You are on page 1of 15


abstract. The complex phenomenon of solid wax deposition in wax saturated crude oils subject to thermal gradients has been treated in a number of papers under very specic assumptions (e.g. thermodynamical equilibrium between dissolved wax and the wax suspended in the oil as a crystallized phase). Here we want to consider a more general framework in which thermodynamical equilibrium may not exist, the whole system may form a gel-like structure in which the segregated solid wax has no diusivity, the thermal eld may evolve due to a non-negligible dierence between the thermal conductivity of the solid wax deposit growing at the cold wall of the container and the conductivity of oil, etc. Riassunto. Il complesso processo della deposizione delle cere nei greggi cerosi sovrasaturi e soggetti a gradiente termico è stato trattato in numerosi articoli sotto speciche ipotesi (ad es. equilibrio termodinamico tra cere disciolte e cere in sospensione in fase cristallina). In questa nota si considera uno schema più generale in cui può non aversi equilibrio termodinamico in cui il sistema può gelicare sì da impedire la diusione delle cere cristalline, in cui il campo termico può essere modicato per eetto del deposito sulla parete fredda, ecc.

1. Introduction

Crude oils are complex mixtures containing parans, aromatics, naphtenics, resins, asphaltenes and other impurities. For the purpose of our analysis, we can consider a solution of a high molecular weight paran (referred to as wax in this paper) in a solvent (referred to as oil ). The solubility of wax in oil decreases very sharply with temperature. At oshore reservoir temperatures the solubility is suciently high to keep wax fully dissolved. But when oil is pumped in subsea pipelines, where the external temperature can be a few degrees centigrades, the decreased solubility causes the formation of wax crystals and at the same time deposition of solid wax on the pipe wall, due to diusional migration of wax in the saturated solution induced by the radial thermal gradient. The enormous economic relevance of this phenomenon stimulated several studies, laboratory experiments and eld measurements (see, e.g., [1], [2], [7], [15], [16], [17]). In the framework of a research contract with Enitecnologie, we have proposed mathematical models ([9], [10], [11]) for the phenomenon and applied them to the interpretation of data form an experimental device called cold nger ([5], [6]). Such models were based on the assumption that the segregated crystals can 1

in any case. and β is a positive constant. If we release assumption (2). t). This situation occurs when temperature is suciently low. If we assume that thermodynamical equilibrium between dissolved and segregated wax is instantaneously reached. t) be the total concentration of wax (either dissolved or crystallized) at each point x and time t. t)) − c(x. Let us denote by cs (T ) the saturation of the solution. where |Q| is the mass of segregated wax dissolving per unit volume and unit time at x. t) − cs (T (x. 1. t) + G(x. t). In the next section we illustrate the general framework of our analysis. t) of the dissolved wax will be given by (2) and. the concentration c(x. i. t) = c(x. t) denotes the concentration of wax crystals (often referred to as segregated wax ). Let ctot (x. the maximum concentration (mass per unit volume) of wax that can be dissolved in the given oil at temperature T . t. t) = −β [cs (T (x. t) ≥ [ctot (x. t)]H (G). (4) (5) Q(x. G ≤ 0. We will also include the case in which the thermal conductivity of the deposit is much smaller than the one of the oil. The rest of the paper is devoted 2 . The main consequence of this circumstance is that the growth of the deposit is accompanied by a quasisteady evolution of the thermal eld. we have the obvious inequality (1) G(x. t) = min(ctot (x. A reasonable assumption is c(x. t))) ctot (x.diuse (though their diusivity is much smaller than the one of the dissolved wax) and they are at any time in thermodynamical equilibrium with the solute. we want to include in our analysis the substantially dierent case in which wax crystals develop such a strong tendency to aggregate in a gel-like structure. (6) H (G) = 0. we prescribe that (1) holds with the equality sign. using for simplicity a one-dimensional geometry.e. (3) c(x. If G(x. G > 0. t))]+ . t) ≤ cs (T (x. cs (T (x. that they can be considered as immobile and not subject to diffusion (under "reasonable" concentration gradients). where [·]+ denotes the positive part. In this circumstance. we have to replace it by imposing that the transition G → c occurs according to a given kinetics. Here. t)).

