Anda di halaman 1dari 23

Downloaded from rspa.royalsocietypublishing.

org on October 4, 2010

Shock interaction with solid particles in condensed matter and related momentum transfer
Fan Zhang, Paul A. Thibault and Rick Link Proc. R. Soc. Lond. A 2003 459, 705-726 doi: 10.1098/rspa.2002.1045

Email alerting service

Receive free email alerts when new articles cite this article sign up in the box at the top right-hand corner of the article or click here

To subscribe to Proc. R. Soc. Lond. A go to: http://rspa.royalsocietypublishing.org/subscriptions

This journal is 2003 The Royal Society

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

10.1098/ rspa.2002.1045

Shock interaction with solid particles in condensed matter and related momentum transfer
By F an Z h a n g1 , Paul A. Thiba ult2 a n d R i c k L i n k3
1 Defence R&D Canada-Su eld, PO Box 4000, Stn Main, Medicine Hat, AB T1A 8K6, Canada (fan.zhang@drdc-rddc.gc.ca) 2 Timescales-Scientic Ltd, 20 Bedford Street, Bedford, NS B4A 1W6, Canada 3 Combustion Dynamics Ltd, 1888 Brunswick Street, Halifax, NS B3J 3J8, Canada

Received 11 April 2002; accepted 9 July 2002; published online 20 January 2003

Detonation propagation in a condensed explosive with metal particles can result in signi cant momentum transfer between the explosive and the particles during their crossing of the leading shock front. Consequently, the assumption of a `phaseinteraction-frozen shock used in multiphase continuum models for detonation initiation and propagation may not be valid. This paper addresses this issue by performing numerical and theoretical calculations in liquid explosives and RDX with various compressible metal particles under conditions of detonation pressure. The results show that the momentum transferred to heavy-metal particles such as tungsten is not signi cant after the shock{particle interaction. However, light-metal particles including aluminium, beryllium and magnesium rapidly accelerate during the shock{ particle interaction. They reach a considerable speed immediately behind the shock front, typically 60{100% of the ow speed of the explosive. It is important to take this signi cant momentum transfer at the shock front into account when modelling the shock initiation and detonation structure for two-phase mixtures of condensed explosive and light-metal particles.
Keywords: shock waves; detonations; multiphase; particle-uid ows

1. Introduction
For detonation propagation in a condensed explosive with metal particles, interaction between the leading shock front and the particles could result in a momentum loss to the explosive that inuences the detonation initiation and structure. Theoretical models for detonation in a uid{solid particle system have mostly been based on twophase uid dynamics models taking mass, momentum and heat transfer between the phases into consideration (Zeldovich & Kompaneets 1960). Dynamic compaction of packed solid particles has also been taken into account for modelling deagration and detonation in granular materials (Baer & Nunziato 1986). In the two-phase uid dynamics models, a frozen shock{particle interaction is often assumed in which the solid particles are not accelerated during crossing of the shock front. Behind the shock front, a drag force is assumed to determine the momentum transfer between the uid phase and the particles. For detonation in two-phase mixtures of gas and
Proc. R. Soc. Lond. A (2003) 459, 705{726 2003 The Royal Society c

705

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

706

F. Zhang and others

solid particles, the shock interaction time in which a particle crosses the shock front is several orders of magnitude smaller than the velocity relaxation time related to the drag. Thus, the particle crosses the shock front with negligible changes in its velocity and the assumption of frozen shock interaction has been proven adequate. Momentum loss behind the shock front plays a role in the detonation velocity de cit for gas{ particle systems with considerable particle combustion beyond the sonic choking point (Lee & Sichel 1986; Gelfand et al . 1991; Zhang et al . 1992; Zhang & Lee 1994; Veyssiere & Ingignoli 2002). However, for detonation in condensed matter containing metal particles, the shock interaction time can be comparable with the velocity relaxation time and therefore the assumption of frozen shock interaction may fail. While the deformation of the metal particles and the multi-reected shock on their surface may cause hot spots in the explosive, the explosive loses momentum during the shock{particle interaction if the reaction time of metal particles is larger than the shock interaction time. Momentum transfer during the shock interaction together with that behind the shock front can both inuence the shock initiation and detonation structure. Detonation velocity de cit was observed experimentally in condensed explosives with 0.1 m m aluminium particles (Baudin et al . 1998; Gogulya et al . 1998; Haskins et al. 2002; Zhang et al . 2002). A spherical particle crossing a planar shock wave is a simple but good example to elucidate the suitability of the frozen shock interaction assumption. Considering a spherical particle of diameter d and density s moving in a continuum uid of density and neglecting any impressed force such as gravity, Newtons law of motion becomes

u us dus ; = v dt

(1.1)

in which the velocity relaxation time of motion is


v

4d2 s : 3 Cd Re

(1.2)

Here, us stands for the particle velocity, u for the uid velocity and for the dynamic viscosity. The viscous drag coe cient, Cd , de ned in terms of the dynamic head of the relative ow and the frontal area of the particle, is a function of the Reynolds number and the Mach number in a compressible uid based on Reynoldss principle of similarity (Schlichting 1979). On the other hand, the shock interaction time, , can be de ned as the time in which a particle crosses the shock front, which is approximately = d=D0 ; (1.3) where D 0 is the shock wave velocity. Thus, the ratio of the shock interaction time over the velocity relaxation time is

=
v

3 Cd Re : 4d s D0

(1.4)

