betaFeedback

WorksPhysics and Astronomy

Mechanism of deflagration-to-detonation transition in gas

Abstract

The deflagration-to-detonation transition on the tip of an elongated flame in a tube is shown to be related to a dynamical saddle-node bifurcation of the inner flame structure leading to a runaway of the pressure in finite time. The comparison with the experiments shows a good agreement.

Introduction and orders of magnitude

The deflagration-to-detonation transition (DDT) is a quasi-instantaneous transition (few microseconds) between two opposite regimes of combustion wave. The phenomenon was observed long ago [1]. However, after more than a century of experimental works and decades of numerical studies reported in an extensive literature, DDT is not yet understood [2][3]. This abrupt transition is explained here by the one-dimensional dynamics of the reactive flow of a self-accelerating flame. Coupling reaction-diffusion and compressibility is a challenging problem. The physical mechanism of the DDT is presented here in a synthetic way for physics-oriented readers, skipping the technical details of the analytical methods that will be published elsewhere [4]. Gaseous detonations are supersonic combustion waves involving a pressure rise Δp/p\Delta p/p ranging from 20 to 50 and propagating with a velocity between 2 000 m/s and 3 500 m/s under normal conditions. Laminar flames are quasi-isobaric reaction-diffusion waves characterized by a markedly subsonic velocity and a negligible pressure drop Δp/p≪1\Delta p/p \ll1. In gas, the laminar flame velocity relative to the unburned gas ULU_L ranges from few tens cm/s in the least energetic mixtures to 9 m/s in the most energetic ones. The rate of heat release is governed by inelastic collisions of molecules associated with an activation energy E\mathcal{E} larger than the thermal agitation kBTk_B T. The flame temperature TbT_b and the reduced activation energy β≡E/kBTb\beta\equiv\mathcal{E}/k_B T_b are in the ranges 1800 to 3500 K and 2 to 8 respectively; the larger TbT_b, the smaller β\beta. According to the kinetic theory of gas, the reaction rate 1/τr1/\tau_r is expressed in terms of the collision frequency 1/τcoll1/\tau_{\mathrm{coll}} in the form of an Arrhenius law τcoll/τr∝βe−β\tau_{\mathrm{coll}}/\tau_r \propto\beta e^{-\beta}. However, due to a complex chemical kinetics network [5], the exothermic reaction rate decreases drastically below a crossover temperature Tc∈[850−1200 K]T_c \in[850 - 1200\ \mathrm{K}] and combustion cannot develop for T<TcT < T_c. The reaction rate 1/τr1/\tau_r at TbT_b is typically 1/τr≈106 s−11/\tau_r \approx10^6\ \mathrm{s}^{-1}. Detonations consist of a strong inert shock of few mean free paths thick (≈aτcoll\approx a\tau_{\mathrm{coll}}, aa is the sound speed) followed by a macroscopic zone of reaction of thickness ≈aτr\approx a\tau_r through which the diffusive transports are negligible τr≫τcoll\tau_r \gg\tau_{\mathrm{coll}}. Expressed in terms of the molecular diffusivity D≈a2τcollD \approx a^2\tau_{\mathrm{coll}}, according to the Zeldovich-Frank-Kamenetskii (ZFK) analysis [6] in the limit β≫1\beta\gg1, the laminar flame velocity relative to the burned gas is

Ub∝β−3/2Db/τrbU_b \propto\beta^{-3/2}\sqrt{D_b/\tau_{rb}} where the subscript b refers to the burned gas, see [3] for a didactic presentation. The flame thickness and the transit time of a fluid particle across the flame are respectively d=Db/Ub≈2−4×10−1 mmd = D_b/U_b \approx2 - 4 \times10^{-1}\,\mathrm{mm} and tb=d/Ub∝β3/2≈10−4 st_b = d/U_b \propto\beta^{3/2} \approx10^{-4}\,\mathrm{s} under normal conditions. The flame Mach number ε≡Ub/ab∝e−β/2/β\varepsilon\equiv U_b/a_b \propto e^{-\beta/2}/\beta is about 0.03–0.1 in very energetic mixtures (stoichiometric H2/O2H_2/O_2 or C2H4/O2C_2H_4/O_2 mixtures) and 10−310^{-3} in hydrocarbon-air mixtures and Δp/p≈ε2\Delta p/p \approx\varepsilon^2.

