Abstract

Fractional diffusion equations describe an anomalous diffusion on fractals. In this paper, by means of the successive approximation method and other analysis technique, we present a local positive solution to Cauchy problem for a fractional diffusion equation with singular nonlinearity. The fractional derivative is described in the Caputo sense.

1. Introduction

In this paper, we consider the existence of a local positive solution to the Cauchy problem for fractional diffusion equation with singular nonlinearity𝐷𝛼𝑑𝑒(𝑑,π‘₯)=𝐷2Ξ”π‘’βˆ’π‘’βˆ’πœˆ,π‘₯βˆˆπ‘…π‘›],π‘‘βˆˆ(0,𝑇,𝜈>0,𝑒(0,π‘₯)=πœ™(π‘₯)∈𝐢LB(𝑅𝑛),π‘₯βˆˆπ‘…π‘›,(1.1) where 𝐷𝛼𝑑 is a regularized fractional derivative (the Caputo derivative) of order 0<𝛼≀1 with respect to 𝑑, defined by𝐷𝛼𝑑1𝑒(𝑑,π‘₯)=ξ‚Έπœ•Ξ“(1βˆ’π›Ό)ξ€œπœ•π‘‘π‘‘0(π‘‘βˆ’πœ)βˆ’π›Όπ‘’(𝜏,π‘₯)π‘‘πœβˆ’π‘‘βˆ’π›Όξ‚Ήπ‘’(0,π‘₯),0<𝛼≀1,(1.2)∫1/Ξ“(1βˆ’π›Ό)𝑑0(π‘‘βˆ’πœ)βˆ’π›Όπ‘’(𝜏,π‘₯)π‘‘πœ=𝐼𝑑1βˆ’π›Όπ‘’(𝑑,π‘₯) is the Riemann-Liouville fractional integral of order 1βˆ’π›Ό with respect to 𝑑; 𝐷 is a positive constant, 𝑛β‰₯1,𝐢LB(𝑅𝑛)=πœ™βˆˆπΆ(𝑅𝑛)βˆΆπœ™>0inπ‘…π‘›πœ™onlywithmin=minπ‘…π‘›πœ™>0,thereexistπ‘˜β‰₯0and𝑀>0onlysuchthat|π‘₯|βˆ’π‘˜ξ€Ύ.πœ™(π‘₯)≀Λ(Ξ›>0)for|π‘₯|β‰₯𝑀(1.3)

Fractional diffusion equations describe an anomalous diffusion on fractals (physical objects of fractional dimension, like some amorphous semiconductors or strongly porous materials); see [1, 2]. We know that there is a few papers dealing with solutions to some initial value problems for linear fractional diffusion equations, by means of the Laplace transform, Fourier transform, and so forth; see [3–6]. There are also some papers concerned with initial value problems for nonlinear fractional diffusion equations, in which the existence of solutions is considered, using some estimates and the monotone iterative method; see [7–9]. For the details for fractional diffusion equations, please see articles [10, 11] written by Kilbas and Trujillo. However, as far as we know there are few papers which consider the existence of local positive solution to the Cauchy problem (1.1). In this paper, we consider the local existence of positive solutions to singular problem (1.1).

2. Main Result

In this section, we will establish the local existence of positive solutions to the Cauchy problem (1.1). In what follows, 𝐢=𝐢(β‹―) denotes positive constants, besides the arguments inside the parenthesis, which vary from line to line.