Substitution of typical numbers for a gaseous detonation into equation (1.4) shows that, even for particles of 0.1{1 m m, the shock interaction time is about three to four orders of magnitude smaller than the relaxation time. Thus, the particle crosses the shock front with negligible changes in its velocity and the assumption of frozen shock
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

Shock interaction with solid particles in condensed matter

707

interaction is reasonable for detonation in two-phase mixtures of gas and solid particles. However, for detonation in condensed matter such as in liquid explosive with aluminium particles of 0.1{1 m m, the shock interaction time is about the same order of magnitude as, or one order smaller than, the relaxation time. Therefore, the particle could be accelerated to a considerable velocity during passage of the shock front, which makes the assumption of a frozen shock{particle interaction questionable. The objectives of the present paper are to calculate the momentum transfer during the shock interaction with compressible metal particles in host condensed matter and to understand the mechanism for the momentum transfer at the shock front of a detonation wave. Numerical simulations were performed and analysed both for a single particle and for multiple particles that cross a planar shock front. The study was carried out for various material combinations of particles and host matter in the condensed phase, and for shock strengths under conditions of detonation pressures which exceed the yield strength of solid particles. Quasi-one-dimensional analysis was also conducted to obtain more insight into the mechanism of momentum transfer when the particle crosses the shock front.

2. Shock interaction with a single particle


Classic detonation theories for homogeneous condensed explosives have been based on the ZND model in which detonation occurs through a frozen shock transition followed by an induction period, where the shocked temperature results in vibrational, rotational and electronic excitation followed by dissociation of the molecules (Zeldovich & Kompaneets 1960). During the frozen shock transition, explosives are assumed to be chemically unchanged. Various hot-spot mechanisms have been developed to interpret the nature of the detonation initiation in solid heterogeneous explosives (Campbell et al . 1961; Dremin et al. 1970; Lee & Tarver 1980). The frozen shock transition, however, has remained a fundamental assumption following the ZND model. Since the late 1970s, it has been speculated that detonation initiation could start within the shock front due to the non-equilibrium kinetics (Dremin & Klimenko 1981; Walker 1988; Dremin 1992). Experimental proof of this postulate must be inferred through observation inside the shock front and, therefore, remains a di cult diagnostics challenge. In the present paper, only mechanical interactions between the condensed explosive and the particles during passage of the shock front were addressed. Hence, it was assumed that decomposition of the condensed explosive would not start within the shock front. It was also assumed that the reaction time-scale of solid particles is larger than the shock interaction time so that the particles can be considered chemically frozen within the shock front. Because of the assumption of the chemically frozen shock transition, the explosive material was described using the Murnaghan equation of state n K0 1 + p0 ; (2.1) p= 0 n

where p, , K0 and n are the pressure, density, bulk modulus and the pressure derivative of the bulk modulus, respectively; the subscript `0 denotes the initial state. Note that temperature does not appear as a variable in the Murnaghan equation since it is mainly concerned with mechanical interactions. The response of the solid
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

708

F. Zhang and others

particle material during the shock interaction was described using the HOM equation of state (Mader 1998): ps = ( s =v s )(es eH ) + pH (2.2) with c20 (vs 0 vs ) s (2.3) ; eH = 1 pH (vs 0 vs ): 2 [vs 0 ss 0 (vs 0 vs )]2 The HOM equation of state is based on the shock Hugoniot and the Mie Gruneisen equation of state. In equations (2.2) and (2.3), e is the speci c energy, v is the speci c volume, s is the Gruneisen coe cient, and cs 0 and ss 0 are the coe cient to the linear t of the shock and particle speed: Ds = cs 0 +ss 0 us . The subscripts `s and ` H denote the solid particle material and the Hugoniot state, respectively. The Euler equations for mass and momentum were used to model the ow of the non-reacted explosive, and the governing equations for inviscid plastic ow were employed for modelling the dynamic response in metal particles. The interaction between the explosive and the particles was solved by matching their boundary conditions of pressure and particle velocity. Numerical solution of these closure equations were obtained using the IFSAS code (Thibault & Moen 1997), in which the Euler equations were solved using a Lax{Wendro/FCT algorithm with sixth-order phase error correction and the plastic dynamic equations were solved using a nite-element scheme (Boris 1976; Book & Fry 1984; Hallquist 1983). Deviatoric deformation of the particle material that does not produce any volume change was also incorporated. The resolution for the calculations corresponded to 20 cells or elements for a particle radius, and a cylindrically axi-symmetrical mesh was used to represent a spherical particle immersed in the uid. As for initial and boundary conditions, a planar shock wave was applied as an in-ow boundary condition from the upstream of the particle, whereas the downstream and lateral boundaries were treated as a free boundary extrapolated with an expanding cell technique. For a particle crossing a planar shock front, the inviscid governing equations and the rate-independent material response models do not introduce any additional length- or time-scales. Therefore, there exists a simple geometric scaling rule: at any given time, td2 , the surrounding ow eld and the dynamic response of a particle with diameter d2 can be geometrically scaled to the same state as a particle with diameter d1 at a time td1 : td2 =td1 = d2 =d1 : (2.4) pH =

Consequently, calculations need only to be conducted for one particle diameter, and the results can be scaled to any other particle diameters. In the following calculations, the particle diameter was arbitrarily chosen as 10 m m. According to the geometry scaling equation (2.4), for example, the time for a 0.1 m m particle is scaled down to 1/100th of that for the 10 m m particle. Note that the simple geometry scaling would no longer be valid for viscous ows since the additional length-scale associated with viscosity introduces the Reynolds number as an additional scaling parameter. Similarly, kinetic phenomena such as strain-rate eects or chemical reactions would also introduce additional scaling parameters. (a) E ects of a particles material density Various metal particles were subjected to a 101.3 kbar shock in a liquid with the same initial density and compressibility as water ( 0 = 1 g cm3 , c0 = 1477 m s1 ,
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

Shock interaction with solid particles in condensed matter


1 ns 2 ns 3 ns 4 ns

709

Figure 1. Pressure contours for a magnesium particle subjected to a 101.3 kbar liquid shock.