Experimental data and objective of the analysis

We limit our attention to DDT in long smooth-walled tubes filled with stoichiometric H2/O2H_2/O_2 or C2H4/O2C_2H_4/O_2 mixtures. Just before transition in micro-scale tubes (radius ≈\approx few mms) the front of the laminar flame is elongated and the flow is laminar [7][8], see Figure 1. The DDT is more

Figure 1. Sketch of an elongated flame in a smooth-walled tube

difficult to analyze in larger tubes because the flame front can be turbulent [9], the problem becoming stochastic in nature. Conclusions can nevertheless be drawn since the salient features of the abrupt transition in moderately large tubes (50×50 mm250 \times50\ \mathrm{mm}^2) are not so different, especially as the flow stays laminar ahead of the wrinkled flame front [10][11]. The salient features can be summarized as follows: spontaneous onset of detonation occurs locally outside the boundary layers in a small explosion center on a laminar flamelet near the leading edge of the flame brush. In some turbulent flames [9] the explosion center is in the boundary layer where the DDT could be due to a Zeldovich mechanism [12] reinforced by the compressible waves [3]. This case is not considered here where the attention is focused on an inherent DDT mechanism of laminar flames accelerated by a self-generated convection flow. In a tube of diameter between 5 and 50 mm in which the reactive mixture is ignited at the center of the closed end, the surface area of the laminar flame first increases exponentially, goes through a transient tulip-shaped form [13] and then increases slowly before the abrupt transition. Correlatively, a longitudinal convection flow is generated by the increase of surface-area of the flame front, putting the overall flame in motion at velocity UP>UbU_P > U_b. Because of the laminar nature of the flamelets, the longitudinal flow of unburned gas relative to a flamelet orthogonal to the tube axis is equal to the laminar flame velocity UL≪aU_L \ll a. The detonation onset occurs spontaneously when appropriate local conditions are attained in the unburned gas adjacent to the leading edge of the self-accelerating flame [2] [10]. This occurs when the overall speed of the flame front reaches a critical value UP∗U_P^* comparable to the local sound speed [8][10][11]. The acceleration-induced compression waves warm up the unburned gas by compressional heating but, as noticed a long time ago, the temperature stays well below the crossover temperature TcT_c so that the reactive mixture just ahead of the tip of the elongated flame remains chemically frozen leaving unexplained the DDT. We will not discuss the origin of the growth of surface area Σ(t)\Sigma(t) responsible for accelerating the flame front. The flame elongation in a tube with stick walls is governed by the quasi-isobaric expansion of the gas across the flame front while the abrupt transition under consideration is a consequence of the compression waves in the external flow. In the following, a small growth rate of the total surface area ϵ≡(tb/Σ)dΣ/dt≪1\epsilon\equiv(t_b/\Sigma)\mathrm{d}\Sigma/\mathrm{d}t \ll1 is a given parameter (not to be confused with ε\varepsilon). The analysis of the accelerating flame on the tip of the elongated front sketched in fig.1 is performed in the limit of small flame Mach number ε≡Ub/ab≪ϵ≪1\varepsilon\equiv U_b/a_b \ll\epsilon\ll1 using the ZFK flame model. The objective is twofold. Firstly, identify the critical condition from which the slightest increase in elongation, no matter how small, produces the compressibility-induced DDT onset. This part is an extension of a recent work in which the self-accelerating flame was treated as a planar discontinuity [14], as in the pioneering analysis identifying a turning points of the self-similar solutions of wrinkled flames [15]. The second objective is to show that the DDT corresponds to a finite-time singularity of the inner structure of an accelerating quasi-planar flame orthogonal to the direction of propagation and put in motion by the convective flow. The analysis takes advantage of three simplifications: firstly, a local quasi-planar geometry on the tip (back-flow model [16]), secondly, a negligible pressure drop across the flame Δp/p≈ε2\Delta p/p \approx\varepsilon^2 (smaller than the pressure variation of order ϵ\epsilon across the compression waves in the external flow), and finally a scale separation, the laminar flame thickness dd being much smaller than both the tube radius, d/r<10−1d/r < 10^{-1} and the length ll involved in the compression waves, d/l<ϵd/l < \epsilon.

Quasi-steady self-similar solution

To begin, consider the acceleration-induced compression waves in the unburned gas, neglecting unsteadiness elsewhere. Due to a quasi-isobaric gas expansion, the longitudinal flow increases across the flamelet, from ubu_b on the burned gas side to uuu_u on the unburned gas side, uu>ubu_u > u_b. If the inner structure is in steady state, mass conservation yields ub+Ub=uu+UL=UPu_b + U_b = u_u + U_L = U_P where UPU_P is the nearly sonic speed of the convected flame front while UbU_b and ULU_L are the laminar flame velocities (markedly subsonic) relative respectively to the burned and unburned gas, ρbUb=ρuUL\rho_b U_b = \rho_u U_L (mass conservation) and ρb/ρu=Tu/Tb∈[0.13−0.08]\rho_b/\rho_u = T_u/T_b \in[0.13 - 0.08] (quasi-isobaric condition), subscripts uu and bb denoting respectively the unburned gas ahead of the flame and the burned gas behind the flame. If the inner flame structure is the same everywhere on the elongated front (this is not an essential assumption as discussed few lines below (1)), the mass of fresh gas burned by unit-time yields ρuULΣ=ρbUbΣ\rho_u U_L \Sigma= \rho_b U_b \Sigma. Assuming that the flow is at rest between the closed end of the tube and the straight-section delimiting the burned gas enclosed in the finger flame, the mass rate of burned gas is ρb(dH/dt)Φ\rho_b(\mathrm{d}H/\mathrm{d}t)\Phi, Φ\Phi and HH denoting respectively the cross section of the tube and the tube length behind the finger flame (dH/dt≈UP\mathrm{d}H/\mathrm{d}t \approx U_P), see fig.1. This yields the classical result UP≈(Σ/Φ)UbU_P \approx(\Sigma/\Phi)U_b. The relations ub=UP−Ubu_b = U_P - U_b and uu=ub+Ub−ULu_u = u_b + U_b - U_L then lead to the flow velocities ubu_b and uuu_u on the flamelet in terms of UbU_b and the flame elongation S≡Σ/Φ−1S \equiv\Sigma/\Phi- 1 (S≈2L/RS \approx2L/R for an elongated flame of length LL in a cylindrical geometry),