x = L are impermeable to wax and oil. Thus we consider the slab x ∈ (0. L). t)]σ ˙ (t) = Dcx (σ (t). We will also assume that the densities of segregated wax. The solution is assumed to be diluted (even when saturated) in the sense that oil will be considered to be immobile. Since wax diusivity is much smaller than thermal diusivity we can assume that thermal equilibrium is instantaneously reached. Therefore. 2. T2 ). although wax is displaced by diusion (this is true in practical the case of a linear solubility curve (cs = 0). T (L. L We also assume that cs (T ) is a given increasing function and we refer -as it is often done. G(x. just to simplify notation). 0) = c∗ > cs (T2 ). although in Sect. The law of advancement of the interface x = σ (t) is given by the mass balance (12) [ρ − G(σ (t). x ∈ (0. L) whose boundaries are kept at temperatures (7) T (0. cs (T1 ) = 0. t > 0 3 . We suppose that (9) (10) (11) ctot (x. The basic idea of our model is that the dissolved wax diuses towards the cold wall and creates a solid deposit of increasing thickness σ (t). we have the linear prole (8) T0 (x) = T1 + x (T2 − T1 ). for a more general model. The statement of the problem is completed prescribing that the boundaries x = 0. if thermal conductivity is constant. i. A model problem In this paper we will study a one-dimensional problem (conning to the planar symmetry. x ∈ (0. 7 we outline the peculiarities of general cs (T ) and study in some detail a case in which cs > 0 in (T1 . x ∈ (0. 0) = cs (T0 (x)). For the necessity of being concise most of the proofs will just be sketched. t) = T1 . L). essentially for its mathematical interest. t). t) = T2 > T1 . L). and that we have initially thermodynamical equilibrium. when the temperature varies on a suciently small range (T1 . c(x. dissolved wax and oil are equal and that the latent heat involved in the transition from crystallized to dissolved wax is negligible (see the discussion in [7]).e. T2 ]. 0) = G0 (x) = c∗ − cs (T0 (x)) > 0. see [10]).to the study of various specic situations.

t) = T1 + (T2 − T1 ) (18) T (x. 4 . Moreover (13) ρ > maxG0 (x) = G0 (0). 1). t) → c∞ . As we pointed out above. We conclude this section by writing the expression for the temperature. the assumption of diluted solution allows us to neglect these terms. The wax diusivity D will be taken constant for simplicity. Remark 2. According to the already mentioned assumptions.1 (mass conservation). t) = T1 + (T2 − T1 ) x . t). Irrespectively of the assumptions on thermodynamical equilibrium (hypotheses (2) or (4)-(5)) and on the diusivity of c (including or not its possible dependence on G). t) = k0 Tx (σ (t)+. σ (t)). t > 0. σ (t) + ϑ(1 − σ (t)) x ∈ (0. L).2 (asymptotics). and c(x.such that c∞ = cs (T (σ∞ )). Remark 2. t > 0. and we impose continuity of thermal ux at the interface (16) We nd: (17) T (x. σ (t) + ϑ(1 − σ (t)) σ (t) + ϑ(x − σ (t)) . although it is likely to depend on G. kd Tx (σ (t)−. t > 0. the diusing wax that deposits on the wall induces the displacement of oil and mass conservation should take into account this additional eect which would also aect the evolution equation for c and G with the appearance of a convective term. With no loss of generality we can rescale length and take L = 1. As a matter of fact. We set (15) ϑ = kd /k0 where kd and k0 are the thermal conductivities of the deposit and of the solution (the latter being considered to be independent of c and G). the problem with D = D(G) with D (G) < 0 presents several interesting mathematical aspects. the asymptotic situation for t → ∞ is such that σ → σ∞ . ∀x ∈ (σ∞ . In principle. temperature is a linear function of x in each layer of constant conductivity. x ∈ (σ (t). which is the unique solution of (14) ρσ∞ + cs (T (σ∞ ))(L − σ∞ ) = c∗ L. Note that the same situation is reached irrespectively of the fact that diffusion of G is included or not in the model and for any strictly monotone cs (T ).where ρ is the density of the solid layer which is supposed to incorporate all the segregated wax present in the region crossed by the interface.