1 ns

2 ns

3 ns

8 ns

Figure 2. Pressure contours for a tungsten particle subjected to a 101.3 kbar liquid shock.

K0 = 21:78 kbar and n = 7:15). The shocked ow properties of the liquid are density 1 = 1:639 g cm3 , ow velocity u1 = 1987 m s1 and shock velocity D0 = 5097 m s1 , where the values of ow velocity and shock velocity are relative to a xed observer. The particle motion, deformation and pressure distribution in the particle and the surrounding uid are displayed as pressure contour plots in gures 1 and 2 for magnesium and tungsten. Times of the plots refer to the beginning of the calculation when the incident shock is 2 m m before the leading edge of the particle. It can be seen that the shock interaction with the particle results in a reected shock into the liquid and a shock transmitted into the particle. The particle is severely deformed and distorted due to the complex lateral expansion and the fact that the shock pressure exceeds the yield stress of the material. The distortion is directly related to the relative velocity between the leading and trailing edges of the particles, which are displayed in gure 3 for magnesium. Local velocities inside the particle are rst produced during the propagation of the transmitted shock. The velocity at the leading edge of the
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

710
2000 velocity (m s- 1) 1500 1000 500 0 - 500 0

F. Zhang and others

4 time (ns)

Figure 3. Leading- and trailing-edge velocity histories for a magnesium particle subjected to a 101.3 kbar liquid shock. 2000

Mg Al

1500 s- 1) us (m

1000

Ni W

500

0 0 2 4 6 time (ns) 8

10

Figure 4. Particle mass-centre velocity histories in magnesium, aluminium, nickel and tungsten subjected to a 101.3 kbar liquid shock.

particle is very close to the analytical value of the one-dimensional wave transmission, 1474 m s1 for light magnesium and 407 m s1 for heavy tungsten. The local velocities decay behind the transmitted shock and along the propagation distance due to the unloading eect of lateral expansion. Furthermore, the rarefaction that is reected o the trailing edge of the particle competes against lateral expansion and causes the secondary acceleration. Due to the non-uniform local velocities inside the particle, the particle mass-centre velocity is calculated based on a mass average over the particle elements. Figure 4 provides the histories of particle mass-centre velocity for four metals: magnesium, aluminium, nickel and tungsten with increasing initial density. The particles velocity histories display two phases: a shock{particle interaction phase which occurs over a time = 1:96 ns from equation (1.3), followed by a drag phase. It can be seen from this gure that the particle acceleration decreases with the particle material denProc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

Shock interaction with solid particles in condensed matter


Table 1. Velocity transmission factors

711

material

s0 (g cm

cs 0 (m s 1 )
3

p1 (kbar)

magnesium

liquid 0 = 1:0 g cm 1.770 4700

beryllium aluminium nickel uranium tungsten

1.870 2.785 8.860 18.98 19.30

7975 5350 4646 2540 4060


3

50.65 101.3 202.6 101.3 101.3 101.3 101.3 101.3 202.6 101.3 101.3
3

0.776 0.790 0.803 0.781 0.600 0.227 0.108 0.108 0.114 0.940 0.754 1.008 0.802

magnesium aluminium magnesium aluminium

RDX 0 = 1:4 g cm 1.770 4700 2.785 5350 RDX 0 = 1:8 g cm 1.770 4700 2.785 5350

101.3 101.3

sity. While the heavy particles such as tungsten accelerate mildly, the light particles accelerate rapidly during the shock{particle interaction. For example, the magnesium particle accelerates to a velocity of 1600 m s1 that corresponds to ca. 80% of the surrounding shocked uid velocity of 1987 m s1 . A velocity transmission factor can be de ned as the ratio of the particle masscentre velocity us after the shock interaction time over the shocked uid velocity u1 : = us =u1 : (2.5) The velocity transmission factor varies between 0 for perfect reection o a rigid body to 1 for a perfect transmission into a particle with the same material properties as the unreacted uid. The computed velocity transmission factor is displayed in table 1 for the four metal particles. Similar to the velocity distribution, the pressure distribution inside the particle is initially highly non-uniform, as exhibited in gure 5, which displays pressure histories for elements at the leading edge, centre and trailing edge of the magnesium and tungsten particles. The shock pressure at the leading edge of the particle is very close to the values from the one-dimensional wave transmission theory, 166.8 kbar for magnesium and 358.3 kbar for tungsten. However, the shock pressure decays in the particle due to the lateral expansion; the rate of decay is much faster in tungsten than in magnesium due to the higher acoustic impedance of tungsten. The relative pressure between the leading and tailing edge for both magnesium and tungsten approaches zero shortly after the shock interaction. Thus, the particle pressure, computed on a mass average over the particle elements, rapidly reaches and oscillates around the initial liquid shock pressure of 101.3 kbar. Finally, given density and pressure, we can also compute the particle temperature using the HOM equations of state for temperature based on the Walsh and Christian method (Mader
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

712
200 static pressure (103 atm) 150 100 50 0 - 50 400 static pressure (103 atm) 300 200 100 0 - 100 0 (b) 2 (a) 0

F. Zhang and others

6 time (ns)

10

Figure 5. Pressure history for leading edge, centre and trailing edge of a magnesium (a) and a tungsten (b) particle subjected to a 101.3 kbar liquid shock.