Lemma 2.1 (see [4]). The linear initial value problem π·π›Όπ‘‘π‘’βˆ’π·2]Δ𝑒=π‘ž(𝑑,π‘₯),π‘‘βˆˆ(0,𝑇,π‘₯βˆˆπ‘…π‘›π‘’,𝐷>0,(0,π‘₯)=πœ™(π‘₯),π‘₯βˆˆπ‘…π‘›,(2.1) with a bounded continuous function πœ™ (locally HΓΆlder continuous if 𝑛>1) and a bounded jointly continuous and locally HΓΆlder continuous, in π‘₯, function π‘ž, has the following integral representation of solution: ξ€œπ‘’(𝑑,π‘₯)=𝑅𝑛𝑍0(+ξ€œπ‘‘,π‘₯βˆ’πœ‰;πœ‰)πœ™(πœ‰)π‘‘πœ‰π‘‘0ξ€œπ‘‘πœπ‘…π‘›π‘Œ0(π‘‘βˆ’πœ,π‘₯βˆ’πœ‰;πœ‰)π‘ž(𝜏,πœ‰)π‘‘πœ‰,(2.2) where 𝑍0πœ‹(𝑑,π‘₯βˆ’π‘¦)=βˆ’π‘›/2𝐷||||π‘₯βˆ’π‘¦βˆ’π‘›π»2012×14π‘‘βˆ’π›Ό||||π‘₯βˆ’π‘¦2|||ξ‚„,π‘Œ(1,𝛼)(𝑛/2,1),(1,1)0πœ‹(𝑑,π‘₯βˆ’π‘¦)=βˆ’π‘›/2𝐷||||π‘₯βˆ’π‘¦βˆ’π‘›π‘‘π›Όβˆ’1𝐻2012×14π‘‘βˆ’π›Ό||||π‘₯βˆ’π‘¦2|||ξ‚„.(𝛼,𝛼)(𝑛/2,1),(1,1)(2.3)𝐻2012 is an 𝐻-function defined by 𝐻2012(𝑧)=𝐻2012𝑧|||ξ‚„=1(1,𝛼)(𝑛/2,1),(1,1)ξ€œ2πœ‹π‘–π’žΞ“(𝑛/2+𝑠)Ξ“(1+𝑠)Γ𝑧(1+𝛼𝑠)βˆ’π‘ π‘‘π‘ ,(2.4) where π’ž is an infinite contour [12]. Moreover, the functions 𝑍0 and π‘Œ0 are nonnegative. 𝑍0 satisfies the relation ξ€œπ‘…π‘›π‘0(𝑑,π‘₯βˆ’π‘¦)𝑑𝑦=1.(2.5)

Remark 2.2. From [4], π‘Œ0(𝑑,π‘₯) is the Riemann-Liouville derivative of 𝑍0(𝑑,π‘₯) in 𝑑 of the order 1βˆ’π›Ό (for π‘₯β‰ 0, 𝑍0(𝑑,π‘₯)β†’0 as 𝑑→0, so that the Riemann-Liouville derivative 𝐷𝑑1βˆ’π›Όπ‘0(𝑑,π‘₯) coincides in this case with the regularized fractional derivative πΌπ›Όπ‘‘πœ•π‘0(𝑑,π‘₯)/πœ•π‘‘ and that 𝐷𝑑1βˆ’π›Όπ‘0πœ‹(𝑑,π‘₯)=βˆ’π‘›/2𝐷||||π‘₯βˆ’π‘¦βˆ’π‘›π‘‘π›Όβˆ’1𝐻2012×14π‘‘βˆ’π›Ό||||π‘₯βˆ’π‘¦2|||ξ‚„.(𝛼,𝛼)(𝑛/2,1),(1,1)(2.6) Therefore, together with the formulas 𝐼𝛼𝑑𝐼𝛽𝑑𝑓(𝑑,π‘₯)=𝐼𝛽𝑑𝐼𝛼𝑑𝑓(𝑑,π‘₯) and 𝐼1𝑑𝐷1𝑑𝑓(𝑑,π‘₯)=𝑓(𝑑,π‘₯)βˆ’π‘“(0,π‘₯), 𝑓 is summable with respect to 𝑑, we obtain that the solution (2.2) has the following form: ξ€œπ‘’(𝑑,π‘₯)=𝑅𝑛𝑍0(1𝑑,π‘₯βˆ’πœ‰;πœ‰)πœ™(πœ‰)π‘‘πœ‰+ξ€œΞ“(𝛼)𝑑0(π‘‘βˆ’πœ)π›Όβˆ’1Γ—ξ€œπ‘‘πœπ‘…π‘›π‘0(π‘‘βˆ’πœ,π‘₯βˆ’πœ‰;πœ‰)π‘ž(𝜏,πœ‰)π‘‘πœ‰=𝐸𝑑Δ1πœ™+ξ€œΞ“(𝛼)𝑑0(π‘‘βˆ’πœ)π›Όβˆ’1𝐸(π‘‘βˆ’πœ)Ξ”π‘ž(𝜏,π‘₯)π‘‘πœ,(2.7) where πΈπ‘‘Ξ”πœ‹πœ™=βˆ’π‘›/2π·ξ€œπ‘…π‘›||||π‘₯βˆ’π‘¦βˆ’π‘›π»2012×14π‘‘βˆ’π›Ό||||π‘₯βˆ’π‘¦2|||ξ‚„(1,𝛼)(𝑛/2,1),(1,1)πœ™(𝑦)𝑑𝑦.(2.8)
Consequently, from the Lemma 2.1 and Remark 2.2, we see that the solution of (1.1) has the form 𝑒(𝑑,π‘₯)=𝐸𝑑Δ1πœ™βˆ’ξ€œΞ“(𝛼)𝑑0(π‘‘βˆ’π‘ )π›Όβˆ’1𝐸(π‘‘βˆ’π‘ )Ξ”π‘’βˆ’πœˆ([]𝑠,π‘₯)𝑑𝑠,forall(𝑑,π‘₯)∈0,𝑇×𝑅𝑛.(2.9)