.3). 4). t ∈ (0. c(x. 5 . In most practical cases (20) So. t) is expressed in terms of σ (t) by (18). there will be a time interval (0.thermodynamical equilibrium with full saturation (Sect. σ (t) + ϑ(1 − σ (t)) σ (t) + ϑ(1 − σ (t)) Dissolution of wax crystals occurs except at x = 0. t) T1 . During this period. In the next two sections we will study in some detail the following cases: . This means that the thermal eect of the deposit can be taken into account. 3. but also in the bulk because of the increase of saturation concentration. t1 ). Case of thermodynamical equilibrium: full saturation Assume (2) holds and let the initial situation be described by (10). if ϑ = O(1). T (σ (t). where T (x. t) = T1 + (T2 − T1 ) σ (t) . t1 ) such that (22) G(x. still desregarding the corresponding modication on the geometry of the problem. σ (t) + ϑ(1 − σ (t)) t > 0. Then.thermodynamical equilibrium with partial saturation (Sect. dissolved wax migrates towards the cold wall with a ow rate equal to (23) Therefore (24) (25) [ρ − G(σ (t). t) > 0. x ∈ (σ (t). Dcs T2 − T1 γ ≡ .3. where segregated wax has to dissolve to replace the migrating dissolved wax. (21) σ (t) ≤ σ∞ β 1.In particular (19) T (σ (t). t ∈ (0. σ (t) + ϑ(1 − σ (t)) σ (0) = 0. meaning that even a thin deposit can have a relevant inuence on the thermal prole. (11). We start by considering the case of non-vanishing diusivity DG of wax crystals. t)). t)]σ ˙ (t) = γ . 1). but this is no longer true if ϑ 1. Remark 2. t1 ). t) = cs (T (x.

To this we can associate a pair (σ. owing to thermodynamical equilibrium. t). G has to be found by solving the problem   DG Gxx − Gt = cs Tt =    ϑ(1 − ϑ)(1 − x)    . t ∈ (0. Moreover ϑ ϑ Gn < G0 (0) at all internal points. The procedure can be started taking for instance Γ1 (t) = G0 (0). t ∈ (0. t). Let us sketch the procedure. G) which turns out to be a solution of our free boundary problem. ρ − Γn σn (0) = 0. Compute σn (t) by integrating (27) [σn + ϑ(1 − σn )]σ ˙n = γ . 1). thus identifying σn (t) as the positive root of the equation (28) 1 2 (1 − ϑ)σn + ϑσn = γ 2 t 0 dτ . σ (t) + ϑ(1 − σ (t))         Gx (σ (t). x ∈ (σ (t). t) by solving problem (26) with σ = σn and dening Γn+1 (t) = Gn (σn (t). G0 (0)) for t ∈ (0. by means of successive approximations. t) < − . t1 ). t) = − . Thus we possess enough information to say that there exists a subsequence {Γn } converging uniformly to a Hölder continuous limit Γ(t) taking values in (0. It is also possible to obtain a uniform estimate of the rst vanishing time of Gn (x. 0 s 0 (26) which is coupled with (24). using the fact that the sink term in the dierential equation 1−ϑ γ is dominated by cs σ ˙0 (1 − x) and that DG Gnx (1. t) = 0. t ∈ (0.g. = cs σ ˙ (t)   [σ (t) + ϑ(1 − σ (t))]2       γ DG Gx (1. Gn ) along the same ϑ(ρ − G (0)) 6 . γ Next we compute Gn (x. Let Γn (t) be a Hölder continuous function such that Γn (0) = G0 (1) and 0 ≤ Γ(t) ≤ G(0). Problem (24)-(26) is a free boundary problem of a non-standard form for the heat equation. positive by denition and γ ≡σ ˙ 0. Its solution can be sought e.Hence. t1 ). x ∈ (0. as it can be shown by taking the limit of (σn . t). t). a-priori bounded by σ ˙n < ϑ(ρ − G0 (0)) 0 We x a time interval (0. 1). such that t < . (25). t1 ). so that σn (t) < 1. It is easy to obtain an a-priori uniform estimate of Γn+1 in a Hölder space. 0) = G (x) = c∗ − c (T (x)).        G(x. ρ − Γn (τ ) which possesses a Hölder continuous derivative.