1998; Walsh & Christian 1955). Consequently, the temperature histories computed also show a rapid equilibration to the Hugoniot temperature of the metal particle, i.e. 450 K, 371 K, 329 K and 314 K for magnesium, aluminium, nickel and tungsten, respectively. Strictly speaking, the late-time calculations in the drag phase after the shock interaction is not correct without using governing equations for viscous uid. However, some qualitative phenomena observed in the calculations for the shock interaction and shortly thereafter may bring attention to an appropriate description of the drag under conditions of detonation pressures. Due to the severe particle deformation, particle acceleration in the drag phase can be relatively complex. For instance, the considerable increase in cross-sectional area associated with the above-mentioned lateral expansion and distortion can play a role in the drag force exerted on the particle. The sharp edge produced at the trailing edge can also result in ow separation and a further increase in the drag coe cient. Furthermore, the large deformation can also lead to shedding of small particles at the sharp edge. Hence, the drag law cannot be expressed simply in terms of the local relative ow properties, but must also account for the history-dependent particle shape. (b) E ects of particle acoustic impedance The dependence of the velocity transmission factor on a particles acoustic impedance, s 0 cs 0 , was explored by performing calculations using materials with similar initial densities but with dierent sound speeds. Beryllium was compared
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

Shock interaction with solid particles in condensed matter


1 ns 2 ns 4 ns

713

Figure 6. Pressure contours for a beryllium particle subjected to a 101.3 kbar liquid shock. 1 ns 2 ns 8 ns

Figure 7. Pressure contours for a uranium particle subjected to a 101.3 kbar liquid shock.

with magnesium for the low-density regime while uranium and tungsten were compared for the high-density regime. Figures 6 and 7 show the pressure contour plots for beryllium and uranium, respectively. It can be seen from a comparison of gure 6 with gure 1 that the transmitted shock propagates faster in beryllium than in the softer magnesium, which also displays a slightly larger deformation. Similarly, comparison between gures 7 and 2 indicates that the transmitted shock propagates faster in tungsten than in the softer uranium. In spite of these dierences, the particles mass-centre velocity imparted after the shock interaction remains almost the same as shown in gure 8, thus resulting in a similar velocity transmission factor as displayed in table 1. One noticeable dierence is the higher distortion produced for the lower-impedance materials. In particular, much higher distortion is observed in the softer uranium than tungsten after the shock interaction process. The higher distortion causes a larger drag force that causes the particle velocity for uranium to diverge from that for tungsten, as seen in gure 8.
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

714
2000 (a)

F. Zhang and others

1500 us (m s- 1)

1000

500

Mg Be

0 0 600 (b) 2 4 6 8

400 us (m s- 1) 200 W U 0 0 2 4 6 time (ns) 8 10

Figure 8. Particle mass-centre velocity histories for magnesium and beryllium (a) and for tungsten and uranium (b).

(c) E ects of shock strength To determine the dependence of the velocity transmission factor on the incident shock strength, additional calculations were performed with incident shock pressures of 50.7 kbar and 202.6 kbar. The pressure contour plots for these calculations are presented in gure 9 for tungsten. An increase in shock strength does result in a faster particle distortion rate which plays a role in calculating the drag force. However, within the shock interaction process, the increase in velocity transmission factor with shock strength is relatively small as seen in table 1. The velocity transmission factors in table 1 are calculated based on the particles mass-centre velocity at the shock interaction time , which is 2.5 ns, 1.96 ns and 1.48 ns for the 50.7 kbar, 101.3 kbar and 202.6 kbar shocks, respectively.
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

Shock interaction with solid particles in condensed matter


1 ns 2 ns 3 ns 6 ns

715

Figure 9. Pressure contours for a 50.7 kbar (a) and 202.6 kbar (b) liquid shock interacting with a tungsten particle.

(d ) E ects of host matter The studies presented above have dealt with particles in a liquid matter. To examine the inuence of dierent condensed matter on the velocity transmission factor, a solid explosive, RDX, is chosen to replace the liquid. RDX is commonly considered as a heterogeneous explosive with a grain size distribution and cannot be rigorously described by a homogeneous uid model. Moreover, mixing of metal particles with RDX introduces the further complexity of heterogeneity. However, the bulk response of unreacted RDX to shock waves is still worthwhile studying in respect of the inuence of shock Hugoniots and material densities of host matter on the shock{particle interaction process. Thus, curve ts of the Murnaghan equation of state were conducted to RDX shock Hugoniots (Dremin et al . 1970; Marsh 1980) for initial densities of 1.44 g cm3 and 1.6{1.8 g cm3 . This yields the following parameters of the equation of state: K0 = 52:77 kbar and n = 4:006 for 1.44 g cm3 and K0 = 109:8 kbar and n = 7:294 for 1.6{1.8 g cm3 , respectively.
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

716
1.1 1.0 0.9

F. Zhang and others

0.8 0.7 0.6


0.5 3000

4000 5000

6000 7000 D0 (m s- 1)

8000 9000 10000

Figure 10. Velocity transmission factors for magnesium and aluminium particles subjected to RDX shocks. , Mg, RDX (1.8 g cm 3 ); N, Al, RDX (1.4 g cm 3 ); O, Al, RDX (1.8 g cm 3 ).

