 Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

 Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

 References Arnold VI. Lectures on Partial Differential Equations. Berlin: Springer-Verlag (2004). 10.1007/978-3-662-05441-321954293 Calogero F Degasperis A. Spectral Transform and Solitons: Tolls to Solve and Investigate Nonlinear Evolution Equations. Amsterdam: North-Holland Publishing Company (1982). Galaktionov VA Svirshchevskii SR. Exact Solutions and Invariant Subspaces of Nonlinear Partial Differential Equations in Mechanics and Physics. Boca Raton, FL: Chapman &Hall/CRC Press (2007). 10.1201/9781420011623 Kruskal M. Nonlinear wave equations. In: Moser J, editor. Dynamical Systems, Theory and Applications (Lecture Notes in Phys. V. 38.). Berlin: Springer-Verb (1975). p. 310–54. 10.1007/3-540-07171-7_9 Polyanin AD Zaitsev VF. Handbook of Nonlinear Partial Differential Equations, 2nd Edn. Boca Raton, EL: Taylor & Francys Group (2012). Gefter SL Piven' AL. Linear partial differential equations in module of formal generalized functions over commutative ring. J Math Sci. (2021) 257:579–96. 10.1007/s10958-021-05505-0 Gefter SL Piven' AL. Linear partial differential equations in module of copolynomials of several variables over a commutative ring. J Math Phys Anal Geometry. Available at: http://arxiv.org/abs/2407.04122 (accessed 2024). Colombeau JF. Multiplication of Distributions. A tool in mathematics, numerical engineering and theoretical physics. Berlin: Springer-Verlag (1992). 10.1007/BFb0088954 Egorov YV. A contribution to the theory of generalized functions. Russian Math Surv. (1990) 45:1–49. 10.1070/RM1990v045n05ABEH002683 Antosik P Mikusinski J Sikorski R. Theory of Distributions. The Sequential Approach. Amsterdam: Elsevier Science Publ. Co (1973). Maroni P da Rocha Z. A new characterization of classical forms. Comm Appl Anal. (2001) 5:351–62. Garcia-Ardila JC Marcellán F Marriaga ME. Orthogonal Polynomials and Linear Functionals. An Algebraic Approach and Applications. EMS Series of Lectures in Mathematics. Berlin: EMS Press (2001). Estrada R Kanwal RP. A distributional approach to asymptotics theory and applications. Boston: Birkhäuser (2002). 10.1007/978-0-8176-8130-2 Ouchi S. Formal solutions with Gevrey type estimates of nonlinear partial differential equations. J Math Sci Univ Tokyo. (1994) 1:205–37. Balser W Kostov V. Recent progress in the theory of formal solutions for ODE and PDE. Appl Math Comput. (2003) 141:113–23. 10.1016/S0096-3003(02)00325-9 Lysik G Michalik S. Formal solutions of semilinear heat equations. J Math Anal Appl. (2008) 341:372–85. 10.1016/j.jmaa.2007.10.005 Gefter SL Stulova TE. Fundamental solution of the simplest implicit linear differential equation in a vector space. J Math Sci. (2015) 207:166–75. 10.1007/s10958-015-2363-z Gefter SL Piven' AL. Implicit linear differential-difference equations in the module of formal generalized functions over a commutative ring. J Math Sci. (2021) 25:409–22. 10.1007/s10958-021-05381-8 Hörmander L. The Analysis of Linear Partial Differential Equations. 1. Distribution Theory and Fourier Analysis. Berlin: Springer-Verlag (1983). Lidl R Niederreiter H. Finite Fields. Cambridge: Cambridge University Press (1996). 10.1017/CBO9780511525926 Graham RG Knutt DJ Patashnik O. Concrete Mathematics. A Foundation of Computer Sciences. Boston, MA: Addisson-Wesley (1994). Gould HW. Some generalizations of vandermonde's convolution. Amer Math Monthly. (1956) 63:84–91. 10.1080/00029890.1956.11988763 Tutte WT. A census of Hamiltonian polygons. Can J Math. (1962) 14:402–17. 10.4153/CJM-1962-032-x22026406 Sloane NJA. The On-Line Encyclopedia of Integer Sequences. Available at: https://oeis.org/ (accessed 2024).27882283