UP=(S+1)Ub,ub=SUb,uu=(S+1−θ0)Ub(1)U_P = (S+1)U_b,\qquad u_b = SU_b,\qquad u_u = (S+1-\theta_0)U_b \tag*{(1)}

where 1−θ01-\theta_0 is the reduced heat release, θ0≡Tu0/Tb0<1\theta_0 \equiv T_{u0}/T_{b0} < 1, the subscript 0 denoting the initial condition. The greater the flame surface area Σ\Sigma, the larger the longitudinal flows ubu_b and uuu_u on the flamelet. When approaching the abrupt transition, Eqs. (1) are meaningful independently of the precise definition of SS (or the length LL of the elongated flame). The crucial points are ub∝Ubu_b \propto U_b, uu∝Ubu_u \propto U_b and a rate of change of UbU_b which diverges at the transition, no matter how small dS/dt\mathrm{d}S/\mathrm{d}t, see (9) below. Therefore, the rate of change of SS does not matter. Consider an elongated flame accelerating from an initial state (subscript 0) constituted by a self-similar solution with the lead shock at infinity; t≤0t \leq0: uniform flow of unburned gas, uu=uu0u_u = u_{u0}, p=pu0p = p_{u0}. The flame acting as a semi-transparent piston, the downstream-running compression waves ahead of the flame are isentropic as long as the variation of ULU_L is not larger than its initial value because, the acceleration-induced shocks being weak for ϵ≪1\epsilon\ll1, δuu/au0=O(ϵ)\delta u_u/a_{u0} = O(\epsilon), δpu=ρu0au0δuu\delta p_u = \rho_{u0}a_{u0}\delta u_u, δpu/pu0=γδuu/au0∼ϵ\delta p_u/p_{u0} = \gamma\delta u_u/a_{u0} \sim\epsilon, the entropy jump of ϵ2\epsilon^2-order is negligible. Denoting x=XP(t)x = X_P(t) the position of the flame dXP/dt=UP(t)\mathrm{d}X_P/\mathrm{d}t = U_P(t) and uu(t)u_u(t) the flow just ahead of the flame, the compressible flow of unburned gas is, ε≪1\varepsilon\ll1:

{0<x−XP(t)<au0t:u=uu(t−x−XP(t)au0),au0t<x−XP(t):u=uu0.(2)\begin{cases} 0 < x-X_P(t) < a_{u0}t : u=u_u\left(t-\dfrac{x-X_P(t)}{a_{u0}}\right),\\ a_{u0}t < x-X_P(t) : u=u_{u0}. \tag*{(2)} \end{cases}

The instantaneous pressure and temperature of unburned gas on the flame δpu(t)\delta p_u(t) and δTu(t)\delta T_u(t) are expressed in terms of uu(t)u_u(t) by the acoustic relations δpu/pu0=γδuu/au0\delta p_u/p_{u0}=\gamma\delta u_u/a_{u0} and δTu/Tu0=(γ−1)δuu/au0\delta T_u/T_{u0}=(\gamma-1)\delta u_u/a_{u0}. Using Ub0/au0=ε/θ0U_{b0}/a_{u0}=\varepsilon/\sqrt{\theta_0} where ab0/au0=1/θ0a_{b0}/a_{u0}=1/\sqrt{\theta_0}, one gets

δpupu0=γθ0εδuu(t)Ub0,δTuTb0=(γ−1)θ0εδuu(t)Ub0.(3)\frac{\delta p_u}{p_{u0}}=\frac{\gamma}{\sqrt{\theta_0}}\varepsilon\frac{\delta u_u(t)}{U_{b0}},\qquad \frac{\delta T_u}{T_{b0}}=(\gamma-1)\sqrt{\theta_0}\varepsilon\frac{\delta u_u(t)}{U_{b0}}. \tag*{(3)}