600

400 ps (kbar) 200 0 3000

4000 5000

6000

7000

8000 9000 10000

D0 (m s- 1)

Figure 11. Mass-averaged particle pressures at the shock interaction time for magnesium and aluminium subjected to RDX shocks. , Mg, RDX (1.8 g cm 3 ); N, Al, RDX (1.4 g cm 3 ); O, Al, RDX (1.8 g cm 3 ).

The velocity transmission factor was computed for magnesium and aluminium particles at various RDX shock strengths up to a shock velocity of D0 = 8930 m s1 , which corresponds to the Chapman{Jouguet detonation of RDX at an initial density of 1.8 g cm3 with a shock interaction time of = 1:12 ns. It is expected that the particle distortion rate is increased with shock pressure on the high-density uid. However, as displayed in gure 10, the variation of the velocity transmission facProc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

Shock interaction with solid particles in condensed matter


1.0 0.8 0.6

717

a
0.4 0.2 liquid RDX1.4 RDX1.8 curve fit 1D theory 0.2 0.4 0.6 p0 / ps0 0.8 1.0

Figure 12. Velocity transmission factors for various particle materials subjected to a 101.3 kbar shock in a liquid and RDX.

tor after the shock interaction is within 10% for a range of RDX shock velocity from 4000 m s1 to 9000 m s1 . It is noticeable that the velocity transmission factor becomes larger than 1 for a magnesium particle immersed in 1.8 g cm3 RDX. This is due to the transmitted expansion caused by the smaller density of magnesium (1.77 g cm3 ) than that for RDX. At the shock interaction time, the mass-averaged particle pressure rapidly reaches the initial RDX shock pressures over the shock speed range between 4000 m s1 and 9000 m s1 (see gure 11). The velocity transmission factor computed for various metal particles immersed in dierent condensed matters at several shock pressures are summarized in table 1. Comparison of the results for shocks in liquid and RDX with various metal particles ( gure 12) reveals that the velocity transmission factor depends on the initial density ratio of host condensed matter over particle, 0 = s 0 , rather than on the host condensed matter or the particle alone. Among the material properties and shock strength investigated, is most dependent on this initial density ratio.

3. Shock interaction with multiple particles


(a) A linear cluster of particles To understand the complex shock interaction processes for multiple particles, onedimensional calculations were performed for a linear cluster of aluminium particles subjected to a 101.3 kbar liquid shock that served as an in-ow boundary condition from the upstream of the particles. The linear cluster was composed of twelve 10 m m thick aluminium slugs with a 10 m m gap lled with liquid between slugs. Figure 13 displays an oscillatory structure for the mass-centre velocity history of the leading particle of the linear cluster. The rst sharp velocity rise corresponds to the shock transmitted into the metal and agrees with the analytical value of 1110 m s1 . This pulse is followed by a second pulse which is due to the rarefaction wave produced when the shock reaches the metal{liquid interface at the trailing edge
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

718
2000 1500 velocity (m s- 1) 1000 500 0 - 500 0 2

F. Zhang and others

6 8 time (ns)

10

12

14

Figure 13. Velocity history of the leading particle in a one-dimensional linear cluster of 12 aluminium particles with a 10 m m gap, subjected to a 101.3 kbar liquid shock.

1 ns

2 ns

3 ns

4 ns

5 ns

Figure 14. Pressure contours for two cylindrical aluminium particles with a 5 m m gap subjected to a 101.3 kbar liquid shock. There is a 2.5 m m gap between the particles and the side wall.

of the particle. A third weak velocity jump also occurs when the rarefaction wave reects o the metal{liquid interface at the leading edge of the particle. Finally, a sharp velocity drop takes place when the particle is subjected to the wave reected o the second particle. The process is similar for successive particles with the exception that the mean particle velocity slightly increases as the shock wave penetrates the cluster. When the gap between the particles is reduced, the rarefaction wave does not have su cient time to propagate in the particle before the arrival of the wave that is reected o the downstream particle. In this case, oscillations become asymmetric and the mean particle velocity is reduced and becomes closer to that behind the transmitted shock. (b) A matrix cluster of particles The complexity of the shock{particle interaction process is further increased by additional waves produced by neighbouring particles in the lateral direction. Twodimensional calculations were conducted for a matrix cluster of 10 m m cylindrical aluminium particles subjected to a 101.3 kbar liquid shock. Although the results of
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

Shock interaction with solid particles in condensed matter


2000 velocity (m s- 1) 1500 1000 500 0 - 500 0 1 2 3 time (ns) 4 5

719

Figure 15. Velocity histories for the leading, centre and trailing edges of the leading particle of two cylindrical aluminium particles subjected to a 101.3 kbar liquid shock.