a) We start by considering the case DG = 0. implying δσ ≡ 0. 1). Case of thermodynamical equilibrium: partial saturation At time t1 . (29) leads to a Gronwall inequality for δσ t with zero free term. This stage ends at time t2 when r(t2 ) = σ (t2 ). 1) where c < cs and G = 0. t ∈ (0. σ (t) ≤ x < 1. t) = 0. G1 ).   [σ (t) + ϑ(1 − σ (t))]2      (32) G(r(t). if DG = 0). r(t)) dissolved wax diuses with a ow rate still given by (23) and the dierential equation for the function σ (t) representing the thickness of the deposit is still (24) with initial condition (31) σ (t1 ) = σ1 . G2 ) be two solutions. t ∈ (t1 . Let (σ1 . if t1 > 0 (i. t2 ). Gnx are also equibounded).       Gx (σ (t). τ ∈(0. Now we have in the saturated zone:  DG Gxx − Gt = cs Tt =     ϑ(1 − ϑ)(1 − x)   = cs σ ˙ (t) .from (24). where σ1 is calculated from the previous stage.e. x ∈ (σ (t). (25) we deduce an integral inequality of the type (29) t |δσ (t)| ≤ 0 {A|δσ (τ )| + B |δG(τ )|}dτ. Thus the recursive scheme described above is convergent. t1 ). separating the zone x ∈ (σ (t). t) − G2 (σ2 (t). B . t) = 0. r(t)) where c = cs and G > 0 from the zone x ∈ (r(t). a new free boundary x = r(t) appears. δG = G1 (σ1 (t). (σ2 . 1). 0 ≤ t ≤ t1 . Setting δσ = σ1 − σ2 . Since |δG(t)| can be estimated in terms of sup |δσ (τ )| ≡ δσ t (the proof is lengthy and is omitted).        G(x. t1 is also zero and a part of the domain will be unsaturated from the very beginning. 4.subsequence (note that σ ˙ n .t) The time t1 is necessarily bounded. t ∈ (t1 . 7 . for some easily computable positive constants A. This procedure can be iterated up to the time t1 > 0 such that (30) G(x. t). x ∈ (σ (t1 ). t1 ) = G1 (x). t) > 0. Note that if the diusivity of segregated wax vanishes. For x ∈ (σ (t). Uniqueness is not dicult to be proved. t2 ).