For δuu(t)/Ub0\delta u_u(t)/U_{b0} not larger than unity, the pressure variation on the flame is larger than the pressure drop across the flame by a factor 1/ε1/\varepsilon. Neglecting ε\varepsilon-terms, the flow velocity in the steady inner structure of the flame is the same as in steady state uu−ub=Ub−UL≈(1−θ0)Ubu_u-u_b=U_b-U_L\approx(1-\theta_0)U_b. Then, according to (1) ub=SUbu_b=SU_b, uu0=(S0+1−θ0)Ub0u_{u0}=(S_0+1-\theta_0)U_{b0},

δuuUb0+(S0+1−θ0)=(S+1−θ0)UbUb0.(4)\frac{\delta u_u}{U_{b0}}+(S_0+1-\theta_0)=(S+1-\theta_0)\frac{U_b}{U_{b0}}. \tag*{(4)}

Under the steady state approximation of the inner structure of the flame, the flame temperature is simply shifted by the temperature change of the unburned gas just ahead of the flame δTb=δTu\delta T_b=\delta T_u. An equation for δTb/Tb0\delta T_b/T_{b0} is obtained by introducing the last equation (3) into (4), using the expression Ub/Ub0U_b/U_{b0} in terms of δTb/Tb0\delta T_b/T_{b0} given by the flame theory. The ZFK result β≫1\beta\gg1: Ub/Ub0≈e(β/2)δTb/Tb0U_b/U_{b0}\approx e^{(\beta/2)\delta T_b/T_{b0}}, δTb/Tb0=O(1/β)\delta T_b/T_{b0}=O(1/\beta), yields the equation previously obtained with a flame treated as a discontinuity [14]. Introducing the reduced flame temperature ϑ≥0\vartheta\geq0 (and/or the flow just ahead the flame uu/Ub0u_u/U_{b0}) and the reduced elongation σ≥1\sigma\geq1

ϑ≡βδTb2Tb0=bδuuUb0,σ≡S+1−θ0S0+1−θ0,(5)\vartheta\equiv\frac{\beta\delta T_b}{2T_{b0}}=b\frac{\delta u_u}{U_{b0}},\qquad \sigma\equiv\frac{S+1-\theta_0}{S_0+1-\theta_0}, \tag*{(5)}

the function ϑ\vartheta versus σ\sigma is given by the roots of

σeϑ−ϑs0−1=0,(6)\sigma e^\vartheta-\frac{\vartheta}{s_0}-1=0, \tag*{(6)}
s0≡b(S0+1−θ0)≤1,b≡εβ2(γ−1)θ0,(7)s_0\equiv b(S_0+1-\theta_0)\leq1,\qquad b\equiv\frac{\varepsilon\beta}{2}(\gamma-1)\sqrt{\theta_0}, \tag*{(7)}