the cylindrical particle calculations cannot be directly applied to spherical particles, they do provide a qualitative trend for the eect of particle volume fraction on particle velocity. Figure 14 displays pressure contours for two particles with a 5 m m gap in the direction of motion. The eect of neighbouring particles in the lateral direction is simulated by introducing a wall at the plane of symmetry between the particles and the lateral particles with a 5 m m gap, that is, a 2.5 m m gap between the particles and the sidewall. It can be seen that, apart from the wave interaction described in the calculations for a linear cluster of particles, there are also additional shock waves reected o the lateral particles around the leading particle. These lateral shocks compete against the lateral expansion and prevent the local velocities from decaying inside the particle, as seen in gure 15. Collision of these lateral shocks produces a secondary pressure pulse to the leading particle, thus causing a gradual acceleration which is less signi cant than the deceleration eect from the wave that is reected o the downstream particle in the direction of motion. Figure 16 displays a high-loading matrix of cylindrical aluminium particles compacted by a 101.3 kbar liquid shock into a tight cluster of particles. The particles are largely deformed and coalesce due to the time lag between the acceleration of the leading and trailing particles. As shown in gure 17, due to the inuence of the neighbouring particles (particularly the downstream particles), the time in which the initially non-uniform velocity inside the particle reaches and oscillates around an average value behind the shock front is almost twice as much as the time required for the single particle case (see gure 3). It should be noted that geometric arrangement of the particles for a given particle volume fraction may result in dierent inuences. However, resonant eects that can be created in ordered systems may not appear in more realistic stochastic distributional ensemble systems (Baer 2002). Thus, the time required for achieving a relatively uniform velocity within a particle in a realistic system would be expected to be not much more than that for an ordered system. Considering the eects on the shock interaction from the neighbouring particles, the particle mass-centre velocity must be computed at a time longer than the shock interaction time for a single particle. To include the eect of the downstream particle, the particle mass-centre velocity for multiple particles, computed at a time equal to twice the shock interaction time , is summarized in gure 18. The results
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

720
3 ns

F. Zhang and others


7 ns 11 ns

Figure 16. Pressure contours for a matrix of cylindrical aluminium particles subjected to a 101.3 kbar liquid shock.

2000 velocity (m s- 1) 1500 1000 500 0 - 500 0 2 4 6 8 time (ns) 10 12 14

Figure 17. Velocity histories for the leading, centre and trailing edges of the central leading particle in a matrix of cylindrical aluminium particles subjected to a 101.3 kbar liquid shock.

indicate that the transmitted particle velocity decreases with an increase in particle volume fraction .

4. Simple models
(a) Single particles Under the assumption that the condensed explosive and the solid particles are chemically frozen during the particle crossing the shock front, the numerical studies for a single particle suggest that the velocity transmission factor , de ned in equation (2.5), strongly depends on the initial density ratio of host condensed matter to particle. The inuence on from other parameters, such as particle acoustic impedance, shock strength and unreacted explosives shock Hugoniots, is relatively
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

Shock interaction with solid particles in condensed matter


1.0 linear model 2 cylinders 11 cylinders

721

0.9

0.8

0.7

0.6

0.5

0.2

0.4

0.6

0.8

1.0

Figure 18. Particle mass-centre velocities for various matrices of aluminium particles.

2 s3 1 0 s0

s4

host matter (H)

particle (S)

Figure 19. Wave diagram of the one-dimensional model.

weak or could be implicitly included in the eect of the initial density ratio. A curve t of the numerical values of suggests the following correlation, 1 0 0 ; (4.1) a+b = s 0 a+b s0

where a = 3:947 and b = 1:951 (see gure 12). To get more insight into the mechanism for shock interaction, a one-dimensional analytical model based on the wave dynamics is considered. For simplicity, it is assumed that the impedance matching of the host condensed matter and the particle material results in a reected shock in the host matter and a transmitted shock in the particle as shown in gure 19. For the solid particle, the mass and momentum conservation equation across the shock, along with the linear c{s Hugoniot relation, Ds ps ps
0

us

= cs 0 + ss 0 (us
3

us 0 ); us 0 )]:

(4.2)

yield a relation between transmitted pressure ps


3 0

and ow velocity us 3 , (4.3)

s 0 (us 3

us 0 )[cs 0 + ss 0 (us 3

Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

722

F. Zhang and others

In equations (4.2) and (4.3), the plus sign is for a right-crossing shock and the minus sign for a left-crossing shock. The variables D s , us , ps , and s are the shock velocity, ow velocity, pressure and density for the particle. The subscript `0 represents the particle initial state and `3 is for the transmitted shock state. For the host condensed matter, the mass, momentum and energy conservation equations crossing the reected shock are used, and they result in p (4.4) u2 = u1 (p2 p1 )(1= 1 1= 2 );

e2 = e1 + 1 (p2 + p1 )(1= 2

1= 2 );

(4.5)

where the plus sign is for a right-crossing shock and the minus sign for a left-crossing shock. The variables u, p, and e denote the ow velocity, pressure, density and the speci c internal energy for the host matter. The subscript `2 denotes the reected shock state and `1 is for the incident shock state. The Murnaghan equation of state (2.1) must also be referred to the incident shock state, i.e. 1 n K1 1 + p1 (4.6) p= 1 n1

with K1 = 1 c2 = 1 c2 ( 1 = 0 )(n0 1) ; 1 0 ln(c1 =c0 ) : n1 = 1 + 2 ln( 1 = 0 ) (4.7)

(4.8)

From (4.6), the speci c internal energy can be obtained: e= p (n1 p1 1 c2 ) 1 : (n1 1)

(4.9)

Applying equation (4.9) to (4.5) results in a Hugoniot = (p), which is substituted into (4.4) to yield a relation u2 = u(p2 ). Solving the host matter relation u2 = u(p2 ) and that for the particle (4.3) along with the interfacial conditions ps
3

= p2 ;

us

= u2

(4.10)

results in the transmitted velocity us 3 =1 q s = u1 u2 [ 1 (n + 1)p


1 1 2

ps
s 3

p1

1 (n1 2

; 1)p1 (n1 p1
3

(4.11)

1 c2 )] 1 (4.12)

in which

ps

= ps

s 0 (us 3

us 0 )(cs 0 + ss 0 (us

us 0 )):