t). 8 .      cx (1. 1)) has been found in the previous stage (remember (30)). x ∈ (r(t).t. c(x. t) = Dcx (r(t)+. t ∈ (t1 . On the other hand. Problem (24). t2 ). then σ ˜ ∈ Σ and that the mapping σ → σ ˜ is compact w. t) solves the following problem in the unsaturated zone x ∈ (r(t). t2 ). t). this time using the formula t σ ˜ (t) = γ 0 [σ (τ ) + ϑ(1 − σ (τ ))]−1 [ρ − u(σ (τ ). t2 ). t) − cs (x. t2 ) with condition (35) r(t1 ) = 1. t2 ) satisfying (31). t) = ctot (x. noting that the diusivity for u has a jump for u = 0.r. t) = cs (r(t). For every given σ in a ball of C1 (t1 . t ∈ (t1 . 1). (31)-(35) is a problem with two free boundaries of explicit/implicit type. (33) and on the new free boundary we have the following condition expressing mass balance (34) γ + DG Gx (r(t)−. the sup-norm. Then a compactness argument can be used to nd σ (t) and conclude the proof of the existence theorem: taken σ in a convenient convex set Σ of Lipschitz continuous increasing functions. leaving a more precise analysis of the well-posedness of the problem to a future paper. and show that if Σ is selected properly. (25).where G1 ≥ 0 (G1 > 0 in [σ (t). we conne ourselves to sketch a possible way of proving existence.      c(r(t). problem (32)(35) can be seen as a diraction problem for the function u(x. t) σ (t) + ϑ(1 − σ (t)) for t ∈ (t1 . τ )]−1 dτ. t ∈ (t1 . 1):  Dcxx − ct = 0. This class of problems has been investigated in the context of fast chemical reactions (see [3]. For the necessity of being concise. t) = 0. [4]) and those techniques can be adapted to t the present situation. solve the free boundary problem for u and compute σ ˜ by means of (24).

In order to follow this procedure it is important to be sure that the region r(t) < x < 1 remains desaturated. In the saturated layer x ∈ (σ (t). Since ∂cs 1−x = cs (T2 − T1 )(1 − ϑ)ϑ σ ˙ > 0. σ + ϑ(1 − σ ) Note that in this scheme G is discontinuous across the free boundary. σ + ϑ(1 − σ ) for x ∈ (σ (t). t) = G0 (r) − cs (T2 − T1 )(1 − ϑ) σ (1 − r) > 0. Up to time t2 when the whole solution becomes unsaturated we have that the deposit grows according to equation (24). t)) − cs (T0 (x))] = σ = G0 (x) − cs (T2 − T1 )(1 − ϑ) (1 − x). (38) is a problem of generalized Stefan type for which we have proved a well-posedness result in [13]. t) solves (33). In the unsaturated region x ∈ (r(t). To be specic. r(t)) we have (37) G(x. Again. t) < 0). knowing the following initial data: (39) σ (t2 ) = σ2 9 . problem (33). Fully unsaturated slab Of course. σ (t) + ϑ(1 − σ (t)) for t ∈ (0. t2 ). the mathematical model becomes insensitive of whatever assumption we have made on the behaviour of the segregated phase. the completion of the proof is achieved via a compactness argument. t) = 0. so that the initial conditions for the two free boundaries are (36) σ (0) = 0. 1) c(x. t) = G0 (x) − [cs (T (x. We have already noted that t1 = 0. t2 ). ux (1. u(r(t). just applying the maximum principle (ut − Duxx < 0. r(t)) and t ∈ (0. once we know that the solution has a wax concentration under saturation and G = 0. For a given increasing σ ∈ C 1 .b) The case of DG = 0 is simpler. t) = Dcx (r(t)+. and the condition on x = r(t) is the following (38) γ −r ˙ (t)G(r(t)−. we assume we have gone through the previous stages and we start from time t2 . ∂t [σ + ϑ(1 − σ )]2 it is easy to check that the function u = c − cs is negative in the domain considered. 5. r(0) = 1. t). with G(r(t)−.