where the parameter s0s_0 characterizing the initial elongation S0S_0 (σ0=1\sigma_0=1) is smaller than unity s0≤1s_0\leq1 because the parameter bb is very small, ranging from 1.2 10−21.2\,10^{-2} in a stoichiometric H2_2-O2_2 flame to 6 10−46\,10^{-4} in hydrocarbon-air flames, b≪1b\ll1 and s0≈bS0≤1s_0\approx bS_0\leq1 of order unity. According to (6), dσ/dϑ=[(1−s0)−ϑ]/s0d\sigma/d\vartheta=[(1-s_0)-\vartheta]/s_0, the maximum of the curve σ\sigma versus ϑ\vartheta (turning point of ϑ(σ)\vartheta(\sigma)) defines a critical flame temperature ϑ∗=1−s0\vartheta^*=1-s_0 and a critical elongation S∗≥S0S^*\geq S_0, σ∗=e(s0−1)/s0≥1 ∀s0\sigma^*=e^{(s_0-1)/s_0}\geq1\ \forall s_0, above which Eq. (6) has no root. The temperature of the downstream-running compression waves in the unburned gas for σ>σ∗\sigma>\sigma^* is no longer compatible with the temperature of the accelerating flame acting as a semi-transparent piston. Such a drastic effect associated to a nonlinear thermal feedback was identified long ago in a different context [15]. It can be illustrated simply by the compression waves issued from an impermeable piston whose velocity is an increasing function of the gas temperature [16]. The critical elongation S∗S^* of (6) increases with the initial one S0S_0, dσ∗/ds0=(1−s0)e(s0−1)/s0/s02>0d\sigma^*/ds_0=(1-s_0)e^{(s_0-1)/s_0}/s_0^2>0 (0<s0≤10<s_0\leq1). There is a common upper bound Smax⁡∗=S0max⁡≈1/bS^*_{\max}=S_{0\max}\approx1/b (s0=1: dσ∗/ds0=0, σ∗=1s_0=1:\ d\sigma^*/ds_0=0,\ \sigma^*=1) and the critical Mach number (uu/a∗)max⁡≈1/[(γ−1)β](u_u/a^*)_{\max}\approx1/[(\gamma-1)\beta] depends only on the thermal sensitivity of the flame and is close to unity as in the experiments [8][10][11]. Eq. (6) has two solutions for σ<σ∗\sigma<\sigma^*: ϑ−(σ)<ϑ∗\vartheta_-(\sigma)<\vartheta^* and ϑ+(σ)>ϑ∗\vartheta_+(\sigma)>\vartheta^*, ϑ+(σ∗)=ϑ−(σ∗)=ϑ∗\vartheta_+(\sigma^*)=\vartheta_-(\sigma^*)=\vartheta^*, dϑ+/dσ<0d\vartheta_+/d\sigma<0 and dϑ−/dσ>0d\vartheta_-/d\sigma>0. According to the thermodynamic laws, the temperature increases by compression, so that ϑ−(σ)\vartheta_-(\sigma) is the physical branch. The smallest critical elongation Smin⁡∗S^*_{\min} is obtained for S0=0S_0=0, s0=b(1−θ0)≪1s_0=b(1-\theta_0)\ll1, σ∗=e(s0−1)/s0≈1/(s0e)\sigma^*=e^{(s_0-1)/s_0}\approx1/(s_0e), Smin⁡∗≈Smax⁡∗/e≈(1−θ0)/(s0e)S^*_{\min}\approx S^*_{\max}/e\approx(1-\theta_0)/(s_0e). Such critical elongations are too large (at least by a factor 10210^2) for the DDT of laminar hydrocarbon-air flames (b≈6 10−4b\approx6\,10^{-4}) to happen in smooth-walled tubes, as it is well known. For very energetic mixtures (b≈1.2 10−2b\approx1.2\,10^{-2}) one gets L∗/R∈[15,40]L^*/R\in[15,40], still larger than in experiments but only by a factor 5. The discrepancy comes from the real chemical kinetics of H2_2-O2_2 flames [5] which cannot be represented in a large range of temperature by an Arrhenius law with a single activation energy. However, with the ZFK result β=2\beta=2 in the form Ub/Ub0≈(Tb/Tb0)e(β/2)δTb/Tb0U_b/U_{b0}\approx(T_b/T_{b0})^{e^{(\beta/2)\delta T_b/T_{b0}}}, Eq. (4) yields critical elongations similar to the experimental data L∗/R≈3L^*/R\approx3 to 10.

Expanding (6) near the critical point (ϑ∗,σ∗)(\vartheta^*,\sigma^*), dσ/dϑ∣ϑ=ϑ∗=0d\sigma/d\vartheta|_{\vartheta=\vartheta^*}=0, yields a quadratic form

σ−σ∗σ∗+(ϑ−ϑ∗)22=0,ϑ∗−ϑ−=2σ∗−σσ∗.(8)\frac{\sigma-\sigma^*}{\sigma^*}+\frac{(\vartheta-\vartheta^*)^2}{2}=0,\qquad \vartheta^*-\vartheta_-=\sqrt{2}\sqrt{\frac{\sigma^*-\sigma}{\sigma^*}}. \tag*{(8)}

showing how the flame temperature TbT_b (or the convective flow uuu_u) behaves when the elongation approaches the critical value σ∗\sigma^*. The key point is the singular derivative dϑ−/dσd\vartheta_-/d\sigma when σ→σ∗\sigma\to\sigma^* from below. Introducing the non-dimensional rate of elongation ε\varepsilon and denoting t∗t^* the time at which the critical elongation σ∗\sigma^* is reached (t≤t∗t\leq t^*)

ε≡tbσ∗dσdt∣t=t∗,Eq. (8)⇒tbdϑ−dt=ε/2(t∗−t)/tb(9)\varepsilon\equiv\frac{t_b}{\sigma^*}\frac{d\sigma}{dt}\bigg|_{t=t^*},\qquad \text{Eq. (8)}\Rightarrow t_b\frac{d\vartheta_-}{dt}=\frac{\sqrt{\varepsilon/2}}{\sqrt{(t^*-t)/t_b}} \tag*{(9)}

the time derivative of the flame temperature increases strongly when approaching t∗t^* (t∗−t≈tbt^*-t\approx t_b) and diverges in finite time t=t∗t=t^*. Smaller the elongation rate ε\varepsilon, sharper the runaway of dϑ−/dtd\vartheta_-/dt.