Remark 2.3. In (2.9), it follows from the definition of 𝐸𝑑Δ,𝑑β‰₯0 and 0≀𝑠≀𝑑, that 𝐸(π‘‘βˆ’π‘ )Ξ” is well defined.

Definition 2.4. We call a function 𝑒 a local positive solution of (1.1) on 𝑅𝑛×[0,𝑇] if(i)π‘’βˆˆπΆ([0,π‘‡ξ…ž]×𝑅𝑛) with 𝑒min∢=min[0,𝑇′]×𝑅𝑛𝑒>0 for any 0<π‘‡ξ…žβ‰€π‘‡,(ii)𝑒(𝑑,π‘₯)=πΈπ‘‘Ξ”βˆ«πœ™βˆ’(1/Ξ“(𝛼))𝑑0(π‘‘βˆ’π‘ )π›Όβˆ’1𝐸(π‘‘βˆ’π‘ )Ξ”π‘’βˆ’πœˆ(s,π‘₯)𝑑𝑠 for all (𝑑,π‘₯)∈[0,π‘‡ξ…ž]×𝑅𝑛.

The next theorem is the main result of this paper.

Theorem 2.5. Let πœ™βˆˆπΆLB(𝑅𝑛). Then, (1.1) has a local positive solution 𝑣.