As the shock in the surrounding condensed matter with a velocity D0 reaches the particle trailing edge, the transmitted shock with a velocity Ds 0 still propagates in the particle if Ds 0 < D0 , or it is already reected o the particle trailing edge and the reected rarefaction runs back into the particle if D s 0 > D0 . Taking into account the shock velocity dierence, a momentum balance rule is assumed at the shock interaction time, i.e. s us d = s us d =
Proc. R. Soc. Lond. A (2003)
s 3 us 3 d 3 ; s 3 us 3 (d

d4 ) +

s 4 us 4 d4 ;

if Ds if Ds

0 0

6 D0 ; > D0 ;

(4.13) (4.14)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

Shock interaction with solid particles in condensed matter


D0 Ds0 us3 d3 d Ds0 -< D0 D0 cs3 - us3 us3 us0 = 0 D0

723

us4 D0 d4

d D s0 > D0 Figure 20. One-dimensional model.

where the geometric quantities are illustrated in gure 20. The subscript `s denotes the mass-averaged state at the shock interaction time , the subscripts `3 and `4 represent the transmitted shock and the reected rarefaction state, respectively. The distances d3 and d4 are de ned by the shock interaction time =

d4 d d3 d : + = = cs 3 us 3 Ds 0 Ds 0 D0

(4.15)

The ow velocity us 4 behind the rarefaction wave can be analytically obtained. However, for simplicity, us 4 2us 3 , s s 3 s 4 and also cs 3 cs 0 are assumed. Substitution of equation (4.15) into equations (4.13) and (4.14) yields = Ds 0 s; D 0 cs 0 us 3 Ds 0 = 1+ D0 Ds 0

for Ds

6 D0 ;

(4.16)

for Ds

> D0 ;

(4.17)

where is de ned in equation (2.5) and s is given in equation (4.11). The theoretical velocity transmission factor obtained from equations (4.16) and (4.17) lies between state `s3 behind the transmitted shock and state `s4 behind the rarefaction. In gure 12, comparison of the theoretical results from equations (4.16) and (4.17) with the numerical calculations subjected to a 101.3 kbar shock shows a fairly good agreement except that the theoretical values are generally larger than the numerical ones, since the one-dimensional theory does not consider the loss to
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

724
9

F. Zhang and others

detonation velocity (km s- 1)

RDX RDX/inert Al RDX/react. Al 0 10 20 30 40 50

Al mass % in RDX/Al Figure 21. Equilibrium Chapman{Jouguet detonation velocities for RDX with aluminium additive.

be caused by lateral deformation and expansion. Note that under conditions of very strong shocks, the shocked host matter density can become larger than the particle material density. In this case, the one-dimensional theory does not take the lateral compression into account and can therefore underpredict the numerical results. (b) Multiple particles For shock interaction with multiple particles immersed in a condensed matter, the particle mass-centre velocity calculated varies with the selection of the time for the shock interaction over multiple particles. It may also vary with the geometric arrangement of the particles for a given particle volume fraction. For complex cases involved in multiple particles, it may simply assume a linear model for the velocity transmission factor between the value for a single particle (i.e. the particle volume fraction = 0) and the value for the transmitted shock in the solid s (i.e. = 1), i.e. m = (1 ) + s : (4.18) In equation (4.18), can be obtained from equation (4.1) or (4.16) and (4.17), and s can be obtained from equation (4.11). The results predicted using equation (4.18) for multiple aluminium particles are displayed in gure 18 and their variation trend in terms of particle volume fraction is in good agreement with the numerical calculations.

5. Final remarks
For a charge of condensed explosive with metal particles, the present study indicates that the momentum transfer from condensed explosive to heavy-metal particles such as tungsten and uranium is insigni cant during the particle crossing of the shock front. However, the momentum transferred to light-metal particles is signi cant, and
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

Shock interaction with solid particles in condensed matter

725

the particle velocity for aluminium, beryllium and magnesium achieves 60{100% of the value of the shocked explosives velocity. The particle velocity after the shock interaction strongly depends on the initial density ratio of explosive to metal. The inuence of other parameters, such as particle acoustic impedance, shock strength and bulk unreacted explosives shock Hugoniot, can be implicitly included in the eect of the initial density ratio. The transmitted particle velocity decreases with increase in the particle volume fraction. A theoretical model is also established to gain the insights into the main shock interactions that dominate the momentum transfer. The signi cant momentum transfer during a particle crossing the shock front must be taken into account when modelling the shock initiation and detonation structure for two-phase mixtures of condensed explosive and light-metal particles. Its importance can be illustrated with the equilibrium Chapman{Jouguet detonation for RDX with molecular aluminium with a mixture density of 1.66 g cm3 using the Cheetah code (Fried et al . 1998), and the results are shown in gure 21. Detonation velocity de cit is generally predicted to increase with the mass of the particle additive, regardless of the reactive or chemically frozen nature of the particles. For chemically frozen particles, the considerable velocity de cit is originated in the momentum and heat transferred to the particles up to their equilibrated state. For reactive particles, apart from a negative oxygen balance and changes in the explosives detonation products, the inuence of the momentum and heat transferred to the rest of the particles and the particle combustion products still remains. Unlike the metallic molecules, metallic particles may not be fully equilibrated with the detonation products of the explosive within the detonation wave. However, for light small particles, both the signi cant momentum transfer during the shock{particle interaction and that occurring behind the shock front play an important role in the detonation velocity de cit.
This work was partly supported under the auspices of the DND contract W7708-8-R748. The authors thank Dr Alexander Gonor for his valuable suggestions and comments.