Finite time singularity and DDT mechanism

(9) corresponds to a finite time singularity of the flame acceleration while the critical velocities ub∗u_b^*, uu∗u_u^* and Ub∗U_b^* stay finite. In fluid mechanics a shock wave is spontaneously generated on the piston when the acceleration becomes singular. This suggests the formation of a shock inside the inner structure of the laminar flame at t=t∗t=t^*. The phenomenon is not that simple. As shown in the next paragraph, due to unsteady effects not included in (9), the DDT mechanism is associated with a run away of the pressure and temperature occurring at tt slightly above t∗t^*. The pressure being quasi-uniform inside the inner flame structure, the latter is blown off as a whole at finite time, leading to a sharp decrease of the flame thickness followed quasi-instantaneously by the detonation offset. Involving time and length scales of the order of the mean free path and collision time respectively, neither the formation of the lead shock of the detonation nor the transient overshoot of pressure and/or temperature cannot be accurately described by the macroscopic equations of fluid mechanics. However, once the overdriven detonation is formed, the decay to the Chapman-Jouguet regime can be properly obtained by treating the lead shock as a discontinuity across which the Rankine-Hugoniot jump conditions are applied. Notice that the singularity is not due to the nonlinear wave breaking usually involved in the formation of shock waves but by the finite-time divergence of the flame acceleration in (9), as said before.

Dynamical saddle node bifurcation induced DDT

The understanding of the abrupt transition requires to perform an unsteady analysis of the full reactive flow, not only of the downstream-running compression waves ahead of the flame described in (2). Two additional unsteady effects have to be investigated, one concerning the flow of burned gas, and the other the inner structure of the laminar flame. Due to the complexity of the curved flow of burned gas sketched in fig.1, the first unsteadiness can be only roughly evaluated; hopfully a detailed solution is not actually needed. The back-flow ub(t)u_b(t) on the tip of the elongated flame results from a cumulative effect of burned gas flows issued from the lateral flames that are fed by the fresh gas of the tongues quasi-parallel to the wall, ub(t)=(2/R)∫XP−LXPUb(t−Δtbg)dxu_b(t)=(2/R)\int_{X_P-L}^{X_P}U_b(t-\Delta t_{bg})dx where Ub(t)U_b(t) is the laminar flame velocity on the tip and Δtbg(x)∝(XP−x)/2a\Delta t_{bg}(x)\propto(X_P-x)/2a is the delay for transferring to the tip the transient flow of burned gas of the lateral flame at a distance XP−xX_P-x from the tip. For a change in laminar flame speed sufficiently slow (L/a)dUb/dt≪Ub(L/a)dU_b/dt\ll U_b, (tb/Ub)dUb/dt=O(ϵ)(t_b/U_b)dU_b/dt=O(\epsilon), the delayed model (1) ub(t)=S(t)Ub(t)u_b(t)=S(t)U_b(t) is replaced by a delayed back-flow evaluated by integrating Δtbg(x)\Delta t_{bg}(x)

ub(t)≈S(t)[Ub(t)−ΔtwdUb/dt],Δtw≈L/(2a)(10)u_b(t)\approx S(t)[U_b(t)-\Delta t_w dU_b/dt],\qquad\Delta t_w\approx L/(2a) \tag*{(10)}

where L∗/(2atb)<1/ϵL^*/(2at_b)<1/\epsilon is anticipated, L∗/d<2/(ϵϵ)L^*/d<2/(\epsilon\epsilon).

The unsteady analysis of the ZFK flame structure is performed in the limit ϵ≪ϵ≪1\epsilon\ll\epsilon\ll1, b=O(1)b=O(1), retaining unsteady terms of order ϵϵ\epsilon\epsilon while neglecting terms of order ϵ2\epsilon^2. This leads to an equation similar to (4)

(uu−ub)/Ub0=(1−θ0)eβ(θb−1)/2[1+O(ϵ)](11)(u_u-u_b)/U_{b0}=(1-\theta_0)e^{\beta(\theta_b-1)/2}[1+O(\epsilon)] \tag*{(11)}