Proof. Defining 𝜌=πœ™min>0, we will find 0<𝑑0≀𝑇 and ΜƒπœŒ>0 depending on 𝜌, 𝜈, and 𝑛 such that (1.1) has a positive solution 𝑣(𝑑,π‘₯) satisfying 𝑣(𝑑,π‘₯)β‰₯ΜƒπœŒfor(𝑑,π‘₯)∈0,𝑑0×𝑅𝑛.(2.10) Let 𝐹(𝑒)=𝐸𝑑Δ1πœ™βˆ’ξ€œΞ“(𝛼)𝑑0(π‘‘βˆ’π‘ )π›Όβˆ’1𝐸(π‘‘βˆ’π‘ )Ξ”π‘’βˆ’πœˆ(𝑠,π‘₯)𝑑𝑠.(2.11) For 0<ΜƒπœŒ<𝜌, which will be determined below, we construct a sequence {π‘’π‘˜} as follows π‘’π‘˜(𝑑,π‘₯)=𝐸𝑑Δ1πœ™βˆ’ξ€œΞ“(𝛼)𝑑0(π‘‘βˆ’π‘ )π›Όβˆ’1𝐸(π‘‘βˆ’π‘ )Ξ”π‘’βˆ’πœˆπ‘˜βˆ’1(𝑠,π‘₯)𝑑𝑠,π‘˜=0,1,2,…,(2.12) with π‘’βˆ’1(𝑑,π‘₯)=ΜƒπœŒ. For 0<ΜƒπœŒ<π‘šπ‘–π‘›{1,𝜌/2}, by (2.5), we have πΈπ‘‘Ξ”πœ™=πœ‹βˆ’π‘›/2π·ξ€œπ‘…π‘›||||π‘₯βˆ’π‘¦βˆ’π‘›π»2012×14π‘‘βˆ’π›Ό||||π‘₯βˆ’π‘¦2|||(ξ‚„β‰₯πœ‹(1,𝛼)𝑛/2,1),(1,1)πœ™(𝑦)π‘‘π‘¦βˆ’π‘›/2πœŒπ·ξ€œπ‘…π‘›||||π‘₯βˆ’π‘¦βˆ’π‘›π»2012×14π‘‘βˆ’π›Ό||||π‘₯βˆ’π‘¦2|||(ξ‚„(1,𝛼)𝑛/2,1),(1,1)𝑑𝑦=𝜌β‰₯2ΜƒπœŒ.(2.13) Also, we can obtain that πΈπ‘‘Ξ”πœ‹πœ™=βˆ’π‘›/2π·ξ€œπ‘…π‘›||||π‘₯βˆ’π‘¦βˆ’π‘›π»2012×14π‘‘βˆ’π›Ό||||π‘₯βˆ’π‘¦2|||ξ‚„β‰€πœ‹(1,𝛼)(𝑛/2,1),(1,1)πœ™(𝑦)π‘‘π‘¦βˆ’π‘›/2π·ξ€œπ‘…π‘›||||π‘₯βˆ’π‘¦βˆ’π‘›π»201214π‘‘βˆ’π›Ό||||π‘₯βˆ’π‘¦2Γ—|||𝐢||𝑦||ξ€Έ(1,𝛼)(𝑛/2,1),(1,1)1+π‘˜=πœ‹π‘‘π‘¦βˆ’π‘›/2π·ξ€œπ‘…π‘›πœ‚βˆ’π‘›π»201214||πœ‚||2|||ξ‚„ξ€·(1,𝛼)(𝑛/2,1),(1,1)×𝐢1+|π‘₯|+𝑑𝛼/2||πœ‚||ξ€Έπ‘˜π‘‘πœ‚β‰€2π‘˜πΆ(1+|π‘₯|)π‘˜+2π‘˜πΆπ΅π‘‘π‘˜π›Ό/2≀𝐢(1+|π‘₯|)π‘˜,(2.14) where 𝐡=(πœ‹βˆ’π‘›/2∫/𝐷)π‘…π‘›πœ‚π‘˜βˆ’π‘›π»2012[(1/4)|πœ‚|2||(1,𝛼)(𝑛/2,1),(1,1)]π‘‘πœ‚.
On the other hand, if 0<𝑑≀(Ξ“(1+𝛼)ΜƒπœŒ1+𝜈)1/𝛼, by (2.5), we have 𝑒0(𝑑,π‘₯)=𝐸𝑑Δ1πœ™βˆ’ξ€œΞ“(𝛼)𝑑0(π‘‘βˆ’π‘ )π›Όβˆ’1𝐸(π‘‘βˆ’π‘ )Ξ”ΜƒπœŒβˆ’πœˆπ‘‘π‘ β‰₯2ΜƒπœŒβˆ’ΜƒπœŒβˆ’πœˆΞ“π‘‘(1+𝛼)𝛼β‰₯ΜƒπœŒ.(2.15) Moreover, it is obvious that 𝑒0(𝑑,π‘₯)≀𝐢(1+|π‘₯|)π‘˜,for0<𝑑≀𝑇,π‘₯βˆˆπ‘…π‘›.(2.16) Hence, it holds that ΜƒπœŒβ‰€π‘’0(𝑑,π‘₯)≀𝐢(1+|π‘₯|)π‘˜ξ€Ίfor(𝑑,π‘₯)∈0,𝑑0×𝑅𝑛,(2.17) where 𝑑0=min{𝑇,(Ξ“(1+𝛼)ΜƒπœŒ1+𝜈)1/𝛼}. From the previous arguments, it is very easy to show that ΜƒπœŒβ‰€π‘’π‘˜(𝑑,π‘₯)≀𝐢(1+|π‘₯|)π‘˜ξ€Ίfor(𝑑,π‘₯)∈0,𝑑0×𝑅𝑛andallπ‘˜β‰₯0.(2.18) Now, we show that ||π‘’π‘˜βˆ’π‘’π‘˜βˆ’1||β‰€πœˆπ‘˜βˆ’1ξ‚΅2ΜƒπœŒπœˆ+1ξ‚Άπ‘˜1𝑑Γ(1+π‘˜π›Ό)π‘˜π›Ό,βˆ€π‘˜β‰₯1.(2.19) In fact, we have that ||𝑒1βˆ’π‘’0||≀1ξ€œΞ“(𝛼)𝑑0(π‘‘βˆ’π‘ )π›Όβˆ’1𝐸(π‘‘βˆ’π‘ )Ξ”||𝑒0βˆ’πœˆβˆ’ΜƒπœŒβˆ’πœˆ||≀2π‘‘π‘ ΜƒπœŒπœˆ1𝑑Γ(1+𝛼)𝛼≀2𝜈0ΜƒπœŒπœˆ+11𝑑Γ(1+𝛼)𝛼.(2.20) If we assume that (2.19) holds for π‘˜=π‘šβˆ’1(π‘š>2), next, we will show that (2.19) is valid for π‘˜=π‘š. Indeed, since π‘’βˆ’πœˆπ‘šβˆ’1βˆ’π‘’βˆ’πœˆπ‘šβˆ’2=βˆ’πœˆπœ‰βˆ’(𝜈+1)ξ€·π‘’π‘šβˆ’1βˆ’π‘’π‘šβˆ’2𝑒=βˆ’πΆ(𝑑,π‘₯)π‘šβˆ’1βˆ’π‘’π‘šβˆ’2ξ€Έ,(2.21) where πœ‰=π‘ π‘’π‘šβˆ’1+(1βˆ’π‘ )π‘’π‘šβˆ’2 with π‘ βˆˆ(0,1), it follows from (2.18) that 𝐢(𝑑,π‘₯)β‰€πœˆΜƒπœŒβˆ’(𝜈+1).
Hence, we can obtain that ||π‘’π‘šβˆ’π‘’π‘šβˆ’1||≀1ξ€œΞ“(𝛼)𝑑0(π‘‘βˆ’π‘ )π›Όβˆ’1𝐸(π‘‘βˆ’π‘ )Ξ”||π‘’βˆ’πœˆπ‘šβˆ’1βˆ’π‘’βˆ’πœˆπ‘šβˆ’2||≀1π‘‘π‘ ξ€œΞ“(𝛼)𝑑0(π‘‘βˆ’π‘ )π›Όβˆ’1𝐸(π‘‘βˆ’π‘ )Ξ”||𝑒𝐢(𝑑,π‘₯)π‘šβˆ’1βˆ’π‘’π‘šβˆ’2||β‰€πœˆπ‘‘π‘ ΜƒπœŒπœˆ+11ξ€œΞ“(𝛼)𝑑0(π‘‘βˆ’π‘ )π›Όβˆ’1𝐸(π‘‘βˆ’π‘ )Ξ”||π‘’π‘šβˆ’1βˆ’π‘’π‘šβˆ’2||β‰€πœˆπ‘‘π‘ ΜƒπœŒπœˆ+1πœˆπ‘šβˆ’2ξ‚΅2ΜƒπœŒπœˆ+1ξ‚Άπ‘šβˆ’111Ξ“(1+(π‘šβˆ’1)𝛼)Γ—ξ€œΞ“(𝛼)𝑑0(π‘‘βˆ’π‘ )π›Όβˆ’1𝐸(π‘‘βˆ’π‘ )Δ𝑠(π‘šβˆ’1)𝛼𝑑𝑠=πœˆπ‘šβˆ’12π‘šβˆ’1ξ€·ΜƒπœŒπœˆ+1ξ€Έπ‘šΞ“(1+(π‘šβˆ’1)𝛼)𝑑Γ(1+(π‘šβˆ’1)𝛼)Ξ“(1+π‘šπ›Ό)π‘šπ›Όβ‰€πœˆπ‘šβˆ’1ξ‚΅2ΜƒπœŒπœˆ+1ξ‚Άπ‘š1𝑑Γ(1+π‘šπ›Ό)π‘šπ›Ό.(2.22) Therefore, (2.19) is valid for all π‘˜=1,2,…. Thus, we see that, for 0<𝑑≀𝑑0,π‘₯βˆˆπ‘…π‘›, it holds that ||π‘’π‘˜βˆ’π‘’π‘˜βˆ’1||β‰€πœˆπ‘˜βˆ’1ξ‚΅2ΜƒπœŒπœˆ+1ξ‚Άπ‘˜1𝑑Γ(1+π‘˜π›Ό)0π‘˜π›Ό.(2.23)
Define πœπ‘˜=βˆ‘π‘˜π‘—=1(π‘’π‘—βˆ’π‘’π‘—βˆ’1). Since |πœπ‘˜βˆ‘|β‰€π‘˜π‘—=1πœˆπ‘—βˆ’1(2/ΜƒπœŒπœˆ+1)𝑗(𝑑0𝑗𝛼/Ξ“(1+𝑗𝛼)), we can claim that πœπ‘˜β†’πœ uniformly in [0,𝑑0]×𝑅𝑛 as π‘˜β†’βˆž and 𝜁∈𝐢([0,𝑑0]×𝑅𝑛). In fact, from the Stirling formula, we have that βˆšΞ“(1+𝑛𝛼)=ξ‚€2πœ‹π‘›π›Όπ‘›π›Όπ‘’ξ‚π‘›π›Όπ‘’πœƒ/12𝑛𝛼,(2.24) where 0<πœƒ<1. We let βˆ‘βˆžπ‘—=1πœˆπ‘—βˆ’1(2/ΜƒπœŒπœˆ+1)𝑗(𝑑0π‘—π›Όβˆ‘/Ξ“(1+𝑗𝛼))=βˆžπ‘—=1π‘₯𝑗; thus, we have that π‘₯𝑗+1π‘₯𝑗=2πœˆπ‘‘π›Ό0Ξ“(1+𝑗𝛼)ΜƒπœŒπœˆ+1=Ξ“(1+(𝑗+1)𝛼)2πœˆπ‘‘π›Ό0ΜƒπœŒπœˆ+1√2πœ‹π‘—π›Ό(𝑗𝛼/𝑒)π‘—π›Όπ‘’πœƒ/12π‘—π›Όβˆš2πœ‹(𝑗+1)𝛼((𝑗+1)𝛼/𝑒)(𝑗+1)π›Όπ‘’πœƒ/12(𝑗+1)𝛼≀2πœˆπ‘‘π›Ό0ΜƒπœŒπœˆ+1√2πœ‹(𝑗+1)𝛼((𝑗+1)𝛼/𝑒)π‘—π›Όπ‘’πœƒ/12π‘—π›Όβˆš2πœ‹(𝑗+1)𝛼((𝑗+1)𝛼/𝑒)(𝑗+1)π›Όπ‘’πœƒ/12(𝑗+1)𝛼=2πœˆπ‘‘π›Ό0ΜƒπœŒπœˆ+1π‘’πœƒ/12𝑗(𝑗+1)𝛼(𝑗+1)π›Όπ‘’ξ‚Άβˆ’π›Ό,0≀limπ‘—β†’βˆžπ‘₯𝑗+1π‘₯𝑗≀limπ‘—β†’βˆž2πœˆπ‘‘π›Ό0ΜƒπœŒπœˆ+1π‘’πœƒ/12𝑗(𝑗+1)𝛼(𝑗+1)π›Όπ‘’ξ‚Άβˆ’π›Ό=0,(2.25) that is, limπ‘—β†’βˆžπ‘₯𝑗+1π‘₯𝑗=0;(2.26) then, by the convergence principle of D’Alembert, we can claim that πœπ‘˜β†’πœ uniformly in [0,𝑑0]×𝑅𝑛 as π‘˜β†’βˆž and 𝜁∈𝐢([0,𝑑0]×𝑅𝑛). Moreover, we can obtain that π‘’π‘˜=𝑒0+πœπ‘˜β†’π‘’0+𝜁∢=𝑣(𝑑,π‘₯) uniformly in [0,𝑑0]×𝑅𝑛, as π‘˜β†’βˆž, and 𝑣(𝑑,π‘₯)∈𝐢([0,𝑑0]×𝑅𝑛). Combining with (2.18), we have 𝑣(𝑑,π‘₯)β‰₯ΜƒπœŒin0,𝑑0×𝑅𝑛,(2.27) and, letting π‘˜β†’βˆž in (2.12), we have 𝑣(𝑑,π‘₯)=𝐸𝑑Δ1πœ™βˆ’ξ€œΞ“(𝛼)𝑑0(π‘‘βˆ’π‘ )π›Όβˆ’1𝐸(π‘‘βˆ’π‘ )Ξ”π‘£βˆ’πœˆ(𝑠,π‘₯)𝑑𝑠.(2.28) This mean that 𝑣(𝑑,π‘₯) is a local positive to the Cauchy problem (1.1). Thus, the proof is complete.

Remark 2.6. Under the assumption that πœ™min𝑇=𝜌>2𝛼Γ(1+𝛼)1/(1+𝑣),(2.29) by a similar proof to that of Theorem 2.5, we can claim that problem (1.1) has positive solution on the interval [0,𝑇]×𝑅𝑛. In fact, we may let 𝑇𝛼Γ(1+𝛼)1/(1+𝑣)ξ‚†πœŒ<ΜƒπœŒ<1,2;(2.30) then we can obtain that 𝑒0(𝑑,π‘₯)=𝐸𝑑Δ1πœ™βˆ’ξ€œΞ“(𝛼)𝑑0(π‘‘βˆ’π‘ )π›Όβˆ’1𝐸(π‘‘βˆ’π‘ )Ξ”ΜƒπœŒβˆ’πœˆπ‘‘π‘ β‰₯2ΜƒπœŒβˆ’ΜƒπœŒΜƒπœŒβˆ’1βˆ’πœˆπ‘‘Ξ“(1+𝛼)𝛼β‰₯2ΜƒπœŒβˆ’ΜƒπœŒΜƒπœŒβˆ’1βˆ’πœˆπ‘‡Ξ“(1+𝛼)𝛼β‰₯2ΜƒπœŒβˆ’ΜƒπœŒ=ΜƒπœŒ.(2.31) The remaining proofs may be finished by the same way as the Theorem 2.5.

3. Conclusion

In this paper, we obtain the existence of a local (or nonlocal) positive of problem (1.1), that is, our main results: Theorem 2.5 and Remark 2.2. It is well known that fractional-order models are more accurate than integer-order models, that is, there are more degrees of freedom in the fractional-order models. Problem (1.1) may appear in several applications in mechanics and physics and in particular can be used to model the electrostatic micro-electromechanic System devices, in which the derivatived 𝑒𝑑(𝑑,π‘₯) should be replace by 𝐷𝛼𝑑𝑒(𝑑,π‘₯),0<𝛼≀1.

Acknowlegment

This paper is supported by the NNSF of china (10971221), the Ministry of Education for New Century Excellent Talent Support Program (NCET-10-0725), and the Fundamental Research Funds for the Central Universities (2009QS06).