At this point global existence and uniqueness for (43)-(48) is a well known result (see [13]). cx (1. 6. t > t2 . (no dissolution of the deposit) so that σ ˙ ≥ 0 by denition. we have a basic dierence. Indeed. we can prove that we have cx (x. t) to be an increasing function of time at any point x: in the present case immediate desaturation takes place. 1). t)) can be proved as follows. We have the following diusion problem for c(x. t ). showing that u = c − cs < 0.(40) (41) where (42) G(x. 10 x ∈ (σ (t). Then we apply the same argument we have used at the end of the previous section. x ∈ (σ2 . that is a consequence of the fact that (excluding the trivial case ϑ = 1) the deposition causes cs (x. c(σ (t). t) with one free boundary: (43) (44) (45) (46) (47) (48) Dcxx − ct = 0. x ∈ (σ (t). 1). because to keep c = cs one should have Q < 0. Now. Moreover. c(x. Hence we may say that σ ˙ (t) > 0 for all t > t2 . x ∈ (σ2 . c(x. as an easy consequence of the maximum principle. t2 ) = 0. x ∈ (σ2 . t) > 0 in (t . t2 ) = c2 (x). t > 0. 1). . implying that cx (σ (t). thus contradicting the assumption σ ˙ = 0. t)]H (G). We start by considering the case DG = 0 and we study the following problem (49) Gt (x. t) = 0. 1). the solution has to be unsaturated throughout. if for some time interval (t . σ (t2 ) = σ2 ∈ (0. 1). and cs (T (x)) is found from (19). 1). x ∈ (σ2 . t > t2 . since we are assuming cs = 0. The fact that c(x. t). t)]+ . Case of nonequilibrium In this section we will assume (4)-(6) and we suppose that the same initial conditions (9)-(11) are prescribed. t)). t > t2 . t) > 0 as long as cx (σ. t) = cs (T (σ (t). First replace (47) by ρσ ˙ = −D[cx (σ (t). thus contradicting (5). t)) − c(x. 1). t > t2 . we can easily see that c takes its minimum for x = σ . ρσ ˙ (t) = −Dcx (σ (t). t) remains below cs (T (x. t ) we have σ ˙ = 0. c2 (x) < cs (T (x)). t2 ) = c2 (x). t) = −β [cs (T (x. t) > 0. With respect to previous sections.

t)]H (G). t > 0. t) is Hölder continuous and that its Hölder norm depends on the Lipschitz constant of σ . Hence. Problems of this kind are non-standard. We prescribe σ (t) in a Lipschitz class Σ of monotone functions so that T (x. the r. σ (0) = 0. t)). t) = cs (T (σ (t). t) = 0. a unique xed point is found. t) = 0. It is not dicult to show that this mapping is contractive in Σ with respect to the sup-norm in a suciently small time interval. through H (G). providing a solution to our problem.(56). t > 0. t)) − c(x. 0) = G0 (x) = c∗ − cs (T0 (x)). The resulting function c(x. t)) − c(x. From (49) we nd a formal expression for G (57) t G(x. x ∈ (σ (t). So. provided that suitable conditions are satised. 11 . and no ux conditions are to be assigned on the boundaries Gx (1.h. t). Finally. 0) = cs (T0 (x)). 1). which is actually C 1 . t) = 0.(50) (52) (53) (54) (55) (56) G(x. c(x. t > 0. but the existence of one unique classical solution can be proved by the methods of [14] and we note that the argument works also if we allow D to depend on G. c(σ (t). t) = G0 (x) − β 0 [cs (T (x. with the free boundary conditions [ρ − G(σ (t). t) is given by (18) and thus (57) expresses function G as a functional of c(x. 1). of (51) is also a functional of c. Extension to arbitrary time intervals is then obtained by means of standard arguments. Now (49) is replaced by (58) (59) DG Gxx − Gt = β [cs (T (x. Gx (σ (t). if σ (tG ) < x. cx (1. and is σ −1 (x) otherwise. t). (51) Dcxx − ct = −β [cs (T (x. t)]H (G). t) is easily shown to be such that cx (σ (t). Then. x ∈ (0. For the sake of brevity we just sketch the procedure to prove the wellposedness of problem (49)-(56). We use again a xed point argument. holding for x > σ (t) and t ≤ tG (x) which is the unknown time of extinction of G at the location x. x ∈ (0. a mapping σ → σ ˜ can be dened using (55). τ )]dτ. t)]σ ˙ (t) = Dcx (σ (t). let us just see how the arguments change when DG = 0. 1). τ )) − c(x. t > 0.s.