but where the flame temperature θb(t)≡Tb(t)/Tb0\theta_b(t)\equiv T_b(t)/T_{b0} is no longer related to the instantaneous value of the flow in the unburned gas uu(t)u_u(t) by the first relation in (5). The unsteadiness induced variation of flame temperature δθb(t)≡δTb(t)/Tb0=O(ϵϵ)\delta\theta_b(t)\equiv\delta T_b(t)/T_{b0}=O(\epsilon\epsilon) and of mass flux across the reaction sheet δeβ(θb−1)/2=O(βϵ)\delta e^{\beta(\theta_b-1)/2}=O(\beta\epsilon) are obtained from the unsteady analysis of the inner flame structure. This involves technical difficulties but the unsteady effects are easily understood: according to the solution of the energy equation perturbed by a transient compressional heating, heat conduction introduces a time delay Δtθ\Delta t_\theta proportional to the transit time across the flame Δtθ/tb=O(1)\Delta t_\theta/t_b=O(1), the compressible transfer, (d/ab)/(d/Ub)≈ϵ(d/a_b)/(d/U_b)\approx\epsilon being negligible. Therefore the instantaneous mass flux across the reaction sheet differs from the quasi-steady solution by a term proportional to the time-derivative of the pressure (dpu/dt)/pu0=γ(duu/dt)/au0=O(ϵϵ/tb)(dp_u/dt)/p_{u0}=\gamma(du_u/dt)/a_{u0}=O(\epsilon\epsilon/t_b). This introduces a derivative-term proportional to (Δtθ/Ub)duu/dt=O(ϵ)(\Delta t_\theta/U_b)du_u/dt=O(\epsilon) into the right-hand side of (11). Introducing δuu(t)\delta u_u(t) given by (3) and the delayed back-flow (10) δub(t)/S(t)=−(Δtw/Ub)dUb/dt=O(ϵ)\delta u_b(t)/S(t)=- (\Delta t_w/U_b)dU_b/dt=O(\epsilon) into the left-hand side of (11) yields a dynamical equation for the flame temperature. Near the critical point, one gets the first equation (8) completed by a time-derivative term δKdϑ/dt\delta K d\vartheta/dt on the right-hand side with a non-dimensional coefficient K∝(Δtw−Δtθ)/tbK\propto(\Delta t_w-\Delta t_\theta)/t_b. Using (σ−σ∗)/σ∗=ϵ(t−t∗)/tb(\sigma-\sigma^*)/\sigma^*=\epsilon(t-t^*)/t_b, the so-obtained equation characterizes a dynamical saddle-node bifurcation. Conveniently rescaled, ϑ−ϑ∗→y∝(σ∗/ϵ0∣K∣)1/3(ϑ−ϑ∗)\vartheta-\vartheta^*\to y\propto(\sigma^*/\epsilon_0|K|)^{1/3}(\vartheta-\vartheta^*), (σ−σ∗)/σ∗→τ∝(σ∗/ϵ0∣K∣)2/3(σ−σ∗)/σ∗(\sigma-\sigma^*)/\sigma^*\to\tau\propto(\sigma^*/\epsilon_0|K|)^{2/3}(\sigma-\sigma^*)/\sigma^*, this equation takes a normal form extensively used for sharp transitions in physics, biophysics and catastrophic events [17]

τ+y2=±dy/dτ(12)\tau+y^2=\pm dy/d\tau \tag*{(12)}

where ±\pm is the sign of the coefficient KK and τ=0\tau=0, y=0y=0 is the turning point. The trajectories in the phase space show that the branch of negative quasi-steady-solutions yˉ(τ)=−−τ\bar{y}(\tau)=-\sqrt{-\tau} is stable with the ++ sign (K>0K>0) and unstable with the −- sign (K<0K<0). Therefore, the unsteady inner structure of the flame contributing to KK by a negative term −Δtθ/tb<0-\Delta t_\theta/t_b<0, destabilizes the flame subjected to an instantaneous back-flow. The branch of physical solutions (8) is stabilized if the delay introduced by the flow of burned gas is longer than the transit time across the inner structure Δtw−Δtθ=O(tb)\Delta t_w-\Delta t_\theta=O(t_b), K>0K>0. This is the case if the finger flame is much longer than the laminar flame thickness, L/d>1/(ϵϵ)L/d>1/(\epsilon\epsilon) (L≥1L\geq1 cm in energetic mixtures). The key point of the saddle-node bifurcation (12) is that, starting from the stable branch τ=τi<0\tau=\tau_i<0: y=−−τiy=-\sqrt{-\tau_i}, ∣τi∣>1|\tau_i|>1, the solution is blown off at τ=τc≈2.338\tau=\tau_c\approx2.338, y(τ)≈1/(τc−τ)y(\tau)\approx1/(\tau_c-\tau) [17], in contrast to y‾−(τ)\overline{y}_-(\tau). Unsteadiness makes the entire laminar flame structure blown off few time after the turning point. The smaller the elongation rate ϵ≪1\epsilon\ll1, the closer to the turning point the pressure runaway (DDT onset).

Limitation of the analysis and perspectives

Equation (12) is obtained by an expansion around the turning point. Therefore, the asymptotic behavior y(τ)≈1/(τc−τ)y(\tau)\approx1/(\tau_c-\tau) for τ=τc≈2.338\tau=\tau_c\approx2.338 is not guaranteed for the exact solution of the macroscopic equations satisfying the boundary conditions ahead of and behind the self-accelerating flame under the effect of the back-flow. Nevertheless, the onset of a finite time singularity is not doubtful because terms of higher order than in (12) are expected to reinforce the singularity. This is illustrated by the divergence of the acceleration dϑ/dτ∝1/τ∗−τ\mathrm{d}\vartheta/\mathrm{d}\tau\propto1/\sqrt{\tau^*-\tau} for a flame structure in steady state becoming dϑ/dτ∝1/(τc−τ)2∝(ϑc−ϑ)2\mathrm{d}\vartheta/\mathrm{d}\tau\propto1/(\tau_c-\tau)^2\propto(\vartheta_c-\vartheta)^2 when a weak unsteadiness is taken into account. Moreover, when a second derivative term d2y/dτ2\mathrm{d}^2y/\mathrm{d}\tau^2 becomes larger than the first derivative on the right-hand side of (12) the asymptotic solution of such a first Painlevé transcendent y≈6/(τc−τ)2y\approx6/(\tau_c-\tau)^2 shows a stronger finite time singularity. A numerical analysis of the one-dimensional back flow model including the full chemical kinetics is welcomed to analyze carefully the finite time singularity.