References
Baer, M. R. 2002 Modeling heterogeneous energetic materials at the mesoscale. Thermochim. Acta 384, 351{367. Baer, M. R. & Nunziato, J. W. 1986 A two-phase mixture theory for the de agration-todetonation interaction (DDT) in reactive granular materials. Int. J. Multiph. Flow 12, 861{ 889. Baudin, G., Lefrancois, A., Bergues, D., Bigot, J. & Champion, Y. 1998 Combustion of nanophase aluminium in the detonation products of nitromethane. In 11th Int. Detonation Symp. ONR 33300-5, pp. 989{997. Arlington: O ce of Naval Research. Book, D. L. & Fry, M. A. 1984 Air blast simulations using ux corrected transport codes. NRL Memorandum 5334. Washington, DC: Naval Research Laboratory. Boris, J. P. 1976 Flux corrected transport modules for solving generalized continuity equations. NRL Memorandum 3237. Washington, DC: Naval Research Laboratory. Campbell, A. W., Davis, W. L., Ramsay, J. B. & Travis, J. R. 1961 Shock initiation of solid explosives. Phys. Fluids 4, 511{521. Dremin, A. N. 1992 Shock discontinuity zone e ect: the main factor in the explosive s decomposition detonation process. Phil. Trans. R. Soc. Lond. A 339, 355{364.
Proc. R. Soc. Lond. A (2003)

Downloaded from rspa.royalsocietypublishing.org on October 4, 2010

726

F. Zhang and others

Dremin, A. N. & Klimenko, V. Yu. 1981 On the e ect of shock wave front on the reaction origin. In Progress in astronautics and aeronautics (ed. J. R. Bowen, N. Manson, A. K. Oppenheim & R. I. Soloukhin), vol. 75, pp. 253{268. New York: AIAA. Dremin, A. N., Savrov, S. D., Tromov, V. S. & Shvedov, K. K. 1970 Detonation waves in condensed media. Moscow: Nauka. (Translated by Foreign Technology Division, Wright{Patterson Air Force Base, FTD-HT-23-1889-71.) Fried, L. E., Howard, W. M. & Souers, P. C. 1998 CHEETAH 2.0 users manual. UCRL-MA117541. Livermore, CA: Lawrence Livermore National Laboratory. Gelfand, B. E., Frolov, S. M. & Nettleton, M. A. 1991 Gaseous detonations|a selective review. Prog. Energy Combust. Sci. 17, 327{371. Gogulya, M. F., Dolgoborodov, A. Yu., Brazhnikov, M. A. & Baudin, G. 1998 Detonation waves in HMX/Al mixtures. In 11th Int. Detonation Symp. ONR 33300{5, pp. 979{988. Arlington: O ce of Naval Research. Hallquist, J. O. 1983 Theoretical manual for DYNA3D. UCID-19401. Livermore: Lawrence Livermore National Laboratory. Haskins, P. J., Cook, M. D. & Briggs, R. I. 2002 The e ect of additives on the detonation characteristics of a liquid explosive. In Shock Compression of Condensed Matter, 2001, pp. 890{893 (ed. M. D. Furnish, N. N. Thadhani & Y. Horie). Melville: American Institute of Physics. Lee, D. & Sichel, M. 1986 The Chapman{Jouguet condition and structure of detonations in dust{oxidizer mixtures. In Progress in astronautics and aeronautics (ed. J. R. Bowen, J. C. Leyer, & R. I. Soloukhin), vol. 106, pp. 505{521. New York: AIAA. Lee, E. L. & Tarver, C. M. 1980 Phenomenological model of shock initiation in heterogeneous explosives. Phys. Fluids 23, 2367{2372. Mader, C. 1998 Numerical modeling of explosives and propellants, pp. 309{311; 353{356. Boca Raton, FL: CRC Press. Marsh, S. P. 1980 LASL shock Hugoniot data, p. 625. Berkeley and Los Angeles: University of California Press. Schlichting, H. 1979 Boundary-layer theory, pp. 12{19. McGraw-Hill. Thibault, P. A. & Moen, K. 1997 IFSAS user manual. Version 1.6. Halifax: Combustion Dynamics Ltd. Veyssiere, B. & Ingignoli, W. 2002 Existence and characteristics of the detonation cellular structure in hybrid two-phase mixtures. Shock Waves 12. (In the press.) Walker, F. E. 1988 The initiation and detonation of explosives|an alternative concept. UCRL53860. Livermore: Lawrence Livermore National Laboratory. Walsh, J. M. & Christian, R. H. 1955 Equation of state of metals from shock wave measurements. Phys. Rev. 97, 1554{1556. Zeldovich, Ia. B. & Kompaneets, A. S. 1960 Theory of detonation, pp. 133{191. Academic. Zhang, F. & Lee, J. H. S. 1994 Friction-induced oscillatory behaviour of one-dimensional detonations. Proc. R. Soc. Lond. A 446, 87{105. Zhang, F., Greilich, P. & Gronig, H. 1992 Propagation mechanism of dust detonations. Shock Waves 2, 81{88. Zhang, F., Yoshinaka, A., Murray, S. B. & Higgins, A. 2002 Shock initiation and detonability of isopropyl nitrate. In 12th Int. Detonation Symp., 11{16 August, San Diego, CA. Arlington: O ce of Naval Research. (In the press.)

Proc. R. Soc. Lond. A (2003)

Anda mungkin juga menyukai