Nevertheless we will investigate it briey because of its mathematical interest. As a general remark. cs (T ) ≥ 0 in (T1 . t) = T0 (x) = T1 + (T2 − T1 )x and this assumption is justied by the absence of solid deposit. remain saturated provided temperature does not change in time. bound to become extinct in a nite time. Thus the general problem is considerably complicated. However this task is not particularly dicult. T2 ). we assume (61) cs (T1 ) = 0. also because the corresponding asymptotic limit of G(x) would be a point mass located in x = 0. We have Gt = βH (G)[cs (T0 (x)) − c(x. we note that if we start from initial conditions (10)-(11) and we have cs (T ) < 0 in some interval (x1 .h. irrespectively of the sign of cs (T2 ). It is clear that the situation described by (60). t) − cs (T0 (x)))Dcs (T0 )2 12 . Once this is accomplished. and points where cs < 0 may become desaturated before they are reached by the desaturation front coming from the warm wall. Indeed. For the sake of brevity we conne to the case DG = 0 and to the absence of thermodynamical equilibrium. then a zone of desaturation will immediately appear in case (4)-(5) (even if T remains constant in time). saturated layers. Desaturation occurs immediately at x = 1. Assuming thermodynamical equilibrium leads to a very similar scenario. On the contrary. x2 ). for each given σ in the Lipschitz class. the two parabolic problems for G and for c are strictly coupled and hence the existence of c and G has to be proved beforehand. however too complicated to be treated here. It can be seen that if D = D(G) with D ≤ 0 the problem presents an interesting mathematical structure. the proof follows the same pattern as in the previous case.Of course. (61) is rather articial from the physical point of view.s. Mass transfer with no deposition In this section we give some ideas on the behaviour of the system when linearity of cs (T ) is no longer assumed. t)] + (62) + H (c(x. unless temperature is time independent. Intervals in which cs = 0 conning on the r. if DG = 0. 7. with an interval where cs > 0. Once more we take D independent of G. In the latter case saturation persists locally for some time (with the exception of x = 1 if DG = 0). We will consider in more detail a situation in which we assume time-independent temperature (60) T (x. are present in correspondence of intervals in which cs > 0.

t ∈ (0. x ∈ (s(t). x ∈ (0. t s(0) = 1. Following the technique of [8] it can be shown that (64)-(68) can be reduced to a Stefan-type problem. t) = cs (T0 (s(t)))T0 . t c(s(t). t ˜). 0) = G0 (x). G(s(t)−. ˜). Now we have to solve a free boundary problem in the unsaturated layer x ∈ (s(t). To this system one has to add the information that the limit of w in the corner point (0. although of a singular kind. 1). t) = cs (T0 (s(t))). x ∈ (0. wx (s(t). t ∈ (0. ˜). t ∈ (0. (67). t ∈ (0.where (63) H (z ) = 1. t) = 0. 1). t ∈ (0. in the unsaturated zone G is decreasing according to the kinetics Gt (x. cx (1. t) = 0. z ≥ 0. ˜). 0. z < 0. ˜). (66). t ˜). n 13 . w(s(t). t ∈ (0. x ∈ (s(t). t) > 0 can be evaluated from (62) (69) (70) (71) Gt (x. s(t)). t ˜). t) = −β (cs − c)H (G) and (71) has to be integrated starting from x = s(t). t ˜). 1). Condition (68) expresses the continuity of solute ux and follows from the continuity of G at the interface. t ˜). t ∈ (0. performing formal dierentiations on (64). 1): (64) (65) (66) (67) (68) ct − Dcxx = H (G)β (cs − c). t ∈ (0. Indeed. t ∈ (0. t) = cs (T0 (s(t)))T02 s ˙ (t). while. Since c ≤ cs either the positive or the negative term is active. t G(x. (68) one arrives at the scheme (72) (73) (74) (75) wt − Dwxx = −w. The solution can be obtained as the limit as n → ∞ of regularized problems in which the initial condition is (76) s(n) (0) = 1 − 1 . t wx (1. 1) along directions marked by a unit vector pointing inside the domain is −∞. For instance. cx (s(t). t) = Dcs (T0 (x))T02 . t) = 0. as long as G > 0 in the unsaturated zone setting ct = w.

Predicting deposit formation in the pipelining of waxy crude oils. Boundary value multidimensional problems in fast chemical reactions. Cannon. Primicerio. M. Arch. Magenes ed. Fasano. M. Fusi. A. to appear on Recent Trends in Mathematical Physics. S. Sem. [6] S. Free Boundary Problems. L. Correra. 441-483. 14 . Fasano. J. 1 n 1 < x < 1. C. Correra. Correra. Rat. Rosso. A.R. L. Primicerio. Studies of wax deposition in the trans Alaska pipeline. [2] E. Azevedo. T. Striegler. Mathematical models for waxy crude oils. II (1979). M. On the movement of a chemical reaction interface. Journal of Petroleum Technology (1981). A. Indiana Math. F. J. [8] A. (E. Fasano. Mech. Teixeira. Wax diusivity under given thermal gradient: mathematical model. Free boundary problems for parabolic equations with Cauchy data on the free boundary. L. Fusi. submitted. 1-13. fn (s(n) (0)) = 0 n2 1 1 1 while in (1 − . submitted. Cannon. A.H. 1 − + 2 ) it is any decreasing function such that f (n) is n n n smooth in [s(n) (0). Burger. 237-247. Fasano. Petroleum Science and Technology 21 (2003). Fasano. Perkins. Fusi. [9] A.A. Anal.(77) w(n) (x. A critical review of the modelling of wax deposition mechanism.K. (2004). 429-454.D. 20 (1970). Modelling wax diusion in crude oils: the cold nger device. 53(1973). [5] S.) vol. Proc. Fusi. Primicerio. 1075-1086. [3] J. Fasano. 393-408. Hill. where f (n) (x) = −ncs (T0 s(n) (0))T02 for 1 − + The feasibility of this procedure can be proved using a technique similar to the one employed in [12] for obtaining existence for a singular diusionconsumption problem. References [1] L. Fasano. D. Merino-Garcia.R. Correra. 0) = f (n) (x). submitted. [4] J. Meccanica 39.D. A.F. [7] S. L. [10] A.M. Heat and mass transfer in non-isothermal partially saturated solutions. 1].

443-455. J. [13] A. SPE (2000). Primicerio. Giorni.S. 67/A 50134 Firenze M. [16] R. I. J. AICh.J. A. Dini" Viale Morgagni. S. 91-108. D. Primicerio Dipartimento di Matematica "U. 50 (2004). Math.V. Saracoglu. Comments and analogies for correlated heat and mass transfer in turbulence ux. 694-723. Dini" Viale Morgagni. Shuler. Venkatesan. Measurements of wax deposition in paran solutions. [17] Chien-Hou Wu. General free-boundary for the heat equation. 67/A 50134 Firenze 15 . Appl. Primicerio. M. Math. Kang-Shi Wang.J. 14 (2004). M. H.M. P. 38 (1981). Proc. O.[11] A. Cheung. M. 2002. Meeting Wax Thermodynamic and Deposition. Fasano. E. Appl. Sci. [15] M.L. Temperature driven mass transport in concentrated saturated solutions. Yongehun Tang. Fasano. Primicerio. New Orleans. Fasano Dipartimento di Matematica "U. 1623-1626. Appl. Creek. 439-460. Progress in Nonlinear Dierential Equations and their Applications 61 (2005). Quart. New results on some classical free boundary problems. Fogler. 1-7. R. Fasano. Math. On a one-dimensional problem related to bioremediation of polluted soils. [12] A. [14] A. 57 (1977).Carlson. Adv. Anal.Kok. Mahematical modelling of wax deposition in crude oil pipeline system. Fasano.