The saddle-node bifurcation induced DDT is not limited to the ZFK flame model used here in the asymptotic analysis of the one-dimensional back flow model summarized in this manuscript. As already mentioned, the key point is the nonlinear temperature dependence of the laminar flame velocity characterizing any combustible mixture. The latter function can be easily computed numerically for any complex chemical network of real combustible mixtures and used instead of the exponential term in (6). Moreover the asymptotic analysis can be extended without much difficulty to reduced chemical networks treating the thin zone of radical production as a discontinuity. The results will be qualitatively the same as that presented here.

Such a DDT mechanism concerns also turbulent wrinkled flame treated as a chaotic array of elongated flames since a convection flow is systematically induced by the front wrinkling so that the tip of each cell of the cellular of a turbulent flame looks like to that of the elongated flame in fig.1. In the cellular structure of a Rayleigh-Taylor unstable flame expanding freely in spherical geometry, the longitudinal and transverse dimension of the cells are similar S≈1S\approx1. In very energetic mixtures the parameter b∝βϵb\propto\beta\epsilon (the product of the thermal sensitivity of the flame velocity by the flame Mach number) might not be very small so that S∗∼1/(ϵb)S^*\sim1/(\epsilon b) could be close to unity. This could be the case for the nuclear reactions involved in stars. Further studies are worth performing to decide whether the dynamical saddle-node bifurcation (12) is a DDT mechanism explaining the explosion of SNIa supernovae.

References

  1. [1]K.I. Shchelkin and Yu.K. Troshin, Gasdynamics of Combustion (Mono Book. corp. Baltimore, 1965).
  2. [3]P. Clavin and G. Searby, Combustion Waves and Fronts in Flows (Cambridge University Press, Cambridge, 2016)DOI
  3. [4]P. Clavin, ”One-dimensional mechanism of deflagration-to-detonation in gas”, J. Fluid. Mech. Submitted (2023)DOI
  4. [5]A. Sanchez and F.A. Williams, ”Recent advances in understanding of flammability characteristics of hydrogen”, Progress in Energy and Combustion Science 41, 1-55 (2014).DOI
  5. [6]Ya.B. Zeldovich and Frank-Kamenetskii, ”A theory of thermal flame propagation”, Acta Phys. Chim. 9, 341-350 (1938).
  6. [8]V. Bikov, A. Koksharov, M. Kuzetsov, and V.P. Zhukov, ”Hydrogen-oxygen flame acceleration in narrow open ended channels”, Combust. Flame 238, 111913 (2022).
  7. [9]P.A. Urtiev and A.K. Oppenheim, ”Experimental observations of the transition to detonation in an explosive gas”, Proc. R. Soc. London A 295, 13-28 (1966).
  8. [10]M.A. Liberman, M.F. Ivanov, A.D. Kiverin, M.S. Kuzetsov, A.A Chukalovsky and T.V. Rakhimova, ”Deflagration-to-detonation transition in highly reactive combustion mixtures”, Acta Astronautica 67, 688-701 (2010).
  9. [11]M. Kuzetsov, M. Liberman and I. Matsukov, ”Experimental study of the preheated zone formation and deflagration to detonation transition”, Combust. Sci and Tech. 182, 1628-1644 (2010).
  10. [12]Ya. B. Zeldovich, ”Regime classification of an exothermic reaction with nonuniform initial condition”, Combust. Flame 39, 211-214 (1980)DOI
  11. [13]C. Clanet and G. Searby, ”On the tulip flame phenomenon”, Combust. Flame 105, 225-238 (1996)DOI
  12. [14]P. Clavin, ”Finite-time singularity associated with the deflagration-to-detonation transition on the tip of an elongated flame-front in a tube”, Combust. Flame 245, 112347 (2022)DOI
  13. [15]B. Deshaies and G. Joulin, ”Flame-speed sensitivity to temperature changes and the deflagration-to-detonation transition”, Combust. Flame 77, 202-212 (1989)DOI
  14. [16]P. Clavin and H. Tofaili, ”A one-dimensional model for deflagration-to-detonation transition on the tip of elongated flames in tubes”, Combust. Flame 232, 111521 (2021)DOI
  15. [17]R.D. Peters, M. Le Berre and Y. Pomeau, ”Prediction of catastrophes: a experimental model”, Phys. Rev. E 86, 026207 (2012)DOI

Paper details

vlnc:W_h8q2h8cn
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents