ADA553075

AFRL-RW-EG-TR-2011-159
BASIC DETONATION PHYSICS ALGORITHMS
Douglas V. Nance
AFRL/RWPC
101 W. Eglin Blvd.
Eglin AFB, FL 32542-6810
December 2011
INTERIM REPORT
DISTRIBUTION A. Approved for public release, distribution unlimited. 96th ABW/PA

Approval and Clearance # 96ABW-2011-0548 dated 28 November 2011.
AIR FORCE RESEARCH LABORATORY

MUNITIONS DIRECTORATE
 Air Force Materiel Command  United States Air Force  Eglin Air Force Base, FL 32542
Force MaterialCommand
NOTICE AND SIGNATURE PAGE
Using Government drawings, specifications, or other data included in this document for any purpose
other than Government procurement does not in any obligate the U.S. Government. The fact that the
Government formulated or supplied the drawings, specifications, or other data does not license the
holder or any other person or corporation, or convey any rights or permission to manufacture, use, or
sell any patented invention that may relate to them.
This report was cleared for public release by the 96th Air Base Wing, Public Affairs Office, and is
available to the general public, including foreign nationals. Copies may be obtained from the Defense
Technical Information Center (DTIC) < http://www.dtic.mil/dtic/index/html>.
AFRL-RW-EG-TR-2011- 159 HAS BEEN REVIEWED AND IS APPROVED FOR PUBLICATION

IN ACCORDANCE WITH ASSIGNED DISTRIBUTION STATEMENT.
FOR THE DIRECTOR:
ORIGINAL SIGNED ORIGINAL SIGNED

______________________________________ _____________________________________
Craig M. Ewing, DR-IV, PhD Douglas V. Nance
Technical Adviser Program Manager
Strategic Planning and Assessment Division
This report is published in the interest of scientific and technical information exchange, and its
publication does not constitute the Government’s approval or disapproval of its ideas or findings.
Form Approved
REPORT DOCUMENTATION PAGE OMB No. 0704-0188
The public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources,
gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of
information, including suggestions for reducing the burden, to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and Reports (0704-0188),
1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to any
penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number.
PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS.
1. REPORT DATE (DD-MM-YYYY) 2. REPORT TYPE 3. DATES COVERED (From - To)
01-12-2011 INTERIM 01-10-2011 - 31-10-2011
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
Basic Detonation Physics Algorithms N/A
5b. GRANT NUMBER
N/A
5c. PROGRAM ELEMENT NUMBER
62602F
6. AUTHOR(S) 5d. PROJECT NUMBER
Douglas V. Nance 2502
5e. TASK NUMBER
67
5f. WORK UNIT NUMBER
63
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION
REPORT NUMBER
AFRL/RWPC
101 W. Eglin Blvd. AFRL-RW-EG-TR-2011-159
Eglin AFB, FL 32542-6810
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR'S ACRONYM(S)

AFRL/RWPC AFRL-RW-EG
101 W. Eglin Blvd.
Eglin AFB, FL 32542-6810 11. SPONSOR/MONITOR'S REPORT
NUMBER(S)
12. DISTRIBUTION/AVAILABILITY STATEMENT
Distribution A: Approved for public release, distribution unlimited. (96ABW-2011-0548)
28 Nov 11
13. SUPPLEMENTARY NOTES

NONE
14. ABSTRACT
This report presents the theory behind a series of detonation physics algorithms used to simulate the detonation of a condensed
explosive. The numerical scheme implemented in this case is the Roe flux difference splitting scheme due to Glaister. This report
contains a detailed discussion of the mathematical derivation along with a printing of the source code. We also include a discussion
of methods for implementing Lagrangian tracking algorithms for solid inclusions within the condensed explosive.
15. SUBJECT TERMS

detonation, explosive, flux, jacobian
16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF RESPONSIBLE PERSON
a. REPORT b. ABSTRACT c. THIS PAGE ABSTRACT OF Douglas V. Nance
PAGES
UL 19b. TELEPHONE NUMBER (Include area code)
UNCLAS UNCLAS UNCLAS 102
Standard Form 298 (Rev. 8/98)
Reset Prescribed by ANSI Std. Z39.18
TABLE OF CONTENTS
Section Page
1 INTRODUCTION 1
1.0 Numerical Detonation Physics 1

1.1 A Map for this Report 2
2 GOVERNING EQUATIONS 4
2.1 The Reactive Euler Equations 4

2.2 Mixture Equations of State 5
2.3 Solid Explosive Equations of State 6
2.4 Detonation Products Equation of State 8
3 SYSTEM EIGEN-STRUCTURE 10
3.1 Flux Jacobian Matrices 10

3.2 Eigenvalues 12
3.3 Eigenvectors 13
4 BUILDING THE NUMERICAL SCHEME 18
4.1 Pressure Derivatives 18

4.2 Finite Volume Discretization 21
4.3 Temporal Discretization 22
4.4 The Numerical Flux 23
4.5 A Higher Order Scheme 25
4.6 Boundary Conditions 27
5 PARTICLE MOTION 28
5.1 Coupling Terms 28

5.2 Particle Laws of Motion 29
6 RESULTS 32
6.1 Simple Plane Wave Detonation 32

6.2 Detonation of Pure HMX 35
6.3 Detonation of HMX Containing Metal Particles 37
Distribution A. Approved for public release, distribution unlimited. (96ABW-2011-0548)

i
7 CONCLUSIONS 39
8 RECOMMENDATIONS 39
REFERENCES 40
APP. A SOURCE CODE 42

ii
LIST OF FIGURES
Figure Page
1 Interface Notation 23
2 Problem 1 Detonation Field Density, Time 3.0 33
3 Problem 1 Detonation Field Velocity, Time 3.0 33
4 Problem 1 Detonation Field Pressure, Time 3.0 34
5 Problem 1 Detonation Field Reaction Progress Variable, Time 3.0 34
6 Numerical Detonation Solution Hayes-I/JWL in HMX at 3 μs. Horizontal Axis

is Distance in Meters 36
7 Numerical Detonation Solution Hayes-II/JWL in HMX at 3 μs. Horizontal Axis

is Distance in Meters 36
8 Radial Locations for Steel Particles Embedded in a Mass of Detonating HMX 37
9 Radial Velocities for Steel Particles Embedded in a Detonating Mass of HMX 38

iii
1 INTRODUCTION
Steady increases in large scale circuit integration indicate that the Twenty-First
Century will promise significant advances in High Performance Computing (HPC)
machinery. Today, one may obtain desk-side Linux systems containing eight processors
(and thirty-two or more cores) for comparatively reasonable prices. Moreover, common
laptop systems wield significant computing power with central processing unit (CPU)
speeds in the neighborhood of 3.0 GHz (maybe more by the time this report is certified)
and random access memory (RAM) storage capability in hundreds of Gigabytes (GB). In
the realm of “Big Iron”, the Department of Defense (DoD) High Performance Computing
(HPC) Modernization Office recently began operating clusters each with tens of
thousands of cores, and the Department of Energy laboratory community has even larger
systems. These developments have significant implications for the relatively small
Computational Physics research community. This research community represented by
disciplines such as high energy physics, quantum chemistry and computational fluid
dynamics has an ever increasing need for computer memory and for parallel processing
speed.
Computational Fluid Dynamics (CFD) has drawn on HPC resources for many
years to help with aircraft and fluid system design. Some problems like high Reynolds
number direct numerical simulations are still computationally inaccessible, but these
situations are fewer in number than just one decade ago. For instance, we routinely solve
problems involving the large eddy simulation (LES) of compressible turbulence with
good results. Older techniques such as Reynolds-Averaged Navier-Stokes (RANS)
simulation now teeter on the brink of obsolescence. Moreover, massive computing power
now permits us to invade new territory previously relegated to analytical solutions
supported by many assumptions and highly simplified, under-resolved computational
studies. Quantum physics now benefits widely from HPC science in the areas of quantum
chemistry and molecular dynamics. These areas of physics now impact design
engineering. Although it occupies only a very small part of the research community,
detonation physics, a close relative of CFD, can benefit handsomely from ever more
powerful computational techniques and equipment.
1.0 Numerical Detonation Physics
Numerical Detonation Physics applies many of the same computational

techniques employed by CFD. The primary reason is because detonations are powered by
the propagation of the detonation wave, a powerful shock wave that transforms the
unreacted explosive into detonation product species. Like the shock waves encountered in
transonic and supersonic flow, detonation waves must be “captured” in the material field
by using special numerical techniques. Gas phase detonations, e.g., the explosive burn of
acetylene gas, are true detonations but they lack some of the complexity associated with
the detonation of condensed (solid or liquid) explosives. Gas phase detonation is usually
initiated by high temperature. It follows that temperature is the dominant term in the
reaction rate expression. One should also not make light of the fact that we actually have

good, quantitative models for gas phase detonation chemistry. The science behind the
detonation of condensed explosives is not so evolved.
The detonation of a condensed explosive is most often modeled as a shock-driven

process. Macroscopic observation seems to indicate that a shock wave is often required to
detonate these explosives. Many solid explosives simply “burn” when exposed to a
flame, at least when considered over relatively short time periods. Exposure to a shock
impulse is often needed to initiate the run to detonation for an explosive. This physics
problem is complicated greatly because of the smallness of scales concerning the
detonation wave. The detonation wave covers a thin region, a fraction of a millimeter for
most ideal or Carbon-Hydrogen-Nitrogen-Oxygen (CHNO) explosives like Trinitro-
toluene (TNT). The head of the detonation wave lies at the entrance to the detonation
reaction zone. This is the tiny region in space where the detonation chemical reactions
take place. For condensed explosives, we do not know these chemical reactions. We
know only, in some sense, their end products, and if we detonate two like samples of an
explosive, we may obtain two different product spectrums. For this reason, condensed
explosives are relatively crude chemical mixtures. Still, the detonation process itself may
be addressed by the direct application of the conservation laws for mass, momentum and
energy. This same approach is used for CFD problems, but for explosives we are required
to apply equations of state for both the unreacted explosive material and the detonation
products. It is also important that we consider heterogeneous explosives. These materials
contain non-explosive additives like plastic binders and metal particles. In future
treatments of this problem, we will also be required to treat the material behavior
(material strength versus applied stress) of the solid explosive in response to shock
excitation.
1.1 A Map for this Report
This report is intended to assist in the process of transitioning detonation physics

algorithms into the Large Eddy Simulation with LInear Eddy Modeling in 3 Dimensions
(LESLIE3D) multiphase physics computer program. The discussions that follow describe
the algorithms applied in the source code included in Appendix A. Although these
algorithms are tested and validated to some extent, it is nont recommended that they be
coded directly into LESLIE3D. Rather, the Harten, Lax and van Leer (HLL) family of
algorithms should be used for flux difference splitting in lieu of Roe’s method. Moreover,
inhomogeneous terms in the equations should be addressed through Strang splitting.1
The report is organized as follows. In Section 2, we describe the governing

equations for the detonation problem based upon the work of Xu et al.2 Within this set of
equations, we add the terms coupling the detonation flow field to the particle field. We
show that reaction rate, particle coupling and geometric effects may be incorporated as
source terms. The equations of state used for the solid explosive and for the detonation
products are also presented in this section. The advective terms, of critical importance in
the shock-capturing scheme, are clearly delineated. Section 3 describes the eigen-
structure for the system of governing equations. The flux Jacobian matrix is developed
2
for the reactive Euler equations adapted for a real gas equation of state. Then we develop
a set of eigenvalues and eigenvectors needed in order to accurately capture the detonation
wave. In Section 4, we discuss the overall numerical scheme and temporal discretization
procedure used in our detonation computer program. We also discuss the development of
the numerical flux vector in detail. Section 5 contains the terms governing the motion of
Lagrangian particles including the drag laws. In Section 6, we provide the results for
three example calculations. After performing a calculation to verify proper code
performance, we simulate the detonation of a spherical mass of HMX loaded with metal
particles. We show a series of detonation waveforms for this explosive, and we go on to
include the resulting particle trajectories and velocities. We also make some basic
comparisons between the results produced by our computer program to archival
explosive performance data for HMX. Finally, in Section 7, we draw several important
conclusions from our development. We also make recommendations for follow-on work
needed to support the installation of detonation physics algorithms in LESLIE3D.

3
2 GOVERNING EQUATIONS
To address the detonation problem, we follow a body of research documented in

the general scientific literature.2 By doing so, we can escape some of the uncertainties
associated with the older programmed burn detonation models.3 We do make a departure
from the core reference in that our development disregards the issue of compaction in the
solid explosive.2 Instead, it is assumed that our explosive is a solid mass at or near the
theoretical maximum density. The present approach allows the reaction zone to be clearly
resolved within the limitations of the grid refinement. As a result, the forces applied to
particles may be resolved more accurately.
2.1 The Reactive Euler Equations
The reactive Euler equations are frequently used to represent detonation flow
fields based upon a reaction progress equation and a mixture equation of state.2 The
equations for the conservation of mass, momentum, energy and reaction progress may be
readily expressed in vector form. The equation for a detonation field set in one space
dimension may be written as
U F
  S G  S Rx  S P (2.1.1)
t x
where
U  [  ,  u, E,   ]T (2.1.2)
is the vector of conserved variables, and
F  [  u,  u 2  P, u( E  P),  u  ]T
(2.1.3)
is the flux vector. Also,
j
SG   [  u,  u 2 , u ( E  P),  u  ]T (2.1.4)
x
S Rx  [0, 0, 0,  r]T (2.1.5)
S P  [0, Fs , Q s , 0]T (2.1.6)
We may also write the total energy per unit volume as

E  e u2 (2.1.7)
2
4
where e is the internal energy per unit mass. The equation of state may be written in the
general form
P  P(  , e,  ) (2.1.8)
where  is the reaction progress variable.
Vectors S G , S Rx and S P contain source terms; as we have shown, these

nonhomogenous terms are kept on the right hand side of the reactive Euler equations and
may be treated independently from the advective terms. Vector S G contains the
geometric source terms that allow the system to be configured for planar, cylindrical or
spherical one-dimensional flow. To adapt (2.1.1) for planar flow, we need only set j  0
in (2.1.4). We may adapt (2.1.1) for cylindrical or spherical one-dimensional flow by
setting j  1 or j  2 , respectively. Vector S Rx contains the reaction rate source term
governing the rate of progress for the detonation reaction. The reaction rate r may be
written in many different forms depending on the explosive.4 The term we have chosen to
use for HMX may be written as
N
 P 
r  k   (1  ) (2.1.9)
 PCJ 
where PCJ is the Chapman-Jouquet pressure for HMX; k , N and  are constants
chosen to fit experimental data.5 Note that this reaction rate law is dependent upon both
pressure and reaction progress. The source term vector S P has been added to the system
by the author. It represents the dynamic coupling between the detonation products and a
field of discrete, massive Lagrangian particles. The coupling is based upon both
momentum and thermal effects.6 The specific forms of the coupling terms are presented
in a later section.
2.2 Mixture Equations of State
For the detonation problem, relevant equations of state are cast in the form of
(2.1.8). This form is complicated since pressure varies as a function of density, internal
energy per unit mass and reaction progress. In this analysis, the reaction progress variable
is analogous to a species mass fraction commonly used in reacting gas flows. Moreover,
it is used to compute the specific internal energy for the detonating mixture by forming a
weighted sum of the equation of state (EOS) for the solid explosive and the EOS for the
detonation products. The resulting expression for specific internal energy is called the
mixture EOS.2 Our governing equations (2.1.1), discretized in accordance with the finite
volume method, rely upon the mixed cell approach. Each flow cell is assumed to contain
a mixture – part solid explosive and part detonation products. The mixture fraction is
given by the reaction progress variable  , and  is defined as the mass fraction of the

5
detonation products in the cell. The density within a cell is the sum of the densities for the
solid ( s ) and gas ( g ) phases, respectively, i.e.,
  s   g (2.2.1)
so  is given by
g
 (2.2.2)

and
s
 1  (2.2.3)

Hence, we have that  is the mass fraction of the gas (detonation products) phase. We
also assert that the internal energy for a given finite volume cell may be expressed as
e   eg  (1  ) es (2.2.5)
where e g and es are the specific internal energies for the gas and solid phases,
respectively. This mixing rule differs from the archived approach based upon specific
volume, but to date, we have not been successful in applying Xu’s closure.7 Assume the
same pressure for both phases with each phase having its own equation of state, i.e.,
eg  e g (  g , P ) (2.2.6)
es  es (  s , P) (2.2.7)
with  g and  s given by (2.2.2) and (2.2.3).
2.3 Solid Explosive Equations of State
In the previous section, we showed that one part of our mixture EOS represents
the solid explosive. In the discussions that follow, we apply two different forms of an
EOS originally developed by Hayes.8 The first form of this EOS (Hayes-I) works very
well for mechanical effects.2 The Hayes-I EOS is given as
P  P0  P0   s 0  
 s 
N 1
 s0   
es (  s , P)    t3  1
 
  t 
4 

  ( N  1)1   1 (2.3.1)
  s 0  s     s  
 s 0 
g

where

6
g  0  s 0 (2.3.2)
Cvs T0 g
t3  (2.3.3)
s0
H1
t4  (2.3.4)
 s 0 N ( N  1)
In equations (2.3.1) through (2.3.4), P0 , T0 and  s 0 are the ambient pressure,

temperature and unloaded solid density. 0 is the Gruneisen parameter, and Cvs is the
constant volume specific heat for the solid. H 1 and N are parameters used to fit the EOS
to data. Table 1 lists all of the required parameters for this EOS.2
Table 1 - Hayes EOS Data for HMX

H1 1.3 x 1010 N/m2
N 9.8
Cvs 1.5 x 103 J/(Kg K)
Γ0 1.105
P0 101325 Pa
ρs0 1.9 x 103 Kg/m3
T0 300 K
The second form of the Hayes EOS (Hayes-II) functions well mechanically but
also incorporates temperature. The Hayes-II EOS is given as
1 H 1   s   
N
P0   s 0 
es (  s , P)   P  P0  
 
  1    t3  1 
g N   s 0      s 0   s 
  
(2.3.5)
   N 1    
 t 4  s   ( N  1)1 s 0   1
  s 0   s  

This version of the Hayes EOS may be derived by using Reference 1; however,
additional terms are incorporated in (2.3.5) to match the behavior of (2.3.1) at ambient
pressure. The temperature of the solid explosive is given by
1  H1   

N
  s 
T (  s , P)  P  P0     1   T0 (2.3.6)
t3  N   s 0  
  

7
Together, equations (2.3.5) and (2.3.6) constitute a complete equation of state for a solid
explosive.9 These equations use the same data as is listed in Table 1 for HMX. The
Hayes-II EOS also performs very well in one-dimensional detonation studies for solid
HMX.
2.4 Detonation Products Equation of State
As equation (2.2.5) indicates, part of the mixture EOS must address the gaseous
products resulting from the detonation of the solid explosive. For the purposes of this
work, we have selected the Jones-Wilkins-Lee (JWL) EOS.1 The JWL EOS is somewhat
controversial, but nevertheless, it is widely applied in hydrocodes. Also, many explosives
have been characterized for this EOS. We apply the JWL EOS in the following form.
1    g   Rˆ    g   Rˆ 
eg (  g , P )   P  A1  exp   1   B1  exp   2   Q  e0 (2.4.1)
  g  Rˆ1      Rˆ 2    
   g     g 
where A , B ,  , R̂1 and R̂2 are coefficients produced by curve-fitting for the explosive
under consideration. Also, note that
Rˆ1  R1  s 0 , (2.4.2)
and
Rˆ 2  R2  s 0 . (2.4.3)
Q is the heat of detonation for the explosive, and e0 is the reference value for specific
internal energy. There is no firm rule for determining e0 , but we will define e0 as
e0  Cvg T0 . (2.4.4)
Table 2 - JWL Coefficients for HMX

R1 4.2
R2 1.0
ω 0.3
A 7.783 x 1011 Pa
B 7.071 x 1010 Pa
Cvg (1.1 – 0.28x10-3 ρs0) x 103 J/(Kg K)
[7.91 – 4.33 (10-3ρs0 -1.3)2 - 0.934 (10-3ρs0 -1.3)]
Q
x 106 J
Cvg is the constant volume specific heat for the detonation products. The data used for
HMX in the JWL EOS is listed in Table 2.2 For the studies performed later in this work,
8
we select one of the Hayes equations of state in combination with the JWL EOS to form a
mixture EOS.

9
3 SYSTEM EIGEN-STRUCTURE
3.1 Flux Jacobian Matrices
Capturing the structure of the detonation wave constitutes a difficult numerical

issue involving the discretization of the advective term F , where
U
 F1 F1 F1 F1 

  (  u) E  
 
 F2 F2 F2 F2 
F   (  u) E  
A  (3.1.1)
U  F3 F3 F3 F3 

  (  u) E  
 F4 F4 F4 F4 
  (  u) E  

is called the flux Jacobian matrix. The term Fi simply denotes the ith element of the flux
vector F . Equation (3.1.1) is already annotated with the specific elements of U . It is
important to note that our equation of state is cast in a general form, so the calculation of
the specific elements of (3.1.1) is made more complicated. The method for calculating
these matrix entries relies heavily on the derivatives of pressure taken with respect to the
conservative variables.10 For convenience, the pressure derivatives for this Jacobian are
given below. For the three-dimensional case, the detailed derivation of these pressure
derivatives is presented in Reference 11. For pressure given in the form of (2.1.8), let
 P   P   P 
P    ; Pe    ; P    (3.1.2)
   e,  e   ,     ,e
then we may write the pressure derivatives as
 P   u2 E  
   P  Pe   2   P (3.1.3)
    u , E ,     
 P  u
    Pe (3.1.4)
 (  u )   , E ,   
 P  P
   e (3.1.5)
 E   ,  u ,   

10
 P  P
    (3.1.6)
 (   )   ,  u ,   
Clearly, the pressure derivatives taken with respect to the conservative variables depend
on the pressure derivatives defined in (3.1.2). These derivatives, in turn, depend on the
specific form of the equation of state (2.1.8). Accordingly, the derivation of the elements
of (3.1.1) is a complicated process not to be presented here. Instead, the reader is referred
to a work containing like, yet detailed, mathematical derivations.11 For completeness, the
flux Jacobian matrix for (2.1.1) is given below.
 0 1 0 0 
 2  P Pe P 
 a u   u  2  e 
2

    
A (3.1.7)
 2 u 2
 Pe  u 
u (a  H   ) H   Pe u1   P
 
  
  u   0 u 
where
EP
H (3.1.8)

Pe P
  (H  u2 )  (3.1.9)
 
and the frozen speed of sound, a , is given by
P Pe
a 2  P  . (3.1.10)
2
The derivation for this speed of sound is also archived.11
We can also define a vector of non-conservative variables for the reactive Euler
equations as V , where
V  [  , u, P,  ]T . (3.1.11)
As you may surmise, the governing equations may also be written in terms of the non-
conservative variables, and we may define a non-conservative flux Jacobian matrix Â
such that11

11
u  0 0
0 1 
ˆ 
A
u
 0 (3.1.12)
0  a 2 u 0
 
0 0 0 u 
The derivation of the non-conservative reaction progress is a simple exercise. Observe

that the conservative form of this equation is written as
(   ) (   )
 r (3.1.13)
t x
We may expand (3.1.13) as follows.
  (  u )   
    u r (3.1.14)
 t x  t x
The first term in (3.1.14) vanishes since it is just a scalar multiple of the continuity
equation (component one of 2.1.1), so we obtain
 
u r (3.1.15)
t x
as the non-conservative reaction progress equation.
3.2 Eigenvalues
The eigenvalues of the flux Jacobian matrix contain important information on the
physics of our detonation problem. We think of any fluid mechanics problem (as well as
most solid mechanics problems) in terms of interacting waves. The detonation problem
can be decomposed into a set of characteristic waves.2 The speeds at which these waves
propagate are given by the eigenvalues of the flux Jacobian matrix.12 For any square
matrix A , the eigenvalues are defined as the set of numbers  such that
A  I  0 (3.2.1)
where I is the identity matrix. We may note that the conservative matrix (3.1.7) is
heavily populated, so it is very difficult to obtain the eigenvalues by using (3.2.1).
Fortunately, the non-conservative matrix (3.1.12) is a simpler form mathematically
equivalent to (3.1.7), so these matrices must have the same eigenvalues.11 Using (2.3.1),
the eigenvalues of (3.1.12) are easily shown to be
 {u  a, u, u, u  a} (3.2.2)

12
Note that u is an eigenvalue of multiplicity two, so there are two waves with speed u ,
i.e., the entropy and reaction progress waves both propagating at the flow velocity. The
remaining two distinct eigenvalues   u  a denote acoustic waves.12 The dynamics of
the detonation process may be described through the interactions of characteristic waves,
but to completely describe these waves, we must determine the eigenvectors for the
detonation problem.
3.3 Eigenvectors
In order to determine the characteristic waves for (2.1.1), we must determine the
eigenvectors for the conservative Jacobian matrix (3.1.7). When we use the term
eigenvector, in this case, we are referring to a right eigenvector.10
Definition: Given a matrix A C(n  n) with a set of eigenvalues  i  C , i  1,, n , we

define the right eigenvector ri  C(n) associated to the eigenvalue  i such that
A ri   i ri (3.3.1)
Equation (3.3.1) is useful in that it tells us how to find right eigenvectors. To find a right
eigenvector for (3.1.7) associated to an eigenvalue  , we first define the components of
right eigenvector r . Let
r  (1, 2 , 3 , 4 )T (3.3.2)
Now we apply (3.1.7) and (3.3.1) to create a linear system of equations in the
components of r .
 0 1 0 0 
 2  P  P P  1  1 
 a u   u  2  e 
2 e

       2   
 2
 (3.3.3)
 2 u2  Pe  u  3  3 
u (a  H   ) H   Pe u 1   P
   4   
    4 
  u   0 u 
The system (3.3.3) directly leads to a system of four eigenvector equations. The
eigenvector equations do not have a unique solution; in fact, they have an infinite number
of solutions, so care is required in structuring prospective choices for the components of
r to design a proper numerical treatment for the problem. Also, it is important to observe
that the number of linearly independent eigenvectors must be same as the order of the
system. For this detonation problem, the Jacobian matrix is of the fourth order, so we
must determine four linearly independent eigenvectors even though we have only three
distinct eigenvalues; the eigenvalue u is repeated.
We begin the process of determining some specific eigenvector components by

extracting the first eigenvector equation from (3.3.3), i.e.,

13
 2   1 (3.3.4)
We may satisfy equation (3.3.4) by choosing
1  1 ;  2   (3.3.5)
Equation (3.3.5) may be used in (3.3.3) to produce the remaining three eigenvector
equations
 u  P P
a 2  u 2     2u    Pe    e 3    0 (3.3.6)
    
 u2   u  u
u (a  H   )    H     i    Pe 3  P 4  0
2
 (3.3.7)
     
 u      (u   )4  0 (3.3.8)
Based upon (3.3.5), we may produce the eigenvector associated to eigenvalue   u . Set
  u in (3.3.8), and we see that this equation is trivially satisfied with no restrictions on
 4 . Now we set   u in (3.3.7) and (3.3.8); by simplifying, we can show that both of
these equations reduce to the same equation, i.e.,
u2 Pe P
a2    Pe  3  4  0 (3.3.9)
  
Since there are no restrictions on  4 , we may freely choose  4 and solve for  3 .
 a3 P
3  H   (   4 ) . (3.3.10)
Pe Pe
By cleverly choosing the value of  4 , we produce two linearly independent eigenvectors

associated to the eigenvalue   u . If we set  4  0 , we obtain the eigenvector
T
  a 2 Pe 
r  1, u, H    , 0  (3.3.11)
 Pe P 
Alternatively, we obtain a second eigenvector by setting  4  1 , so

T
  a 2 Pe 
r  1, u, H   (  1), 1 (3.3.12)
 Pe P 

14
We may also obtain the eigenvector associated to eigenvalue   u  a ; by
returning to equation (3.3.4), let us choose
1  1 ;  2  u  a (3.3.13)
By substituting (3.3.13) into (3.3.8), we may show that
4   (3.3.14)
We can produce another eigenvector equation associated with this eigenvalue by using
(3.3.14) and setting   u  a in (3.3.6). By doing so and solving for  3 , we have that
3  H  u a (3.3.15)
One may show that (3.3.13), (3.3.14) and (3.3.15) satisfy (3.3.7), and the eigenvector
associated to eigenvalue   u  a is
r  (1, u  a, H  u a,  )T (3.3.16)
We may derive the eigenvector associated to eigenvalue   u  a by the same

procedure. We consider (3.3.4) and then set
1  1 ;  2  u  a (3.3.17)
Equation (3.3.8) can be applied to again obtain the result (3.3.14). By substituting
(3.3.17) and (3.3.14) into (3.3.6), we can solve for  3 ,i.e.,
3  H  u a . (3.3.18)
Subsequently, one can show that (3.3.17), (3.3.18) and (3.3.14) satisfy equation (3.3.7).
Hence, the eigenvector associated to eigenvalue   u  a , may be written as
r  (1, u  a, H  u a,  )T (3.3.19)
Equations (3.3.11), (3.3.12), (3.3.18) and (3.3.19) are the eigenvectors for the reactive
Euler equations in one dimension. We can form R , the matrix of right eigenvectors, by
allowing each eigenvector to form a column of this matrix. Hence,
 1 1 1 1 
 ua u u u  a 

R  a 2 P  a 2 P (3.3.20)
H ua H    H  (  1) H  u a 
 Pe Pe Pe Pe 
   
 0 1

15
It is a straightforward although tedious exercise to show that R , the determinant of R ,
is
2  a3
R  . (3.3.21)
Pe
So far, our development of the eigen-structure for the reactive Euler equations closely
coincides with Glaister’s derivation performed for the real gas equation of state.10 From
(3.3.21), we can see that our eigenvectors are well-defined and constitute a non-singular
system for realistic values of density and the speed of sound with Pe  0 . As a result, R
is invertible under the same conditions, and we can calculate the matrix of left
eigenvectors L with L  R 1 , and by using the adjoint matrix for R (the transpose of
the matrix of cofactors) in conjunction with the definition of the inverse matrix, we have
that
  a P   a
 a H  u 2  (u  a)     a u  
  Pe Pe   Pe 
   
2a (1  )(u  H )  (  a  P (  1))  2u a(  1)
2 2
1   Pe 
L
R   2 1 
 2a   u  H  (  a 2   P )   2u a 
  Pe 
   a P   a
 a H  u 2  (u  a)     a u  
  Pe Pe   Pe 
P 
a a 
Pe

  a 2 P 
2a(1   ) 2a  (  1) 
 Pe Pe  (3.3.22)
 a 2
P  
2a  2a      
 Pe Pe 

a a
P 
Pe 
Each row of the matrix shown in (3.3.22) is a left eigenvector for the Jacobian matrix
found in (3.1.7).
Although we have not yet presented explicit forms for the pressure derivatives,
we have accomplished a great deal of work in this section. Equations (3.2.2), (3.3.20) and
(3.3.22) offer a complete description of the structure of the eigen-space associated with
the flux Jacobian matrix A shown in 3.1.7. Moreover, we can formulate a special
similarity transformation, i.e.,

16
A  R ΛL (3.3.23)
or
Λ  LAR (3.3.24)
and
u  a 0 0 0 
 0 u 0 0 
Λ (3.3.25)
 0 0 u 0 
 
 0 0 0 u  a
is the diagonal matrix of eigenvalues.11 Recall that matrix L is the inverse of R . Our
discussion of the numerical physics behind Roe’s scheme for the reactive Euler equations
is now complete. The Roe formulation is quite important from the theoretical standpoint,
but this method is difficult to implement for two or more non-Cartesian space
dimensions. Fortunately, other flux-based discretization methods such as the Harten, Lax
and van Leer (HLL) family of schemes can easily be applied to this problem. Moreover,
these methods do not require the calculation of pressure derivatives (yet to be discussed)
for the mixture equation of state. This fact affords greater of ease of calculation for a
production numerical scheme.

17
4 BUILDING THE NUMERICAL SCHEME
In this section, we pull together all of the aspects of detonation physics and
mathematics discussed in preceding sections and dedicate our efforts to the solution of
our benchmark problem – simulating the detonation of a finite sphere of HMX. In order
to accomplish this goal, we begin by presenting detailed pressure derivatives for our
mixture equation of state. Then we discuss the details associated with our chosen
numerical integration scheme including formulation of the numerical flux vector.
4.1 Pressure Derivatives
The purpose of this subsection is to document formulas for the pressure

derivatives (3.1.2) of the mixture equations of state. These derivatives must be computed
under the support defined by the set of primitive variables.11 In this work, we consider
two mixture equations of state. The first mixture EOS, called the Hayes-I/JWL EOS is
given by substituting (2.3.1) and (2.4.1) into (2.2.5). The second mixture EOS, referred
to as the Hayes-II/JWL EOS, is created by substituting (2.3.5) and (2.4.1) into (2.2.5).
Either mixture EOS consists of a lengthy formula, so to promote brevity in
documentation, we can relate the two mixtures equations of state to one another. If we
look carefully at the Hayes-I and Hayes-II formulas, (2.3.1) and (2.3.5), respectively, we
see that
H1  
N
   
e e 
II I
 
   1 (4.1.1)
gN   0 
s s
 

These expressions for the internal energy of the solid explosive differ by only one term.
The Hayes-I/JWL mixture EOS may be written as
eMI  (1  ) esI   eg (4.1.2)
Hence, by using (4.1.1), we may write the Hayes-II/JWL mixture EOS as
H1 (1  )  
N
 (1  )   
e  (1  ) e     1   eg (4.1.3)
II I
gN  0 
M s
 

where we have used (2.2.3). A general formula for the Hayes- K /JWL mixture EOS may
be written as
 N

K H 1 (1  )  (1  )   
eM  (1  ) es   II
K I

 
  1   eg (4.1.4)
gN  0  
 
Accordingly, equations (2.3.1) through (2.3.4) may be used to expand (4.1.4) and obtain

18
N 1
     
e  P D   1   0   t 4 (1  ) N    t5 (1  )
K
   0 
M

 1    Rˆ   1    Rˆ 
 A    exp   1   B   exp   2   (Q  e0 )  (4.1.5)
ˆ ˆ
   R1       R2   
H 1 (1  )  (1  )   
N
 K
   1 
gN   0 
II
 
where
1  1
D  (4.1.6)
g 
P
  t3  0 (4.1.7)
0
    ( N  1) t4 (4.1.8)
P0
t5  t 4  (4.1.9)
g
Equation (4.1.5) may be solved for pressure, i.e.,
1
N 1
     
P  eMK   1   0   t 4 (1  ) N    t5 (1  )
D     0 

 1    Rˆ   1    Rˆ 
 A    exp   1   B    exp   2   (Q  e0 )  (4.1.10)
ˆ ˆ
   R1       R2   
 N

H1 N 1   
 (   )    (   ) 
K
 1   1

II
gN   0
Although (4.1.10) is complicated, it is in a convenient form for differentiation through the

use of the quotient rule. We also note that (4.1.10) consists of a sum of eight terms, i.e.,
1 8
P  ci  i ,
D i 1
(4.1.11)
so we may use linearity and differentiate each term individually. If we designate a non-
conservative variable of differentiation as q , q { ,  , e} , then we have that
P 1 8
  i D 
 2
q D
 c  D q
i i .
q 
(4.1.12)
i 1 

19
Equation (4.1.12) presents a very convenient method for evaluating pressure derivatives.
Below, we list explicit equations required in evaluating (4.1.12).
1 1 1

1  eMK ; c1  1;  0;  0; 1 (4.1.13)
  e
 2 0 2 2

 2  1   ; c2   ;  2;  1; 0 (4.1.14)
0    e
N 1 N 2
   3 N  1   
3  (1  )  
N
; c3  t 4 ;  (1  ) N  
 0   0  0 
N 1
(4.1.15)
3    3
  N (1  ) N 1  ; 0
  0  e
 4  4  4
4  1  ; c4  t4 ;  0;  1; 0 (4.1.16)
  e
 1    Rˆ  5
5     exp   1  ; c5  A ; 0
ˆ
   R1    e
5 1  Rˆ 1   Rˆ 
 2  1   1 exp   1  (4.1.17)
          
5  Rˆ1 1 1  Rˆ1 
     exp  
   (  ) 2   Rˆ1   
 
 1    Rˆ   6
 6     exp   2  ; c6  B ; 0
ˆ
   R2    e
 6 1  Rˆ 1   Rˆ 
 2  2   1 exp   2  (4.1.18)
          
 6  Rˆ 2 1 1  Rˆ 2 
     exp  
   (  ) 2   Rˆ 2   
 
 7  7  7
 7   ; c7  Q  e0 ;  0;  1;  0. (4.1.19)
  e

20
N N 1
   H 8 N   
8  (1  ) N 1
     1; c8  1 ;  (1  ) N 1 
 0  gN  0  0 
N
(4.1.20)
8   (1  )  8
 1 ( N  1)  ; 0
  0  e
We also have that
D 1 D 1 D
 2 ;  ; 0 (4.1.21)
    g e
Clearly, we may use (4.1.12) through (4.1.21) to evaluate the pressure derivatives
required by the eigen-space decomposition discussed in Section 3.
4.2 Finite Volume Discretization
Ultimately, we must discretize the governing equations (2.1.1) in order to

numerically solve the detonation problem. We may illustrate the discretization procedure
by considering a simplified form of (2.1.1), i.e.,
U F
 S (4.2.1)
t x
where S is a vector containing all of the source terms. To enact the finite volume
discretization, we integrate (4.2.1) in 1-D space as follows
U F
xi 1 / 2 x
i 1 / 2 x
i 1 / 2
 t
xi 1 / 2
dx   x
xi 1 / 2
dx   S dx
xi 1 / 2
(4.2.2)
Moreover, we obtain
U
xi 1 / 2 xi 1 / 2
  F   S dx (4.2.3)
xi 1 / 2
dx
xi 1 / 2
t xi 1 / 2
xi 1 / 2
Since the limits are fixed in the first term of (4.2.3) and since we assume that U is
continuous on the interval ( xi 1/ 2 , xi 1/ 2 ) , we may interchange the order of integration and
differentiation to find that
 i 1 / 2
x xi 1 / 2
t xi 1 / 2  S dx
U  F  (4.2.4)
xi 1 / 2
dx xi 1 / 2
xi 1 / 2
By observing that the integral in the first term is taken over space, we may evaluate it as

21
~
xi1/ 2
 U dx  U ( x
xi1/ 2
i i 1/ 2  xi 1/ 2 ) (4.2.5)
~
where U i is the average of U  U( x, t ) taken over space in the interval [ xi 1/ 2 , xi 1/ 2 ] .
This interval defines cell i in the finite volume grid. Because of the integration, observe
~ ~
that Ui  Ui ( x) . If we also apply this idea to the source term, (4.2.4) becomes
~
dU i ~
( xi 1/ 2  xi 1/ 2 )  F xi1/ 2  Si ( xi 1/ 2  xi 1/ 2 ) (4.2.6)
x
dt i 1 / 2
the so-called semi-discrete form. Hence,

~
dU i 1 ~
 (Fi 1/ 2  Fi 1/ 2 )  Si (4.2.7)
dt xi 1/ 2  xi 1/ 2
The values of F used in (4.2.7) are evaluated at cell interfaces (natural locations for
possible discontinuities in Euler solutions). As a result, at each interface, F is evaluated
as a numerical flux through the use of an upwind discretization scheme based on the
~
values of U i defined at the cell centers. The upwind scheme, described later in
Subsection 4.4, makes use of the theory developed in Section 3.
4.3 Temporal Discretization
The semi-discrete form (4.2.7) offers certain numerical advantages (or

disadvantages, depending on your point of view). This form effectively decouples the
temporal discretization scheme from the spatial discretization. As a result, we are free to
choose different methods for each discretization. On the other hand, one may argue that it
is unwise to decouple the time and space schemes. Why? Our shock-capturing scheme
fundamentally relies on solutions of the Riemann problem and on characteristics.12
Characteristics adjoin the time and space coordinates in an inextricable manner, so in the
strictest sense, these coordinates cannot be decoupled. This effect has led to the creation
of a large family of schemes based upon Godunov’s method that couple the time and
space discretization.13 Although we do not disagree with these ideas, our development is
evolutionary, so it is very important that we understand our space scheme at a
fundamental level. For these reasons, we will use the decoupled approach involving what
is perhaps the simplest, explicit temporal discretization method. Let us recall (4.2.7) and
discretize the time derivative with a simple forward difference. The current time level is
indicated by the superscript n .
~ ~
U in1  U in 1 ~
 (Fin1/ 2  Fin1/ 2 )  Sin (4.3.1)
t xi
where t  t n1  t n is the numerical time-step, and xi  xi 1/ 2  xi 1/ 2 is the spatial
stepsize. Note that (4.3.1) represents a fully explicit method; by rearranging, we obtain
22
~ ~ ~ F n  Fin1 / 2 
U in1  U in  t S in  i 1 / 2  (4.3.2)
 x 
Basically, equation (4.3.2) implements the Euler time integration method.14 The only
numerical stability control we place on (4.3.2) involves a restriction on the time-step t .
This restriction is enforced through a Courant-Friedrichs-Lewy (CFL) criterion. We
apply a factor of 0.5 to the new predicted time-step given by
 xi 
t pred  min  
 (4.3.3)
 ui  ai
1i i max

4.4 The Numerical Flux
As we mentioned earlier, the flux vector F defined at each interface must be

evaluated via an upwind method in order to facilitate the automatic capturing of shock
waves without numerical oscillations. Our upwind method of choice is Roe’s flux
difference splitting scheme.12 To promote notational clarity, let us designate the
numerical flux vector by the symbol f while retaining the symbol F for the regular flux
vector (2.1.3) defined by the reactive Euler equations. Roe’s numerical flux vector is
simply stated below.11
1 ~
f (FL  FR  A (U R  U L )) (4.3.4)
2
~
where A is the flux Jacobian matrix defined by (3.3.23) and evaluated at the interface in
Figure 1. Interface Notation
question. The (~) notation indicates that this evaluation is conducted with the use of Roe-
averaged variables. The designations L and R are best explained by referring to Figure 1.
The subscript L or R designates that the quantity is defined just to left or right of the
23
interface, respectively. In Figure 1, the interface is located at xi 1 / 2 between cell i and
cell i  1 . Why would the left and right interface values of some property differ? The
answer is very simple. Remember that we stated earlier that our method involves
solutions of the Riemann problem. These solutions admit discontinuities, e.g., shock
waves. Hence, by the nature of a discontinuity, the properties taken to the left and the
right of an interface differ. In the simplest view, we can say that the properties to the left
of the interface taken on the values defined in cell i ; it follows that the properties to the
right of the interface take on the values defined in cell i  1 . This means of selecting the
left and right interface values renders first-order accuracy on uniform meshes. There are
other ways to define these upwind values. A higher order method is discussed in a later
subsection. Our Roe averages are computed from these upwind (L and R) variables.
The Roe average constitutes the physically correct representation of an average at

a discontinuity conforming to the basic ideas of flux difference splitting.15 A
mathematically lengthy derivation is required to produce Roe’s formulas, so we merely
state the results.10
~   L  R (4.3.5)
u  L  uR  R
u~  L (4.3.6)
L  R
~ H L  H R R
H L (4.3.7)
L  R
~ e  L  eR  R
e L (4.3.8)
L  R
~ L  L  R  R
 (4.3.9)
L  R
~ ~ 1 
P  ~ H  e~  u~ 2  (4.3.10)
 2 
~~
~ 2 ~ P Pe
a  P  ~ 2 (4.3.11)

One may note that (3.3.20) through (3.3.22), (3.3.25) and (4.3.11) require Roe-averaged
pressure derivatives. Recall that explicit formulas for these derivatives are presented in
(4.1.12) through (4.1.20). The derivatives are presented in terms of the primitive
variables, so we claim that Roe-averaged values of the pressure derivatives may be
24
obtained by simply evaluating these formulas for the Roe-averaged variables presented in
(4.3.5) through (4.3.10). In practice, this procedure seems to work well.
We may now address the practical evaluation of the numerical flux vector as it is
defined in (4.3.4). The vectors FL and FR are the standard Euler flux vectors (2.1.3)
evaluated for the upwind conservative variables U L and U R (or primitive variables q R
and q L ), respectively. The remaining term
~
A (U R  U L ) (4.3.12)
is denoted as the numerical viscosity expression. The difference between the conservative
variables left and right of the interface may be easily evaluated through the use of (2.1.2).
~
A may be evaluated as follows.
~ ~ ~~
A R ΛL (4.3.13)
where the (~) notation indicates that all of the entries in the matrices are calculated with
~
the use of averaged variables. The matrix Λ is created by taking the absolute value of
~
each element of Λ , the diagonal matrix of eigenvalues. Finally, (4.3.12) is computed by
a series of simple matrix-matrix and matrix-vector multiplications; (4.3.4) is easily
evaluated by using vectors sums.
4.5 A Higher-Order Scheme
The scheme described in the preceding subsection is only accurate to the first
order, and it is highly dissipative, a detriment to the sharp resolution of detonation waves.
In this subsection, we briefly describe an enhancement to the first order scheme that is
third-order accurate on uniform grids. As you may have concluded, the left and right
interface values are constructed from the cell-center values to the left and right of the
interface, respectively. To increase the order of accuracy for the scheme, we instead
reconstruct the interface values using interpolating polynomials involving more than one
cell-center value. One way to apply this idea is through the use of a Monotone Upwind
Scheme for Conservation Laws (MUSCL).12 The equations for the left and right interface
variables are provided below for the interface located at i  1/ 2 . Consider the primitive
variable q , q   , u , P , .
1  1 
qL  qi 1  (1  ) (rL )(qi 1  qi 2 )  (1  )  (qi  qi 1)
4  rL  
(4.4.1)
where   1/ 3 to achieve third-order accuracy, and

25
qi  qi 1
rL  . (4.4.2)
qi 1  qi 2
 is a function designed to serve as a non-limiter limiter. In every case, our interpolated

data must be monotone; otherwise, the interpolation procedure will result in the
formation of non-physical oscillations in the numerical solution.12 The nonlinear limiter
is designed to maintain the monotonicity of smooth sections of data when interpolated to
high order. We have chosen the Van Albada limiter for use in this problem, i.e.,
r2  r
(r )  (4.4.3)
1 r 2
The right interface variable is given by
1  1 
qR  qi  (1  ) (rR )(qi 1  qi )  (1  )  (qi  qi 1) (4.4.4)
4  rR  
For this expression, the ratio used by the limiter is defined as
qi  qi 1
rR  (4.4.5)
qi 1  qi
Equations (4.4.1) through (4.4.5) cannot be implemented without due cognizance. The
left interpolant involves cell-center values located at i  2 , i  1 and i . As a result, we
must ensure that
(qi  qi 1) (qi 1  qi 2 )  0 (4.4.6)
Otherwise, the cell-center data is non-monotone, and the interface values must be set to
the first-order values
qL  qi 1
(4.4.7)
qR  qi
in order to properly smooth the solution. For the right interpolant, we must ensure that
(qi  qi 1) (qi 1  qi )  0 (4.4.8)
or we must use the first-order interpolation values (4.4.7). In addition, after the criteria
(4.4.6) and (4.4.8) are satisfied, we are required to limit on the ratios (4.4.2) and (4.4.5).
Based on the data, these ratios may become undefined, so the limiter function (4.4.3)
must be modified ensure that its value always remains finite. If this interpolation strategy
is used properly, the Roe algorithm becomes a high-resolution flux difference splitting
scheme.
26
4.6 Boundary Conditions
In most cases, we cannot solve partial differential equations without applying

boundary conditions. Even for our simple detonation problem cast in one dimension, we
must apply boundary conditions at x  0 (the center of the sphere) and at x  xMAX (the
outer surface of the sphere). At the center of the sphere, we enforce fully reflective
boundary conditions through the use of a ghost cell installed at i  0 , i.e.,
 0  1
u0  u1
P0  P1 (4.5.1)
0  1
e0  e1
We have assumed that the first flow field cell adjacent to this boundary has the index
i 1.
At the outer surface of the sphere, we apply extrapolated boundary conditions to
mimic a supersonic outflow. We implement this condition by installing a ghost cell at
i  iMAX . We set conditions in this cell as follows.
 IMAX   IMAX -1
uIMAX  u IMAX -1
PIMAX  PIMAX -1 (4.5.2)
IMAX  IMAX -1
eIMAX  eIMAX -1
Boundary conditions (4.5.1) and (4.5.2) function well for the detonation of a finite
spherical mass of HMX.

27
5 PARTICLE MOTION
In this section, we extend our discussion beyond the application of numerical

detonation literature cited thus far. Given the level of interest in Multiphase Blast
Explosives (MBX), it is desirable to incorporate solid particles into our detonation
programming. This effort is new, so our treatment of solid particles is limited, to a certain
extent. Still, our particles have realistic mass and finite radii. They are driven by the
detonation through the use of Lagrangian laws of motion. Our particle algorithms have
only three major limitations:
(i) The particle collection exists in the diffuse limit. Particles are assumed not to
interact with one another.
(ii) Particles are assumed to exist as rigid spheres. The do not deform or change
phase during the detonation event.
(iii) This model is restricted to one dimension. We can only establish initial
particle positions along a single ray.
Based on these assumptions, we can investigate the efficacy of this model in predicting
the post-detonation conditions for a mass of solid HMX loaded with particles.
5.1 Coupling Terms
We may now discuss the coupling terms (source terms) for particles presented in
equations (2.1.1) and (2.1.6). Fs and Q s have relatively simple descriptions. Fs
represents the transfer of momentum between the gas phase and the particle phase while
Q s represents the similar transfer of thermal energy. For spherical particles, these terms
may be written in a simple form.6 Assume that the total number of particles is N p .
4
N
p
du p
Fs     p rp3
 (5.1.1)
p 1 3 dt
~
pN
Q s   4 hp  rp2 (T  Tp ) (5.1.2)
p 1
where  p , rp and u p are the solid density, radius and velocity of the p th particle,
~
respectively. Therefore, du p / dt is the acceleration of the p th particle. Also, T is the
temperature of the gas phase at the surface of the particle, and T p is the particle
~
temperature. Actually, T is the Favre-filtered temperature; this filtering operation is used
to take the presence of turbulence into account. Our simulation is non-viscous, so we
~
simply set T equal to the gas phase temperature T . The parameter h p is the heat transfer
coefficient that governs the transfer of thermal energy at the particle/fluid interface. In
28
general, h p is experimentally determined. By specifying (5.1.1) and (5.1.2), we can
accurately describe the coupling between the gas and particulate phases. Of course, these
equations only apply to particles of fixed mass. Additional terms (including mass
conservation) must be specified for particles that react with the gas phase.
5.2 Particle Laws of Motion
The detonation physics algorithms incorporate discrete, finite-mass particles, so

we apply Lagrangian equations for tracking the movement of particles. Let x p designate
the radial coordinate of the p th particle. Then we have that
dx p
 up (5.2.1)
dt
The particle velocity u p must be determined from the evolution equation given by a
model. We have two alternatives for this model; the first is called the “Spray Model”
which may be described as follows.6
du p 3 CD  Re p
 (u  u p ) (5.2.2)
dt 16  p rp2
where the particle Reynolds number Re p is defined as

2 rp 
Re p  u up (5.2.3)

The drag coefficient for the particle C D is conveyed by the “Spray Drag Law”, i.e.,
 24  Re 2p/ 3 
 1  Re p  1000
Re  
 p 6 

CD   (5.2.4)
 0.44 Re p  1000


 ,  and u are the density, dynamic viscosity and velocity of the gas phase in the
vicinity of the particle. This model is not appropriate for detonation problems, but it still
serves well for testing. For the problem of a detonation with solid inclusions, we apply a
high speed gas flow model originally developed for solid rocket motors.

29
The high speed gas flow model was developed for the multiphase flow field
created by the burn of porous, powdered explosive material.16 In this case, the particle
acceleration is given by
du p  d p2 CD 
 u  u p (u  u p ) . (5.2.5)
dt 8 mp
In order to maintain our notation consistent with the literature, (5.2.5) is written in terms
of the particle diameter d p instead of the radius. Also, m p is the mass of the p th particle.
This high speed drag law provides the drag coefficient through a more complicated
calculation. First, we calculate a “Mach-zero” drag coefficient, C D 0 , i.e.,
 C1  2  0.08


 (0.45   2 ) C1  ( 2  0.08) C2 0.08   2  0.45
CD0  0.37

 C2  2  0.45


(5.2.6)
where Re p is calculated by using (5.2.3), and
24 4.4
C1    0.42 (5.2.7)
Re p Re p
4  150 2 
.
C2  1.75  (5.2.8)

31  1 Re p 

In (5.2.6) and (5.22.8), we have introduced two new parameters  1 and  2 ; they are the
volume concentrations of the gas and particle phases, respectively. These parameters
require interpretation when considering the detonation problem. At the outset of the
problem, the solid explosive has not been detonated, so there is no gas phase at this point.
The best course of action is to compute the initial values of  1 and  2 based upon the
volume of the solid explosive and the volume of particles. Since we are not simulating
details of the shock interaction with metal particles, we calculate  1 and  2 on this basis
of the initial calculation and maintain them fixed for the duration of the detonation. We
must then calculate a final value of C D based on a Mach correction.17 This correction
exists due to the natural variation in the drag coefficient with Mach number. If we do not
wish to implement a drag correction, then we set C D  CD 0 ; otherwise the corrected value
of C D may be calculated from

30
  0.427  
C D  C D 0 1  exp   4.63   , (5.2.9)
  M 
where
u up
M . (5.2.10)
a
By using the particle velocities provided by (5.2.2) though (5.2.4) or (5.2.5) through
(5.2.10), we may integrate (5.2.1) to determine the track of each particle through space
during the detonation.

31
6 RESULTS
From the start of this effort, several versions of our current numerical detonation
computer code have been developed by the author. The purpose of this section is to
present some of the results produced for typical problems. Specifically, we discuss three
results. The first set of results is intended to show that our detonation program is
functioning properly and producing physically correct solutions. In a second calculation,
we address the numerical detonation of a spherical mass of pure HMX. For this problem,
we have computed results by using both the Hayes-I and Hayes-II equations of state for
the solid explosive combined with the JWL EOS for the detonation products. Finally, we
discuss the results for the detonation of a spherical mass of HMX loaded with steel
particles.
6.1 Simple Plane Wave Detonation
This test problem, described in Reference 2, is used to show whether or not the
flux difference splitting scheme is working properly. In this case, we endeavor to solve a
Deflagration to Detonation Transition (DDT) problem in one dimension. Both the
explosive and the detonation products are modeled by using the calorically perfect gas
EOS. The associated mixture EOS is given as
P
e Q (6.1.1)
 (  1)
As discussed in Section 4, we apply fully reflective boundary conditions at x  0 and

extrapolation conditions at x  xMAX . For this problem, we use the reaction rate
expression
 
 E 
r  k (1  ) exp   a  (6.1.2)
 P 
 
where (6.1.2) is in Arrhenius form; k is the reaction rate constant, and E a is a parameter
that behaves like an activation energy. The one-dimensional domain is defined in
0  x 12 . Also, we have that E a  10 ; Q  50 ;   1.4 , and k  7 . The problem is
initialized with u  0 ; P  0 , and   0 everywhere.2 The initial density distribution is
given by
1
 ( x)  , 0  x  12 . (6.1.3)
1 3 exp(  x 2 )
This density distribution initiates the reaction in the region near x  0 by boosting the
reaction rate term.

32
Figure 2. Problem 1 Detonation Field Density, Time = 3.0
Figure 3. Problem 1 Detonation Field Velocity, Time = 3.0

33
Figure 4. Problem 1 Detonation Field Pressure, Time = 3.0
Figure 5. Problem 1 Detonation Field Reaction Progress Variable, Time = 3.0
This problem does not possess an “exact” solution, but Xu et al. have obtained a
fully converged numerical solution using a mesh consisting on 3200 cells.2 This problem
provides an excellent test detonation physics algorithms. Accordingly, we have generated
three numerical solutions on grids comprised of 200, 800 and 3200 cells, respectively.
The numerical solutions for density, velocity, pressure and the reaction progress variable
are provided in Figures 2 through 5, respectively, at the dimensionless time 3.0. In each
figure, solution plots are color-coded to correspond to the mesh used. The behavior
shown in each plot agrees quite well with archived plots.2 We have observed only one
anomaly in our solutions. Strangely enough, on the mesh consisting of only 200 cells,
there are noticeable oscillations in the reaction progress variable. These oscillations
dissipate with increasing mesh density. The explanation for this behavior is not
immediately evident. In some of our solutions, the reaction progress variable has been
observed to hunt between the solid and gaseous equations of state. In fact, this variable is
34
very sensitive and couples strongly to the reaction rate. We apply no post-solution
filtering to this variable. Secondly, we are using a weak time integration scheme with
poor numerical stability performance. The oscillations become less prevalent with
increasing grid density, so the space scheme may be compensating for the time scheme.
This phenomenon bears further investigation as this work continues. We will also re-
examine the nonlinear limiter coding. Nevertheless, our converged solution agrees well
with the converged archival solution.2
6.2 Detonation of Pure HMX
This problem is intended to demonstrate our computer code’s capability for

simulating the detonation of a sphere of pure HMX. This problem permits a test of our
discretization of the geometric source term found in the reactive Euler equations (2.1.1)
and (2.1.4). It also represents our first attempt at capturing the physics of a realistic
detonation event. In this case, we address the detonation of sphere of solid HMX with a
radius of 4.5 cm. The radius of the sphere is divided into 800 cells. Figure 6 shows the
density, velocity, pressure and reaction progress variables for the numerical solution at
three microseconds (μs) detonation elapsed time. As you can see, the Von Neumann
spike is clearly resolved in this solution as is the Taylor wave. Moreover, the Chapman-
Jouquet pressure is captured at the experimentally obtained value of 42 GPa. Also, the
numerical detonation velocity has a value of 1.02 cm/μs which is very close to the
experimentally obtained value of 0.911 cm/μs.21 Of course, the experimental value is
generally taken from tests that mimic plane wave detonation conditions. As a result, we
expect to calculate a different value for the spherical detonation problem. Overall, the
results agree very closely with the archival data. We have also solved this same problem
by using the Hayes-II/JWL mixture EOS. The results of this analysis are given in Figure
7. It is interesting to observe that the Taylor wave is captured in this solution even more
smoothly than it was in the preceding case. The more complex Hayes-II EOS may
actually offer greater stability when used in the mixture EOS. This numerical solution
also offers excellent comparisons with the Chapman-Jouquet pressure and detonation
velocity for HMX. Both mixture equations of state show that the detonation reaction
occurs in a nearly instantaneous manner. As you can see, the reaction progress variable
changes in a nearly discontinuous manner at the detonation front. In either case, our
computer programming captures the appropriate physics for the detonation, and it renders
a wide array of physical data (far more than is shown here).

35
Figure 6. Numerical detonation solution Hayes-I/JWL in HMX at 3 μs. Horizontal axis is distance in
meters.
Figure 7. Numerical detonation solution Hayes-I/JWL in HMX at 3 μs. Horizontal axis is distance in
meters.

36
6.3 Detonation of HMX Containing Metal Particles
This test case is the final detonation problem addressed by this report. We
consider the detonation of a spherical mass of HMX loaded with a radial distribution of
steel particles. The mass of the HMX sphere remains the same as is used for the
preceding problem, and we still have 800 finite volume cells defined along the charge
radius. For this example, we have placed ten particles, at uniform spacing, along the
charge radius. The particles each have a radius of 463 μm and a material density of 7860
kg/m3. We assume the gas viscosity has a value of 1.7x10-5 kg/(m.s). Furthermore, in this
simulation study, we have applied the high speed flow drag law. The results for particle
locations are presented in Figure 8 while the plot of particle velocities is given in Figure
9. The particle tracks shown in Figure 8 clearly indicate the passage of the detonation
wave. For particles farther away from the charge center, the particle tracks show changes
in slope at progressively larger times. The sudden change in track slope concurs with the
nearly discontinuous change seen in the particle velocity traces shown in Figure 9. Also,
in Figure 9, the effect of the drag law can clearly be seen as the particle velocities rise
rapidly in the wake of the detonation wave then fall quickly under the action of drag in
the region behind the wave. We have also applied the Mach correction to the rocket drag
law. In the velocity trace for the particle closest to the charge center, we can see the
velocity begin to level off at 4.5 μs. Available data indicates that the calculated terminal
velocity at or near 375 m/s is an acceptable value. This simulation does not include
thermal effects since we are still in the process of completing our detonation products
EOS.
Figure 8. Radial locations for steel particles embedded in a mass of detonating HMX

37
Figure 9. Radial velocities for steel particles embedded in a detonating mass of HMX

38
7 CONCLUSIONS
In this report, we have presented the governing equations for the direct numerical
simulation of the detonation of a solid explosive material. Proper equations of state have
been discussed for both the solid explosive material and for the gaseous detonation
products. From these equations of state, we have developed a mixture equation of state
relating the specific internal energy for the detonation to the thermodynamic pressure.
The resulting computer program has been tested on an archival detonation problem for
the purpose of comparison. We have presented results for the detonation of a spherical
mass of pure HMX.
More importantly, we have incorporated particle tracking algorithms within the

programming. As a result, the code can now explosively drive particles under the action
of a detonation wave with coupling to a drag law. This mechanism allows the code to
simulate the detonation of a Multiphase Blast Explosive in the diffuse limit of particle
loading. We have built drag laws for both spray and high speed gas flow drag law into the
code. For a test problem, we have simulated the detonation of a mass of HMX loaded
with a radial distribution of steel particles. The trend in post-detonation velocities of these
particles meet our expectations.
8 RECOMMENDATIONS
During the months ahead, detonation physics algorithms are scheduled for
implementation in LESLIE3D. The development of the present work has been a learning
experience accompanied by a large number of difficulties, especially in the
implementation of Roe’s flux difference splitting scheme. A first recommendation is that
the HLL family of schemes be used instead. These schemes are more robust and do not
require the use of pressure derivatives. Also, these schemes already operate well inside of
LESLIE3D. The detonation physics solver will also benefit from the interface tracking
scheme already coded into LESLIE3D. Clearly, the governing equation differ at the
interface between the condensed explosive and the surrounding gas field. This situation
necessitates an interface to maintain code stability.
The detonation physics algorithms discussed here must be adapted for curvilinear
coordinates in three dimensions. For HLL flux forms, this process should not be difficult.
The author has already done some work in this area. However, the pressure and specific
volume (or density) closures associated with the mixture equation of state do require
attention. The Gas-Interpolated Stewart-Prasad-Asay (GISPA) method requires these
closures to address the multiphase physics of detonation. There is no unique set of
closures available for this process, but the chosen closures must be carefull accomplished.
Some difficulty has been encountered in the use of the specific volume closure (due to
Xu), and this difficulty should be investigated and resolved.
The Hayes equation of state for the solid explosive is an older relationship that
characterizes very few explosives. The Mie-Gruneisen equation of state characterizes
many more explosive materials. That is to say, there is data available. However, the
39
mixture equation of state must be rederived for the Mie-Gruneisen formulation. It may be
combined with the JWL adiabat for the detonation products, or with another real gas state
equation. The “Wide-Ranging” detonation equation of state may also be implemented.4
Ultimately, the particle phase algorithms discussed here must be rewritten for
dense phase fields. The detonation of a condensed explosive with solid inclusions is a
dense phase problem. Also, the computer program is currently not properly written even
in the diffuse limit as regards the nonhomogeneous source terms. The integration scheme
should be changed to reflect the use of Strang splitting.1 That is to say, the spatial
integration scheme should be advanced in separate step from the nonhomogeneous terms.
For the latter step, the integration should be conducted in the temporal manner at each
grid cell just like an initial value problem.
REFERENCES
1. Strang, G., “On the construction and comparison of difference schemes”, SIAM J.
Numer. Anal., Vol. 5, No. 3, pp. 506-517, 1968.
2. Xu, S., Aslam, T. and Stewart, D.S., “High resolution numerical simulation of ideal
and non-ideal compressible reacting flows with embedded internal boundaries”,
Combust. Theory Modeling, Vol. 1, pp. 113-142, 1997.
3. Bdzil, J.B., Stewart, D. S. and Jackson, T.L., “Program burn algorithms based on
detonation shock dynamics: Discrete approximations of detonation flows with
discontinuous front models”, Journal of Computational Physics, Vol. 174, No. 2, pp.
870-902, 2001.
4. Wescott, B.L., On Detonation Diffraction in Condensed Phase Explosives, Doctoral

Dissertation, University at Illinois at Urbana-Champaign, 2001.
5. Stewart, D.S., “Tools for Design of Advanced Explosive Systems and Other
Investigations on Ignition and Transient Detonation”, Final Report on a Grant from the
U.S. Air Force Research Laboratory Munition Directorate to the University of Illinois,
2005.
6. Chen, K.H. and Shuen, J.S., “A Coupled Multi-Block Solution Procedure for Spray
Combustion in Complex Geometries”, AIAA Paper 93-0108, American Institute for
Aeronautics and Astronautics, 31st Aerospace Sciences Meeting and Exhibit, January
1993.
7. Stewart, D.S., Electronic Communication, 2006.
8. Hayes, D.B., “A Pnt Detonation Criterion From Thermal Explosion Theory”, Sixth
Symposium (International) on Detonation, Pasadena, California, 1976.
40
9. Davis, W.C., “Complete equation of state for unreacted solid explosive”, Combustion
and Flame, Vol. 120, pp. 399-403, 2000.
10. Glaister, P., “An approximate linearised Riemann solver for the Euler equations for
real gases”, Journal of Computational Physics, Vol. 74, pp. 382-408, 1988.
11. Nance, D.V., “Flux Difference Splitting Algorithms for Real Gas Mixtures”,
Technical Memorandum, Munitions Directorate, Air Force Research Laboratory, March
2006.
12. Hirsch, C., Numerical Computation of Internal and External Flows, Vol. 2, John
Wiley & Sons, New York, 1991.
13. Collela, P. and Woodward, P.R., “The piece-wise parabolic method for gas-
dynamical simulations”, Journal of Computational Physics, Vol. 54, pp. 174-201, 1984.
14. Burden, R.L., Faires, J.D. and Reynolds, A.C., Numerical Analysis, 2nd Ed., Prindle,
Weber & Schmidt, Boston, 1981.
15. Roe, P.L., “Approximate Riemann solvers, parameter vectors and difference
schemes”, Journal of Computational Physics, Vol. 43, p. 357, 1981.
16. Akhatov, I.S. and Vainshtein, P.B., “Transition of porous explosive combustion into
detonation”, Combustion, Explosion and Shock Waves, Vol. 20, No.1, pp. 63-70, 1984.
17. Carlson, D.J. and Hoglund, R.F., “Particle drag and heat transfer in rocket nozzles”,
AIAA Journal, Vol. 2, No. 11, pp. 1980-1984, 1964.

41
APPENDIX A
SOURCE CODE
Instructions:
The source code that follows has been developed over a period of six years, but in
a sporadic manner, as time has permitted. FORTRAN 77 is used throughout the computer
program, and an in-line coding structure has been used. The programming is designed for
research and is thus rather crude. The initial conditions (shock-based initiation) are all
rigidly coded. Different initiation options exist, but they must be enabled or disabled by
commenting. The detonation reaction rate laws are treated in the same way. The desired
reaction rate law must be commented in for the initial conditions and for the first and
second time step segments of the solver. The calorically perfect gas and Jones-Wilkins-
Lee test problems are also activated or deactivated by commenting in/out code segments.
This computer program is written for standard explosives like HMX for which we
have plenty of data. Especially for the Hayes equation of state, a great deal of data input
is required. This data is simply entered directly into the source code. This statement is
also true as pertains to the Jones-Wilkins-Lee detonation product data as well as the
particle field data. This code functions in one dimension only: Cartesian, cylindrical or
spherical. The domain boundaries are contained between x1 and x2. The number of cells
in the detonation field is given by imax-1. The variable NSTP tells the code how many
iterations (time steps) to execute while the variable NDMP tells the code how many
iterations to perform between dump files. The variable IRST controls code execution.
With IRST set at zero, the code begins with the coded initial conditions. With IRST set at
one, the code reads the restart.data file to obtain its starting conditions. The IEOS
variable switches between the mixture equations of state. IEOS equal zero sets calorically
perfect gas conditions. IEOS at one sets JWL conditions while IEOS equal 2 or 3 sets the
Hayes-I/JWL and Hayes-II/JWL formulations. The reader should be advised that the pure
JWL option does not work well. The fault of this equation is that there is not a sufficient
energy separation between the adiabats to result in detonation.
This detonation physics program utilizes a number of flags and control parameters
in order to stabilize code operation. Some of these parameters set tolerances on the
variables (like the reaction progress variable) to prevent “hunting”. Other flags control
solution progress. For instance, internal energy updates are lagged by one iteration to
keep temperature from turning negative. It is also important to observe that the equations
of state used here have constant specific heat formulations. Over time, this limitation
should be lifted, but better equation of state data is required to do so. We also zero the
detonation reaction rate in the far field. As it happens, the flux scheme will erroneously
allow reaction rate to creep up slowly in the unreacted explosive mass. This effect is
damaging to the solution and had to be corrected.
c * * * * * * * * * * * * * EZ1_MASTER * * * * * * * * * * * *
c * * * * * * * * * * * * * * * ** * * * * * * * * * * * * * *
c Program for 1-D detonation test problem
c Simple coding structure
42
c Monotonicity check implemented on extrapolation
c Direct adaptations for calorically perfect gas and JWL
program ez1_master
implicit none
c Parameter statements
integer imax
c parameter (imax = 20001)
parameter (imax = 2001)
integer npar
parameter (npar = 1000)
real*8 c12
parameter (c12 = 0.5d0)
real*8 c13
parameter (c13 = 1d0/3d0)
real*8 c14
real*8 c18
real*8 c23
real*8 c43
real*8 c316
real*8 pi
parameter (pi = 3.141592654d0)
c Variable array declarations

c File I/O
character*12 filex
character*12 parex
c Debug flags
integer idbg1
integer idbgf
integer idbgs
integer idbgp
c Control flags
integer irst
integer ieos
integer igeo
integer irxn
integer ipar
integer idrg
integer imach
43
integer iext
integer iav
integer ilim
integer imon
integer iefx
integer item
c Counters
integer i
integer n,nn,np
integer l,m
integer k
integer nstart
integer nstp
integer ndmp
integer nfil
c Gas phase data

real*8 pamb
real*8 mu
c Calorically perfect EOS data

real*8 gamm
real*8 gam1
c JWL EOS data

real*8 r0
real*8 aj
real*8 bj
real*8 cj
real*8 cjh
real*8 r1
real*8 r2
real*8 wj
real*8 pcj
c Hayes-I EOS data

real*8 cvs
real*8 gh
real*8 h1
real*8 nh
real*8 rgas
real*8 cvg
real*8 cpg
real*8 nhp1
real*8 nhm1
real*8 nhm2
real*8 t3
real*8 t4
real*8 t5
real*8 t7
real*8 alfa
real*8 beta
real*8 thta
c Mixture EOS tolerances

44
real*8 ztol1
real*8 ztol2
c Detonation data
real*8 qdet0
real*8 e0
real*8 eact
real*8 rk
real*8 rk1
real*8 rk2
real*8 pexp
real*8 zexp
real*8 th1
real*8 th2
real*8 rh1
real*8 rh2
real*8 rht
real*8 rhti
real*8 wr1
real*8 wr2
real*8 wr1r
real*8 wr2r
c Grid/Timestep control data

real*8 x1
real*8 x2
real*8 chx
real*8 dx
real*8 xc
real*8 fct
real*8 fct1
real*8 fct2
real*8 time
real*8 tend
real*8 dt
real*8 dt0
real*8 dt1
real*8 dtmx
real*8 cfl
real*8 offs
c Derived data
real*8 et
real*8 ra
real*8 ra2
real*8 ea
real*8 za
real*8 rz
real*8 omz
real*8 rxmin
real*8 bot
real*8 bot2
real*8 botr
real*8 botz
45
real*8 dpdr
real*8 dpde
real*8 dpdz
real*8 a2
real*8 psgn
real*8 kap
real*8 eps
real*8 epsm
real*8 epsp
real*8 off
real*8 tmp
real*8 rl,rr
real*8 ul,ur
real*8 pl,pr
real*8 zl,zr
real*8 el,er
real*8 eel,eer
real*8 hhl,hhr
real*8 dqer,dqwr,dqir
real*8 dqeu,dqwu,dqiu
real*8 dqep,dqwp,dqip
real*8 dqez,dqwz,dqiz
real*8 denm
real*8 dra,drb,drc,drd,dre
real*8 dua,dub,duc,dud,due
real*8 dpa,dpb,dpc,dpd,dpe
real*8 dza,dzb,dzc,dzd,dze
real*8 rat
real*8 phir
real*8 phiu
real*8 phip
real*8 phiz
real*8 phi
real*8 vhi
real*8 sqrl
real*8 sqrr
real*8 rsumi
real*8 rav
real*8 ri
real*8 uav
real*8 zav
real*8 eav
real*8 hav
real*8 aav
real*8 pav
real*8 delr
real*8 delv
real*8 delp
real*8 delz
46
real*8 detr
real*8 pest
c Temperature estimation variables

real*8 tk0
real*8 dtkmx
real*8 denmx
real*8 numr
real*8 e0cr
real*8 eta
real*8 rs
real*8 rg
real*8 de1
real*8 de2
real*8 de3
real*8 de4
real*8 de5
real*8 de6
c Particle phase data

real*8 xp1
real*8 xp2
real*8 dxp
real*8 rdp
real*8 dip
real*8 rop
real*8 pcp
real*8 rep
real*8 ppr
real*8 tcon
real*8 crppr
real*8 nup
real*8 hp
real*8 cdp
real*8 pum
real*8 pam
real*8 delu
real*8 adelu
real*8 hevol
real*8 pvol
real*8 cvol
real*8 p0mas
real*8 pmass
real*8 alf1
real*8 alf2
real*8 alf21
real*8 cd1
real*8 cd2
real*8 cd0
real*8 mach
real*8 dtp
c Array declarations
real*8 x(imax)
real*8 r(0:imax)
real*8 p(0:imax)
47
real*8 u(0:imax)
real*8 z(0:imax)
real*8 ei(0:imax)
real*8 a(0:imax)
real*8 rxr(0:imax)
real*8 c(8)
real*8 top(8)
real*8 topr(8)
real*8 topz(8)
real*8 rp(0:imax)
real*8 pp(0:imax)
real*8 up(0:imax)
real*8 zp(0:imax)
real*8 eip(0:imax)
real*8 etp(0:imax)
real*8 ap(0:imax)
real*8 tk(imax)
real*8 dtk(imax)
real*8 zzl(imax)
real*8 zzr(imax)
real*8 qv(imax,4)
real*8 qvp(imax,4)
real*8 sg(imax,4)
real*8 srx(imax,4)
real*8 sp(imax,4)
real*8 s(imax,4)
real*8 aeg(4)
real*8 evr(4,4)
real*8 cwm(4)
real*8 chk1(4,4)
real*8 chk2(4,4)
real*8 dq(4)
real*8 v1(4)
real*8 vn(4)
real*8 fl(4)
real*8 fr(4)
real*8 fn(imax,4)
real*8 dqv(4)
real*8 derv(imax,2)
c Particle arrays
integer pcel(npar)
real*8 px(npar)
real*8 pu(npar)
real*8 pa(npar)
48
real*8 pxp(npar)
real*8 pup(npar)
real*8 pq(npar)
real*8 ptk(npar)
real*8 ptkp(npar)
ccccccccccccccccccccccccccccccccccccccccccccccccccccccccc
c Main Data Entry Section
c Grid data
x1 = 0d0
c x2 = 200d0
x2 = 3.6d-2
chx = 3.8d-2
c CPG EOS data

gamm = 1.4d0
pamb = 101325d0
rgas = 287d0
c Extrapolation control data

kap = 1d0/3d0
c kap = -1d0
eps = 1d-12
c EOS control tolerances

ztol1 = 1d-2
c ztol1 = 0d0
ztol2 = 0.99d0
c ztol2 = 1d0
c HMX Hayes EOS Data (Xu)

c r0 = 1891d0
c h1 = 1.35d10
c cvs = 1.5d3
c gh = 2.1d3
c nh = 9.8d0
c tk0 = 3d2
c HMX JWL EOS Data (Zukas/Xu)

c aj = 7.783d11
c bj = 0.07071d11
c cj = 0.00643d11
c r1 = 4.2d0
c r2 = 1d0
c wj = 0.3d0
c cvg = (2.4d0 - 0.28d0*r0*1d-3 - 1.3d0)*1d3
c NM Hayes EOS Data

c r0 = 1.13d3
c h1 = 1.32d9
c cvs = 1.446d3
c gh = 1.356d3
49
c nh = 7.144d0
c tk0 = 293d0
c NM JWL EOS Data

c aj = 209.2d9
c bj = 5.689d9
c cj = 0.77d9
c r1 = 4.4d0
c r2 = 1.2d0
c wj = 0.3d0
c cvg = 1.3d3
c RDX Hayes EOS Data

r0 = 1.6d3
h1 = 13d9
cvs = 1.163d3
gh = 1.356d3
nh = 6.3d0
tk0 = 300d0
c RDX JWL EOS Data

aj = 573.187d9
bj = 14.639d9
cj = 0.77d9
r1 = 4.6d0
r2 = 1.4d0
wj = 0.32d0
cvg = 1.2d3
c Detonation reaction data
c CPG Test
eact = 10d0
rk = 16.418d0
c th1 = 0d0
c th2 = 0d0
c rxmin = rk*dexp(-eact)
c qdet0 = 25d0
c HMX Test
c pcj = 42d9
c rk1 = 110d6
c rk2 = 0d0
c pexp = 3.5d0
c zexp = 0.93d0
c th1 = 0d0
c th2 = 0d0
c rxmin = rk1*((pamb/pcj)**pexp)
c qdet0 = (7.91d0 - 4.33d0*(r0*1d-3 - 1.3d0)**2
c & -0.934d0*(r0*1d-3 - 1.3d0))*1d6
c NM Test
c pcj = 12.5d9
c pexp = 1d0
c zexp = 0.95d0
c rk1 = 7.75d10
50
c rk2 = 1.5d12
c th1 = 14500d0
c th2 = 29700d0
c rxmin = rk1*dexp(-th1/tk0)
c qdet0 = 4.530d5
c RDX Test
pcj = 26.5d9
rk1 = 110d6
rk2 = 0d0
pexp = 3.5d0
zexp = 0.93d0
th1 = 0d0
th2 = 0d0
rxmin = rk1*((pamb/pcj)**pexp)
qdet0 = 5.375d6
c Particle data
xp1 = 1.0d-2
xp2 = 5.9d-2
pmass = 4.3d0
rop = 7860d0
rdp = 280d-6
pcp = 446d0
mu = 1.7d-5
c mu = 1.0d-3
tcon = 2.57d-2
c Code control data and flags
c Data
off = 1d-6
cfl = 0.5d0
n = 0
nfil = 0
nstart = 0
nstp = 10
ndmp = 1
dtmx = 1d-2
time = 0d0
tend = 50d0
c Flags
irst = 1
iav = 1
iext = 1
ilim = 1
ieos = 3
igeo = 1
irxn = 1
iefx = 2
ipar = 0
idrg = 1
imach = 1
51
c Debug control
idbg1 = 0
idbgf = 0
idbgs = 0
idbgp = 0
write(*,*) ' Code Control Data:'

write(*,*) ' nstp = ',nstp
write(*,*) ' ndmp = ',ndmp
write(*,*) ' tend = ',tend
if (ipar .eq. 1) write(*,*) ' npar = ',npar
write(*,*) ' '
write(*,*) ' Flags:'
write(*,*) ' irst = ',irst
write(*,*) ' iav = ',iav
write(*,*) ' iext = ',iext
write(*,*) ' ilim = ',ilim
write(*,*) ' '
write(*,*) ' ieos = ',ieos
write(*,*) ' igeo = ',igeo
write(*,*) ' irxn = ',irxn
write(*,*) ' iefx = ',iefx
write(*,*) ' '
write(*,*) ' ipar = ',ipar
write(*,*) ' idrg = ',idrg
write(*,*) ' imach = ',imach
write(*,*) ' '
pause
c Derived data
c Thermal data
cpg = rgas + cvg
ppr = cpg*tcon/mu
crppr = ppr**c13
c EOS Parameters
rh1 = r1*r0
rh2 = r2*r0
wr1 = wj/rh1
wr1r = wr1/r0
wr2 = wj/rh2
wr2r = wr2/r0
cjh = cj*(r0**(-(1d0 + wj)))
nhp1 = nh + 1d0
alfa = nh - 1d0
nhm1 = alfa
nhm2 = nh - 2d0
e0 = cvg*tk0
c Hayes-I EOS
t3 = cvs*tk0*gh/r0
t4 = h1/r0/nh/alfa
t5 = pamb/gh + t4
t7 = pamb/gh + beta + t4
52
thta = t3 - pamb/r0
beta = thta + alfa*t4
c Compute coefficients for Hayes pressure derivatives

c(1) = 1d0
c(2) = beta
c(3) = -t4
c(4) = t5
c(5) = aj
c(6) = bj
c(7) = qdet0 + e0
c(8) = h1/gh/nh
c Particle phase parameters

dip = 2d0*rdp
p0mas = c43*pi*rop*rdp*rdp*rdp
pvol = pmass/rop
if (chx .le. x2) then
write(*,*) ' '
write(*,*) ' chx < x2.'
write(*,*) ' '
stop
else
dx = chx - x2
endif
cvol = c43*pi*x2*x2*x2
c cvol = hevol + pvol
alf2 = pvol/cvol
alf1 = 1d0 - alf2
if (ipar .eq. 1 .and. alf1 .eq. 0d0) then

write(*,*) ' '
write(*,*) ' alf1 = 0!'
write(*,*) ' '
stop
endif
alf21 = alf2/alf1
c Other constants
epsm = c14*(1d0 - kap)
epsp = c14*(1d0 + kap)
gam1 = gamm - 1d0
c Set up the solver report file

open(90,file='rpt.txt',form='formatted')
write(90,*) ' ********** Detonation Solver Report File
**********'
write(90,*) ' '
write(90,*) ' Reaction Data:'
write(90,*) ' qdet = ',qdet0
write(90,*) ' eact = ',eact
write(90,*) ' rk = ',rk
write(90,*) ' rk1 = ',rk1
write(90,*) ' rk2 = ',rk2
write(90,*) ' pexp = ',pexp
write(90,*) ' zexp = ',zexp
53
write(90,*) ' Pcj = ',pcj
write(90,*) ' th1 = ',th1
write(90,*) ' th2 = ',th2
write(90,*) ' '
write(90,*) ' rxmin = ',rxmin
write(90,*) ' '
write(90,*) ' EOS Control Data:'
write(90,*) ' ztol1 = ',ztol1
write(90,*) ' ztol2 = ',ztol2
write(90,*) ' '
write(90,*) ' CPG EOS Data:'
write(90,*) ' gamm = ',gamm
write(90,*) ' gam1 = ',gam1
write(90,*) ' '
write(90,*) ' Hayes-I EOS Data:'
write(90,*) ' H1 = ',h1
write(90,*) ' Cvs = ',cvs
write(90,*) ' g = ',gh
write(90,*) ' N = ',nh
write(90,*) ' T0 = ',tk0
write(90,*) ' '
do nn = 1,8
write(90,*) ' c(',nn,') = ',c(nn)
enddo
write(90,*) ' '
write(90,*) ' alfa = ',alfa
write(90,*) ' beta = ',beta
write(90,*) ' thta = ',thta
write(90,*) ' t3 = ',t3
write(90,*) ' t4 = ',t4
write(90,*) ' t5 = ',t5
write(90,*) ' t7 = ',t7
write(90,*) ' '
write(90,*) ' JWL EOS Data:'
write(90,*) ' r0 = ',r0
write(90,*) ' A = ',aj
write(90,*) ' B = ',bj
write(90,*) ' C = ',cj
write(90,*) ' R1 = ',r1
write(90,*) ' R2 = ',r2
write(90,*) ' W = ',wj
write(90,*) ' Cvg = ',cvg
write(90,*) ' Cpg = ',cpg
write(90,*) ' e0 = ',e0
write(90,*) ' '
write(90,*) ' Particle Data:'
write(90,*) ' pmass = ',pmass
write(90,*) ' rop = ',rop
write(90,*) ' rdp = ',rdp
write(90,*) ' dip = ',dip
write(90,*) ' mu = ',mu
write(90,*) ' tcon = ',tcon
write(90,*) ' ppr = ',ppr
write(90,*) ' p0mas = ',p0mas
write(90,*) ' hevol = ',hevol
write(90,*) ' pvol = ',pvol
write(90,*) ' cvol = ',cvol
54
write(90,*) ' alf1 = ',alf1
write(90,*) ' alf2 = ',alf2
write(90,*) ' '
write(90,*) ' Other Data:'
write(90,*) ' kap = ',kap
write(90,*) ' epsm = ',epsm
write(90,*) ' epsp = ',epsp
write(90,*) ' '
close(90)
write(*,*) ' '

write(*,*) ' Report file ready.'
write(*,*) ' '
c Grid Generation Section
dx = (x2 - x1)/(imax-1)
offs = 0.1d0
do i = 1,imax
x(i) = x1 + (i-1)*dx
c write(*,*) ' i = ',i,' x = ',x(i)
enddo
c Initial Conditions and Restart File Section
c Set initial conditions (no restart)
if (irst .eq. 0) then
c Set time zero primitive variables

do i = 1,imax-1
xc = c12*(x(i) + x(i+1))
c CPG EOS ICs
if (ieos .eq. 0) then
r(i) = 1d0/(1d0 + 3d0*dexp(-xc*xc))
p(i) = 1d0
u(i) = 0d0
z(i) = 0d0
c JWL EOS ICs
else if (ieos .eq. 1) then
r(i) = 1.2d0
c p(i) = 25d0*pamb/(1.00001d0 - dexp(-xc*xc))

55
p(i) = ((x2-xc)*(40d0*pamb/(1.00001d0 - dexp(-xc*xc)))
& + x2*pamb)/x2
c write(70,*) xc,' ',p(i)
c if (xc .lt. offs) then

c p(i) = fct*(xc-offs)*(xc-offs) + pamb
c else
c p(i) = pamb
c endif
u(i) = 0d0
z(i) = 0d0
c Hayes-I/JWL EOS ICs
r(i) = r0
c p(i) = 25d0*pamb/(1.00001d0 - dexp(-xc*xc))
c p(i) = ((x2-xc)*(25d0*pamb/(1.00001d0 - dexp(-xc*xc)))

c & + x2*pamb)/x2
c p(i) = pamb
c HMX or NM
p(i) = 2d0*pcj*dexp(-xc*xc/0.001d0/0.001d0) + pamb
u(i) = 0d0
z(i) = 0d0
tk(i) = tk0
c Hayes-II/JWL EOS ICs
r(i) = r0
c HMX/RDX/NM
c p(i) = 2d0*pcj*dexp(-xc*xc/0.004d0/0.004d0) + pamb
if (i .le. 100) then

p(i) = 5d0*pcj + pamb
else
p(i) = pamb
endif
c NM
c p(i) = 2d0*pcj*dexp(-xc*xc/0.0005d0/0.0005d0) + pamb
u(i) = 0d0
z(i) = 0d0
tk(i) = (p(i) - pamb)/cvs/gh + tk0

56
else
write(*,*) ' '
write(*,*) ' Unknown EOS'
write(*,*) ' '
stop
endif
enddo
c Particle ICs
if (ipar .eq. 1) then
c Check particle bounds

if (xp1 .lt. x1 .or. xp2 .gt. x2) then
write(*,*) ' '
write(*,*) ' Particle X limits are wrong.'
write(*,*) ' '
stop
endif
dxp = (xp2 - xp1)/(npar - 1)

do np = 1,npar
px(np) = xp1 + (np-1)*dxp
pu(np) = 0d0
ptk(np) = tk0
pa(np) = 0d0
pq(np) = 0d0
c write(*,*) px(np),' ',pu(np),' ',pa(np)
enddo
c pause
write(*,*) ' '
write(*,*) ' Particles ready.'
write(*,*) ' '
endif
else if (irst .eq. 1) then
c Read the restart file
write(*,*) ' Reading restart file.'
open(40,file='restart.data',form='unformatted')
read(40) nstart
read(40) nfil
read(40) time
do i = 1,imax-1
read(40) r(i),p(i),u(i),z(i)
enddo
close(40)
else
write(*,*) ' '
write(*,*) ' Unknown restart option.'
57
write(*,*) ' '
endif
c Compute initial derived flow variables for the cells
do i = 1,imax-1
c CPG EOS internal energy and pressure derivatives
ei(i) = p(i)/r(i)/gam1 - z(i)*qdet0
dpdr = gam1*ei(i) + gam1*z(i)*qdet0

dpde = gam1*r(i)
dpdz = gam1*r(i)*qdet0
c JWL EOS internal energy and pressure derivatives
rht = r(i)/r0
rhti = 1d0/rht
ri = 1d0/r(i)
tmp = p(i) - aj*(1d0 - wr1*r(i))*dexp(-rh1*ri)

& - bj*(1d0 - wr2*r(i))*dexp(-rh2*ri)
ei(i) = tmp/wj*ri - z(i)*qdet0
tmp = aj*(rh1*ri*ri - wj*ri - wj/rh1)*dexp(-rh1*ri)

tmp = tmp + bj*(rh2*ri*ri - wj*ri - wj/rh2)*dexp(-rh2*ri)
dpdr = tmp + wj*ei(i) + wj*z(i)*qdet0

dpde = wj*r(i)
dpdz = wj*r(i)*qdet0
c Hayes-I/JWL EOS internal energy and pressure derivatives
ra = r(i)
ra2 = ra*ra
za = z(i)
rz = ra*za
omz = 1d0 - za
c Solid phase limit

if (za .le. ztol1) then
ei(i) = p(i)/gh + beta*r0/ra + t4*((ra/r0)**alfa) - t7
dpdr = beta*r0*gh/ra2
58
& - alfa*gh*t4*(ra**(alfa-1d0))/(r0**alfa)
dpde = gh
c Mixed phases
else if (ztol1 .lt. za .and. za .lt. ztol2) then
c Evaluate denominator functions

bot = omz/gh + 1d0/wj/ra
if (bot .lt. 1d-10) then
write(*,*) ' '
write(*,*) ' Zero denonimator term.'
write(*,*) ' '
stop
endif
bot2 = bot*bot
botr = -1d0/wj/ra2
c Evaluate numerator functions

top(2) = omz - r0/ra
top(3) = (omz**nh)*((ra/r0)**alfa)
top(4) = omz
top(5) = (1d0/wj/ra - za/rh1)*dexp(-rh1/rz)
top(7) = za
c Compute internal energy

ei(i) = bot*p(i)
do nn = 2,7
ei(i) = ei(i) - c(nn)*top(nn)
enddo
top(1) = ei(i)
c Compute derivatives for numerator functions

topr(1) = 0d0
topr(2) = r0/ra2
topr(3) = alfa/r0*(omz**nh)*((ra/r0)**(alfa-1d0))
topr(4) = 0d0
topr(5) = (rh1/wj/rz - 1d0/wj - 1d0)*dexp(-rh1/rz)/ra2
topr(7) = 0d0
c Compute density and internal energy derivatives of pressure

dpdr = 0d0
do nn = 1,7
dpdr = dpdr + c(nn)*(bot*topr(nn) - botr*top(nn))
enddo
dpdr = dpdr/bot2
dpde = 1d0/bot
c Gas phase limit

else
ei(i) = p(i)/wj/ra
& - aj*(1d0/wj/ra - 1d0/rh1)*dexp(-rh1/ra)
& - bj*(1d0/wj/ra - 1d0/rh2)*dexp(-rh2/ra)
& - qdet0 - e0
59
dpdr = wj*ei(i)
& + aj*(rh1/ra2 - wj/ra - wj/rh1)*dexp(-rh1/ra)
& + bj*(rh2/ra2 - wj/ra - wj/rh2)*dexp(-rh2/ra)
& + wj*(qdet0 + e0)
dpdz = aj*(rh1/ra - wj - wj*ra/rh1)*dexp(-rh1/ra)

& + bj*(rh2/ra - wj - wj*ra/rh2)*dexp(-rh2/ra)
& + ra*wj*(qdet0 + e0)
dpde = wj*ra
endif
c Hayes-II/JWL EOS internal energy and pressure derivatives
ra = r(i)
ra2 = ra*ra
za = z(i)
rz = ra*za
omz = 1d0 - za
c Solid phase limit

ei(i) = p(i)/gh + beta*r0/ra + t4*((ra/r0)**alfa) - t7

& - h1/gh/nh*(((ra/r0)**nh) - 1d0)
dpdr = beta*r0*gh/ra2
& - alfa*gh*t4*(ra**(alfa-1d0))/(r0**alfa)
& + h1/r0*((ra/r0)**nhm1)
dpde = gh
c Mixed phases

write(*,*) ' '
write(*,*) ' '
stop
endif
bot2 = bot*bot
botr = -1d0/wj/ra2

top(4) = omz
60
top(7) = za
top(8) = (omz**nhp1)*((ra/r0)**nh) + za - 1d0

ei(i) = bot*p(i)
do nn = 2,8
ei(i) = ei(i) - c(nn)*top(nn)
enddo
top(1) = ei(i)

topr(1) = 0d0
topr(2) = r0/ra2
topr(4) = 0d0
topr(7) = 0d0
topr(8) = nh/r0*(omz**nhp1)*((ra/r0)**nhm1)

dpdr = 0d0
do nn = 1,8
enddo
dpdr = dpdr/bot2
dpde = 1d0/bot
c Gas phase limit

else
ei(i) = p(i)/wj/ra
& - qdet0 - e0
dpdr = wj*ei(i)
& + wj*(qdet0 + e0)

dpde = wj*ra
endif
else
write(*,*) ' '
write(*,*) ' '
stop
endif
c Compute the speed of sound

61
if (dpdr .lt. 0d0) dpdr = dabs(dpdr)
a2 = dpdr + p(i)*dpde/r(i)/r(i)
if (a2 .lt. 0d0) then

write(*,*) ' '
write(*,*) ' Negative initial squared sound speed!'
write(*,*) ' i = ',i
write(*,*) ' '
stop
endif
a(i) = dsqrt(a2)
c Initial reaction rate
c Floor on 1 - z near 0
if (z(i) .gt. ztol2) then
omz = 0d0
else
omz = 1d0 - z(i)
endif
c Test Rate 1
c rxr(i) = rk1*dsqrt(omz)
c if (p(i,j) - 1d9 .lt. 0d0) rxr(i) = 0d0
c if (p(i,j) - 1d9 .eq. 0d0) rxr(i) = 0.5d0*rxr(i)
c CPG Test Rate

c rxr(i) = rk*omz*dexp(-eact*r(i)/p(i)) - rxmin
c HMX Test Rate

c rxr(i) = rk1*(omz**zexp)*((p(i)/pcj)**pexp) - rxmin
c if (rxr(i) .lt. 0d0) rxr(i) = 0d0
c NM Test Rate
c rxr(i) = (rk1*dexp(-th1/tk(i))*omz
c & + rk2*dexp(-th2/tk(i))*z(i))*(omz**zexp) - rxmin
c RDX Test Rate

rxr(i) = rk1*(omz**zexp)*((p(i)/pcj)**pexp) - rxmin
if (rxr(i) .lt. 0d0) rxr(i) = 0d0
enddo
c Write the initial conditions files

if (irst .eq. 0) then
open(21,file='heic.dat',form='formatted')
70 format(1x,d12.6,1x,d12.6,1x,d12.6,1x,d12.6,1x,d12.6,1x,d12.6,
& 1x,d12.6,1x,d12.6)
72 format(1x,d12.6,1x,d12.6,1x,d12.6,1x,d12.6,1x,d12.6,1x,d12.6,
& 1x,d12.6,1x,d12.6,1x,d12.6)
do i = 1,imax-1
xc = c12*(x(i) + x(i+1))
write(21,72) xc,r(i),u(i),p(i),z(i),ei(i),a(i),rxr(i),tk(i)
62
enddo
close(21)
write(*,*) ' ICs ready.'
write(*,*) ' '

open(21,file='paic.dat',form='formatted')
do np = 1,npar
write(21,*) px(np),' ',0d0,' ',pu(np)
enddo
close(21)
endif
endif
pause
c Set the internal energy correction and scale variables
e0cr = 0d0
eta = 0.999d0
c Main Solver Loop
do while (n .lt. nstp .and. time .lt. tend)
c Allocate particles to cells
do np = 1,npar
pcel(np) = int((px(np) - x1)/dx) + 1
c write(*,*) ' px(',np,') = ',px(np)
c write(*,*) ' pcel(',np,') = ',pcel(np)
c write(*,*) ' '
enddo
pum = 0d0
pam = 0d0
do np = 1,npar
if (dabs(pu(np)) .gt. pum) pum = dabs(pu(np))
if (dabs(pa(np)) .gt. pam) pam = dabs(pa(np))
enddo
endif
c Compute time step
dt = 1d2
do i = 1,imax-1
dx = x(i+1) - x(i)
dt0 = dx/(dabs(u(i)) + a(i))
63
dt1 = dx/(dabs(u(i)) + pum)
dt0 = min(dt0,dt1)
c dt1 = 2d1*dx/pam
c dt0 = min(dt0,dt1)
endif
if (dt0 .lt. dt) dt = dt0
enddo
dt = cfl*dt
dt = min(dt,dtmx)
if (idbg1 .eq. 1) then

write(*,*) ' dt = ',dt
write(*,*) ' '
endif
c Set boundary conditions

c Symmetric at x = 0
r(0) = r(1)
u(0) = -u(1)
p(0) = p(1)
z(0) = z(1)
ei(0) = ei(1)
c Fixed at x = xmax
c r(imax) = 1d0
c u(imax) = 0d0
c p(imax) = 1d0
c z(imax) = 0d0
c ei(imax) = p(imax)/r(imax)/gam1
c Extrapolated at x = xmax
r(imax) = r(imax-1)
u(imax) = u(imax-1)
p(imax) = p(imax-1)
z(imax) = z(imax-1)
ei(imax) = ei(imax-1)
if (idbg1 .eq. 1) then

write(*,*) ' BCs:'
write(*,*) ' r(0) = ',r(0)
write(*,*) ' u(0) = ',u(0)
write(*,*) ' p(0) = ',p(0)
write(*,*) ' z(0) = ',z(0)
write(*,*) ' ei(0) = ',ei(0)
write(*,*) ' '
write(*,*) ' r(imax) = ',r(imax)
write(*,*) ' u(imax) = ',u(imax)
write(*,*) ' p(imax) = ',p(imax)
write(*,*) ' z(imax) = ',z(imax)
write(*,*) ' ei(imax) = ',ei(imax)
write(*,*) ' '
endif
c Minimum reaction rate taken at cell imax-1
64
if (z(imax-1) .gt. ztol2) then
omz = 0d0
else
omz = 1d0 - z(imax-1)
endif
c HMX or RDX Test

rxmin = rk1*(omz**zexp)*((p(imax-1)/pcj)**pexp)
c NM Test
c rxmin = (rk1*dexp(-th1/tk(imax-1))*omz
c & + rk2*dexp(-th2/tk(imax-1))*z(imax-1))*(omz**zexp)
c write(*,*) ' rxmin = ',rxmin

c write(*,*) ' rxr = ',rxr(imax-1)
c Compute conservative variables; assemble source terms
do i = 1,imax-1
qv(i,1) = r(i)
qv(i,2) = r(i)*u(i)
et = ei(i) + 0.5d0*u(i)*u(i)
qv(i,3) = r(i)*et
qv(i,4) = r(i)*z(i)
c Compute the source vectors
c Geometric
xc = c12*(x(i) + x(i+1))
sg(i,1) = -r(i)*u(i)/xc
sg(i,2) = -r(i)*u(i)*u(i)/xc
sg(i,3) = -u(i)*(r(i)*et + p(i))/xc
sg(i,4) = -r(i)*u(i)*z(i)/xc
c Reaction rate
if (z(i) .gt. ztol2) then
omz = 0d0
else
omz = 1d0 - z(i)
endif
c CPG Test Rate

c rxr(i) = rk*omz*dexp(-eact*r(i)/p(i)) - rxmin
c HMX Test Rate

c rxr(i) = rk1*(omz**zexp)*((p(i)/pcj)**pexp) - rxmin
c NM Test Rate
65
c rxr(i) = (rk1*dexp(-th1/tk(i))*omz
c & + rk2*dexp(-th2/tk(i))*z(i))*(omz**zexp) - rxmin
c RDX Test Rate

rxr(i) = rk1*(omz**zexp)*((p(i)/pcj)**pexp) - rxmin
if (rxr(i) .lt. 0d0) rxr(i) = 0d0
c Reaction rate terms

srx(i,1) = 0d0
srx(i,2) = 0d0
srx(i,3) = 0d0
srx(i,4) = r(i)*rxr(i)
c Particle phase
sp(i,1) = 0d0
sp(i,2) = 0d0
sp(i,3) = 0d0
sp(i,4) = 0d0
enddo
c Compute particle phase coupling terms
do np = 1,npar
c Momentum
sp(pcel(np),2) = sp(pcel(np),2) - p0mas*pa(np)
c Energy
sp(pcel(np),3) = sp(pcel(np),3) - pq(np)
enddo
endif
c Compute the total source vector
do i = 1,imax-1
c if (sp(i,2) .ne. 0d0) then

c write(*,*) ' i = ',i,' sp = ',sp(i,2)
c endif
do m = 1,4
s(i,m) = igeo*sg(i,m) + irxn*srx(i,m) + ipar*sp(i,m)
enddo
if (idbgs .eq. 1) then

write(*,*) ' i = ',i
write(*,*) ' q1 = ',qv(i,1)
write(*,*) ' q2 = ',qv(i,2)
write(*,*) ' q3 = ',qv(i,3)
write(*,*) ' q4 = ',qv(i,4)
write(*,*) ' '
66
write(*,*) ' s1 = ',s(i,1)
write(*,*) ' s2 = ',s(i,2)
write(*,*) ' s3 = ',s(i,3)
write(*,*) ' s4 = ',s(i,4)
write(*,*) ' '
pause
endif
enddo
ccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc
c Compute the numerical flux at each grid point
71 format(2x,d12.6,2x,d12.6,2x,d12.6,2x,d12.6)
do i = 1,imax
c Left interface variables

if (i .eq. 1) then
c First order at the boundary
rl = r(i-1)
ul = u(i-1)
pl = p(i-1)
zl = z(i-1)
rr = r(i)
ur = u(i)
pr = p(i)
zr = z(i)
else if (2 .le. i .and. i .le. imax-1) then

c Higher-order
if (ilim .eq. 0) then
c Hossaini limiting strategy

dqwr = r(i-1) - r(i-2)
dqer = r(i) - r(i-1)
dqir = r(i+1) - r(i)
phir = c14*(2d0*dqwr*dqer + eps)
& /(dqwr*dqwr + dqer*dqer + eps)
dqwu = u(i-1) - u(i-2)

dqeu = u(i) - u(i-1)
dqiu = u(i+1) - u(i)
phiu = c14*(2d0*dqwu*dqeu + eps)
& /(dqwu*dqwu + dqeu*dqeu + eps)
dqwp = p(i-1) - p(i-2)

dqep = p(i) - p(i-1)
dqip = p(i+1) - p(i)
phip = c14*(2d0*dqwp*dqep + eps)
& /(dqwp*dqwp + dqep*dqep + eps)
dqwz = z(i-1) - z(i-2)

dqez = z(i) - z(i-1)
dqiz = z(i+1) - z(i)
phiz = c14*(2d0*dqwz*dqez + eps)
67
& /(dqwz*dqwz + dqez*dqez + eps)
c Density
rl = r(i-1) + iext*phir*(epsm*dqwr + epsp*dqer)
rr = r(i) - iext*phir*(epsm*dqir + epsp*dqer)
c Velocity
ul = u(i-1) + iext*phiu*(epsm*dqwu + epsp*dqeu)
ur = u(i) - iext*phiu*(epsm*dqiu + epsp*dqeu)
c Pressure
pl = p(i-1) + iext*phip*(epsm*dqwp + epsp*dqep)
pr = p(i) - iext*phip*(epsm*dqip + epsp*dqep)
c Rx Progress
zl = z(i-1) + iext*phiz*(epsm*dqwz + epsp*dqez)
zr = z(i) - iext*phiz*(epsm*dqiz + epsp*dqez)
else if (ilim .eq. 1) then
c Hirsch limiting strategy

dra = r(i+1) - r(i)
drb = r(i) - r(i-1)
drc = r(i-1) - r(i-2)
drd = drb - drc
dre = dra - drb
dua = u(i+1) - u(i)

dub = u(i) - u(i-1)
duc = u(i-1) - u(i-2)
dud = dub - duc
due = dua - dub
dpa = p(i+1) - p(i)

dpb = p(i) - p(i-1)
dpc = p(i-1) - p(i-2)
dpd = dpb - dpc
dpe = dpa - dpb
dza = z(i+1) - z(i)

dzb = z(i) - z(i-1)
dzc = z(i-1) - z(i-2)
dzd = dzb - dzc
dze = dza - dzb
c Check monotonicity
imon = 1
if (dra*drb .lt. 0d0) imon = 0
if (drb*drc .lt. 0d0) imon = 0
if (dua*dub .lt. 0d0) imon = 0
if (dub*duc .lt. 0d0) imon = 0
if (dpa*dpb .lt. 0d0) imon = 0
if (dpb*dpc .lt. 0d0) imon = 0
if (dza*dzb .lt. 0d0) imon = 0
if (dzb*dzc .lt. 0d0) imon = 0
if (imon .eq. 0) then

68
c First-order interface is non-monotonic
rl = r(i-1)
ul = u(i-1)
pl = p(i-1)
zl = z(i-1)
rr = r(i)
ur = u(i)
pr = p(i)
zr = z(i)
else
c First-order interface is monotonic

denm = drb*drb + drc*drc + eps
phi = (drb*drd + eps)/denm
vhi = (drc*drd + eps)/denm
rl = r(i-1) + iext*(epsm*phi*drc
& + epsp*vhi*drb)
denm = dra*dra + drb*drb + eps

phi = (drb*dre + eps)/denm
vhi = (dra*dre + eps)/denm
rr = r(i) - iext*(epsm*phi*dra
& + epsp*vhi*drb)
denm = dub*dub + duc*duc + eps

phi = (dub*dud + eps)/denm
vhi = (duc*dud + eps)/denm
ul = u(i-1) + iext*(epsm*phi*duc
& + epsp*vhi*dub)
denm = dua*dua + dub*dub + eps

phi = (dub*due + eps)/denm
vhi = (dua*due + eps)/denm
ur = u(i) - iext*(epsm*phi*dua
& + epsp*vhi*dub)
denm = dpb*dpb + dpc*dpc + eps

phi = (dpb*dpd + eps)/denm
vhi = (dpc*dpd + eps)/denm
pl = p(i-1) + iext*(epsm*phi*dpc
& + epsp*vhi*dpb)
denm = dpa*dpa + dpb*dpb + eps

phi = (dpb*dpe + eps)/denm
vhi = (dpa*dpe + eps)/denm
pr = p(i) - iext*(epsm*phi*dpa
& + epsp*vhi*dpb)
denm = dzb*dzb + dzc*dzc + eps

phi = (dzb*dzd + eps)/denm
vhi = (dzc*dzd + eps)/denm
zl = z(i-1) + iext*(epsm*phi*dzc
& + epsp*vhi*dzb)

69
denm = dza*dza + dzb*dzb + eps
phi = (dzb*dze + eps)/denm
vhi = (dza*dze + eps)/denm
zr = z(i) - iext*(epsm*phi*dza
& + epsp*vhi*dzb)
endif
else
write(*,*) ' '
write(*,*) ' Unknown limiting strategy'
write(*,*) ' '
endif
c Set ceiling on zl, zr

zl = min(zl,1d0)
zr = min(zr,1d0)
else
c First order at imax

rl = r(i-1)
ul = u(i-1)
pl = p(i-1)
zl = z(i-1)
rr = r(i)
ur = u(i)
pr = p(i)
zr = z(i)
endif
c zzl(i) = zl
c zzr(i) = zr
c Final monotonicity check

imon = 0
rat = (r(i) - r(i-1))*(rr - rl)
if (rat .lt. 0d0) imon = 1
rat = (u(i) - u(i-1))*(ur - ul)
rat = (p(i) - p(i-1))*(pr - pl)
rat = (z(i) - z(i-1))*(zr - zl)
c Set first order interface

if (imon .ne. 0) then
rl = r(i-1)
rr = r(i)
ul = u(i-1)
ur = u(i)
pl = p(i-1)
pr = p(i)
zl = z(i-1)
70
zr = z(i)
endif
c Calculate internal energy

c CPG EOS
el = pl/gam1/rl - zl*qdet0
er = pr/gam1/rr - zr*qdet0
c JWL EOS
rht = rl/r0
rhti = 1d0/rht
ri = 1d0/rl
tmp = pl - aj*(1d0 - wr1*rl)*dexp(-rh1*ri)
& - bj*(1d0 - wr2*rl)*dexp(-rh2*ri)
el = tmp*ri/wj - zl*qdet0
rht = rr/r0
rhti = 1d0/rht
ri = 1d0/rr
tmp = pr - aj*(1d0 - wr1*rr)*dexp(-rh1*ri)
& - bj*(1d0 - wr2*rr)*dexp(-rh2*ri)
er = tmp*ri/wj - zr*qdet0
c Hayes-I/JWL EOS
c Left of interface; set arguments
ra = rl
za = zl
rz = ra*za
omz = 1d0 - za
c Solid phase limit

el = pl/gh + beta*r0/ra + t4*((ra/r0)**alfa) - t7
c Mixed phases
c Evaluate denominator function

write(*,*) ' '
71
write(*,*) ' '
stop
endif

top(4) = omz
top(7) = za
el = bot*pl
do nn = 2,7
el = el - c(nn)*top(nn)
enddo
c Gas phase limit

else
el = pl/wj/ra
& - qdet0 - e0
endif
c Right of interface; set arguments

ra = rr
za = zr
rz = ra*za
omz = 1d0 - za
c Solid phase limit

er = pr/gh + beta*r0/ra + t4*((ra/r0)**alfa) - t7
c Mixed phases

write(*,*) ' '
write(*,*) ' '
stop
endif

top(4) = omz
top(7) = za
72
er = bot*pr
do nn = 2,7
er = er - c(nn)*top(nn)
enddo
c Gas phase limit

else
er = pr/wj/ra
& - qdet0 - e0
endif
c Hayes-II/JWL EOS
c Left of interface; set arguments
ra = rl
za = zl
rz = ra*za
omz = 1d0 - za
c Solid phase limit

el = pl/gh + beta*r0/ra + t4*((ra/r0)**alfa) - t7

& - h1/gh/nh*(((ra/r0)**nh) - 1d0)
c Mixed phases

write(*,*) ' '
write(*,*) ' '
stop
endif

top(4) = omz
top(7) = za
el = bot*pl
do nn = 2,8
73
el = el - c(nn)*top(nn)
enddo
c Gas phase limit

else
el = pl/wj/ra
& - qdet0 - e0
endif
c Right of interface; set arguments

ra = rr
za = zr
rz = ra*za
omz = 1d0 - za
c Solid phase limit

er = pr/gh + beta*r0/ra + t4*((ra/r0)**alfa) - t7

& - h1/gh/nh*(((ra/r0)**nh) - 1d0)
c Mixed phases

write(*,*) ' '
write(*,*) ' '
stop
endif

top(4) = omz
top(7) = za

er = bot*pr
do nn = 2,8
er = er - c(nn)*top(nn)
enddo
c Gas phase limit

else
er = pr/wj/ra
74
& - qdet0 - e0
endif
else
write(*,*) ' '
write(*,*) ' '
stop
endif
c Total energy/mass
eel = el + 0.5d0*ul*ul
hhl = eel + pl/rl
eer = er + 0.5d0*ur*ur
hhr = eer + pr/rr
c if (imon .ne. 0) then

c write(*,*) ' '
c write(*,*) ' Monotonicity violation - ',imon
c write(*,*) ' i = ',i
c write(*,*) ' '
c write(*,*) ' r(i-1) = ',r(i-1)
c write(*,*) ' rl = ',rl
c write(*,*) ' rr = ',rr
c write(*,*) ' r(i) = ',r(i)
c write(*,*) ' '
c write(*,*) ' u(i-1) = ',u(i-1)
c write(*,*) ' ul = ',ul
c write(*,*) ' ur = ',ur
c write(*,*) ' u(i) = ',u(i)
c write(*,*) ' '
c write(*,*) ' p(i-1) = ',p(i-1)
c write(*,*) ' pl = ',pl
c write(*,*) ' pr = ',pr
c write(*,*) ' p(i) = ',p(i)
c write(*,*) ' '
c pause
c endif
c80 format(2x,d12.6,2x,d12.6,2x,d12.6)
c if (n .eq. 177) then
c write(25,80) r(i-1),rr,r(i)
c endif
c Roe averages
if (iav .eq. 1) then
sqrl = dsqrt(rl)
sqrr = dsqrt(rr)
rsumi = 1d0/(sqrl + sqrr)
rav = sqrl*sqrr
uav = (sqrl*ul + sqrr*ur)*rsumi
zav = (sqrl*zl + sqrr*zr)*rsumi
zav = min(zav,1d0)
75
eav = (sqrl*el + sqrr*er)*rsumi
hav = (sqrl*hhl + sqrr*hhr)*rsumi
else
c Test arithmetic averages

rav = 0.5d0*(rl + rr)
uav = 0.5d0*(ul + ur)
zav = 0.5d0*(zl + zr)
eav = 0.5d0*(el + er)
hav = 0.5d0*(hhl + hhr)
endif
pav = rav*(hav - eav - 0.5d0*uav*uav)
c Calculate averaged pressure derivatives

c CPG EOS
dpdr = gam1*eav + gam1*zav*qdet0
dpde = gam1*rav
dpdz = gam1*rav*qdet0
c JWL EOS
ri = 1d0/rav
dpdr = tmp + wj*eav + wj*zav*qdet0

dpde = wj*rav
dpdz = wj*rav*qdet0
c Hayes-I/JWL EOS
ra = rav
ra2 = ra*ra
ea = eav
za = zav
rz = ra*za
omz = 1d0 - za
c Solid phase limit

dpdr = beta*r0*gh/ra2 - alfa*gh*t4*(ra**(alfa-1d0))

& / r0**alfa
dpdz = gh*ea - beta*r0*gh/ra + alfa*gh*t4

& * ((ra/r0)**alfa)
76
dpde = gh
c Mixed phases

write(*,*) ' '
write(*,*) ' '
stop
endif
bot2 = bot*bot
botr = -1d0/wj/ra2
botz = -1d0/gh

top(1) = ea
top(4) = omz
top(7) = za

topr(1) = 0d0
topr(2) = r0/ra2
topr(4) = 0d0
topr(7) = 0d0
topz(1) = 0d0
topz(2) = -1d0
topz(3) = -nh*((omz*ra/r0)**alfa)
topz(4) = -1d0
topz(5) = (rh1/wj/rz/rz - 1d0/rz - 1d0/rh1)*dexp(-
rh1/rz)
rh2/rz)
topz(7) = 1d0

dpdr = 0d0
dpdz = 0d0
do nn = 1,7
dpdz = dpdz + c(nn)*(bot*topz(nn) - botz*top(nn))
enddo
dpdr = dpdr/bot2
dpdz = dpdz/bot2
dpde = 1d0/bot

77
c Gas phase limit
else
dpdr = wj*ei(i)
& + wj*(qdet0 + e0)

dpde = wj*ra
endif
c Hayes-II/JWL EOS
ra = rav
ra2 = ra*ra
ea = eav
za = zav
rz = ra*za
omz = 1d0 - za
c Solid phase limit


& / r0**alfa
& + h1/r0*((ra/r0)**nhm1)

& * ((ra/r0)**alfa)
& + h1/nh*(1d0 - nhp1*((ra/r0)**nh))
dpde = gh
c Mixed phases

write(*,*) ' '
write(*,*) ' '
stop
endif
bot2 = bot*bot
botr = -1d0/wj/ra2
botz = -1d0/gh

78
top(1) = ea
top(4) = omz
top(7) = za

topr(1) = 0d0
topr(2) = r0/ra2
topr(4) = 0d0
topr(7) = 0d0
topr(8) = nh/r0*(omz**nhp1)*((ra/r0)*nhm1)
topz(1) = 0d0
topz(2) = -1d0
topz(4) = -1d0
rh1/rz)
rh2/rz)
topz(7) = 1d0
topz(8) = 1d0 - nhp1*((ra/r0*omz)**nh)

dpdr = 0d0
dpdz = 0d0
do nn = 1,8
enddo
dpdr = dpdr/bot2
dpdz = dpdz/bot2
dpde = 1d0/bot
c Gas phase limit

else
dpdr = wj*ei(i)
& + wj*(qdet0 + e0)

dpde = wj*ra
endif

79
else
write(*,*) ' '
write(*,*) ' '
stop
endif
c Calculate averaged speed of sound

a2 = dpdr + pav*dpde/rav/rav
if (a2 .lt. 0d0) then

write(*,*) ' a2 < 0 !'
write(*,*) ' i = ',i
write(*,*) ' eav = ',eav
write(*,*) ' el = ',el
write(*,*) ' er = ',er
write(*,*) ' rav = ',rav
write(*,*) ' pav = ',pav
write(*,*) ' pl = ',pl
write(*,*) ' pr = ',pr
write(*,*) ' zav = ',zav
write(*,*) ' dpdr = ',dpdr
write(*,*) ' dpde = ',dpde
write(*,*) ' '
write(*,*) ' r+1 = ',rp(i)
write(*,*) ' u+1 = ',up(i)
write(*,*) ' p+1 = ',pp(i)
write(*,*) ' z+1 = ',zp(i)
write(*,*) ' '
write(*,*) ' r-1 = ',rp(i-1)
write(*,*) ' u-1 = ',up(i-1)
write(*,*) ' p-1 = ',pp(i-1)
write(*,*) ' z-1 = ',zp(i-1)
write(*,*) ' '
stop
endif
aav = dsqrt(a2)
if (idbgf .eq. 1) then

write(*,*) ' rav = ',rav
write(*,*) ' uav = ',uav
write(*,*) ' zav = ',zav
write(*,*) ' eav = ',eav
write(*,*) ' hav = ',hav
write(*,*) ' pav = ',pav
write(*,*) ' aav = ',aav
write(*,*) ' '
endif
c Eigenvalues
aeg(1) = dabs(uav - aav)
aeg(2) = dabs(uav)
aeg(3) = dabs(uav)
aeg(4) = dabs(uav + aav)

80
write(*,*) ' aeg1 = ',aeg(1)
write(*,*) ' aeg2 = ',aeg(2)
write(*,*) ' aeg3 = ',aeg(3)
write(*,*) ' aeg4 = ',aeg(4)
write(*,*) ' '
pause
endif
c Right eigenvectors
evr(1,1) = 1d0
evr(1,2) = 1d0
evr(1,3) = 1d0
evr(1,4) = 1d0
evr(2,1) = uav - aav

evr(2,2) = uav
evr(2,3) = uav
evr(2,4) = uav + aav
evr(3,1) = hav - uav*aav

evr(3,2) = hav - rav*a2/dpde + zav*dpdz/dpde
evr(3,3) = hav - rav*a2/dpde + (zav - 1d0)*dpdz/dpde
evr(3,4) = hav + uav*aav
evr(4,1) = zav
evr(4,2) = 0d0
evr(4,3) = 1d0
evr(4,4) = zav

write(*,*) 'EVR:'
write(*,71) evr(1,1),evr(1,2),evr(1,3),evr(1,4)
write(*,*) ' '
endif
c |R|
detr = -2d0*rav*a2*aav/dpde
c Compute primitive variables differences

delr = rr - rl
delv = ur - ul
delp = pr - pl
delz = zr - zl
c Compute characteristic wave magnitudes

omz = 1d0 - zav
cwm(1) = c12*(delp/aav/aav - rav*delv/aav)
cwm(2) = omz*(delr - delp/aav/aav) - rav*delz
cwm(3) = zav*(delr - delp/aav/aav) + rav*delz
cwm(4) = c12*(delp/aav/aav + rav*delv/aav)
c Compute R |eg| L dq
do l = 1,4
81
vn(l) = 0d0
do m = 1,4
vn(l) = vn(l) + aeg(m)*cwm(m)*evr(l,m)
enddo
enddo

write(*,*) ' vn1 = ',vn(1)
write(*,*) ' vn2 = ',vn(2)
write(*,*) ' vn3 = ',vn(3)
write(*,*) ' vn4 = ',vn(4)
write(*,*) ' '
endif
c Compute the Euler flux

fl(1) = rl*ul
fl(2) = rl*ul*ul + pl
fl(3) = rl*ul*hhl
fl(4) = rl*ul*zl
fr(1) = rr*ur
fr(2) = rr*ur*ur + pr
fr(3) = rr*ur*hhr
fr(4) = rr*ur*zr
c Compute numerical flux

do l = 1,4
fn(i,l) = 0.5d0*(fl(l) + fr(l) - vn(l))
enddo

write(*,*) ' FL:'
write(*,*) ' fl1 = ',fl(1)
write(*,*) ' fl2 = ',fl(2)
write(*,*) ' fl3 = ',fl(3)
write(*,*) ' fl4 = ',fl(4)
write(*,*) ' '
write(*,*) ' FR:'
write(*,*) ' fr1 = ',fr(1)
write(*,*) ' fr2 = ',fr(2)
write(*,*) ' fr3 = ',fr(3)
write(*,*) ' fr4 = ',fr(4)
write(*,*) ' '
write(*,*) ' FN:'
write(*,*) ' fn1 = ',fn(i,1)
write(*,*) ' fn2 = ',fn(i,2)
write(*,*) ' fn3 = ',fn(i,3)
write(*,*) ' fn4 = ',fn(i,4)
write(*,*) ' '
pause
endif
enddo
c End of flux calculation loop

82
c Advance the solution in time
do i = 1,imax-1
do l = 1,4
dqv(l) = dt/dx*(fn(i+1,l) - fn(i,l))
enddo
do l = 1,4
qvp(i,l) = qv(i,l) - dqv(l) + dt*s(i,l)
enddo
enddo
c Extract primitive variables

do i = 1,imax-1
rp(i) = qvp(i,1)
up(i) = qvp(i,2)/qvp(i,1)
etp(i) = qvp(i,3)/qvp(i,1)
zp(i) = qvp(i,4)/qvp(i,1)
zp(i) = min(zp(i),1d0)
zp(i) = max(zp(i),0d0)
if (zp(i) .lt. 1d-99) zp(i) = 0d0

if (zp(i) .ge. 0.99d0) zp(i) = 1d0
eip(i) = etp(i) - 0.5d0*up(i)*up(i)
tk(i) = tk0
if (rp(i) .le. 0d0) then

write(*,*) ' '
write(*,*) ' Negative/Zero density'
write(*,*) ' i = ',i
write(*,*) ' r = ',rp(i)
write(*,*) ' u = ',up(i)
write(*,*) ' e = ',etp(i)
write(*,*) ' z = ',zp(i)
write(*,*) ' '
write(*,*) ' Program STOP'
write(*,*) ' '
stop
endif
c If internal energy is negative, apply a fix

if (eip(i) .le. 0d0) then
c write(*,*) ' '

c write(*,*) ' Negative/Zero internal energy'
c write(*,*) ' i = ',i
c write(*,*) ' r = ',rp(i)
c write(*,*) ' u = ',up(i)
c write(*,*) ' E = ',etp(i)
c write(*,*) ' e = ',eip(i)
c write(*,*) ' z = ',zp(i)
83
c write(*,*) ' '
c write(*,*) ' iefx = ',iefx
if (iefx .eq. 0) then

c Absolute value |e| fix
eip(i) = dabs(eip(i))
else if (iefx .eq. 1) then

c Pressure estimation fix
c Estimate pressure using JWL EOS
pest = aj*dexp(-rh1/rp(i)) + bj*dexp(-rh2/rp(i))
& + cjh*(rp(i)**(1d0 + wj))
c Compute detonation e based on JWL pressure

eip(i) = 1d0/wj/rp(i)*
& ( cjh*(rp(i)**(1d0 + wj))
& + aj*wj*rp(i)/rh1*dexp(-rh1/rp(i))
& + bj*wj*rp(i)/rh2*dexp(-rh2/rp(i)) )
c write(*,*) ' '

c write(*,*) ' pest = ',pest
c write(*,*) ' eest = ',eip(i)
c pause

c Time-lagged velocity fix
eip(i) = etp(i) - 0.5d0*u(i)*u(i)

c Zero kinetic energy fix
eip(i) = etp(i)
else
write(*,*) ' '
write(*,*) ' Unknown iefx value.'
write(*,*) ' '
stop
endif
c pause
endif
c Calculate pressure and its derivatives
c CPG EOS
pp(i) = gam1*rp(i)*eip(i) + gam1*rp(i)*zp(i)*qdet0
dpdr = gam1*eip(i) + gam1*zp(i)*qdet0

dpde = gam1*rp(i)
dpdz = gam1*rp(i)*qdet0

84
c JWL EOS
rht = rp(i)/r0
rhti = 1d0/rht
ri = 1d0/rp(i)
tmp = aj*(1d0 - wr1*rp(i))*dexp(-rh1*ri)
tmp = tmp + bj*(1d0 - wr2*rp(i))*dexp(-rh2*ri)
pp(i) = tmp + wj*rp(i)*eip(i) + wj*rp(i)*zp(i)*qdet0

dpdr = tmp + wj*eip(i) + wj*zp(i)*qdet0

dpde = wj*rp(i)
dpdz = wj*rp(i)*qdet0
c Hayes-I/JWL EOS
ra = rp(i)
ra2 = ra*ra
ea = eip(i)
za = zp(i)
rz = ra*za
omz = 1d0 - za
c Solid phase limit

c Compute pressure and its derivatives

pp(i) = gh*(ea - beta*r0/ra - t4*((ra/r0)**alfa)
& + t7)

& /(r0**alfa)

& * ((ra/r0)**alfa)
dpde = gh
c Mixed phases

write(*,*) ' '
write(*,*) ' '
stop
endif
85
bot2 = bot*bot
botr = -1d0/wj/ra2
botz = -1d0/gh

top(1) = ea
top(4) = omz
top(7) = za

topr(1) = 0d0
topr(2) = r0/ra2
topr(4) = 0d0
topr(7) = 0d0
topz(1) = 0d0
topz(2) = -1d0
topz(4) = -1d0
rh1/rz)
rh2/rz)
topz(7) = 1d0

pp(i) = 0d0
dpdr = 0d0
dpdz = 0d0
do nn = 1,7
pp(i) = pp(i) + c(nn)*top(nn)
enddo
pp(i) = pp(i)/bot
dpdr = dpdr/bot2
dpdz = dpdz/bot2
dpde = 1d0/bot
c Gas phase limit

else

pp(i) = wj*ra*ea
& + aj*(1d0 - wj*ra/rh1)*dexp(-rh1/ra)
& + bj*(1d0 - wj*ra/rh2)*dexp(-rh2/ra)
& + wj*ra*(qdet0 + e0)
dpdr = wj*ea
86
& + wj*(qdet0 + e0)

dpde = wj*ra
endif
c Hayes-II/JWL EOS
ra = rp(i)
ra2 = ra*ra
ea = eip(i)
za = zp(i)
rz = ra*za
omz = 1d0 - za
c Solid phase limit


pp(i) = gh*(ea - beta*r0/ra - t4*((ra/r0)**alfa)
& + t7)
& + h1/nh*(((ra/r0)**nh) - 1d0)

& /(r0**alfa)
& + h1/r0*((ra/r0)**nhm1)

& * ((ra/r0)**alfa)
& + h1/nh*(1d0 - nhp1*((ra/r0)**nh))
dpde = gh
c Mixed phases

write(*,*) ' '
write(*,*) ' '
stop
endif
bot2 = bot*bot
botr = -1d0/wj/ra2
botz = -1d0/gh

87
top(1) = ea
top(4) = omz
top(7) = za

topr(1) = 0d0
topr(2) = r0/ra2
topr(4) = 0d0
topr(7) = 0d0
topr(8) = nh/r0*(omz**nhp1)*((ra/r0)**nhm1)
topz(1) = 0d0
topz(2) = -1d0
topz(4) = -1d0
rh1/rz)
rh2/rz)
topz(7) = 1d0
topz(8) = 1d0 - nhp1*((ra/r0*omz)**nh)

pp(i) = 0d0
dpdr = 0d0
dpdz = 0d0
do nn = 1,8
pp(i) = pp(i) + c(nn)*top(nn)
enddo
pp(i) = pp(i)/bot
dpdr = dpdr/bot2
dpdz = dpdz/bot2
dpde = 1d0/bot
c Gas phase limit

else

pp(i) = wj*ra*ea
& + aj*(1d0 - wj*ra/rh1)*dexp(-rh1/ra)
& + bj*(1d0 - wj*ra/rh2)*dexp(-rh2/ra)
& + wj*ra*(qdet0 + e0)
dpdr = wj*ea
& + wj*(qdet0 + e0)
88
dpde = wj*ra
endif
else
write(*,*) ' '
write(*,*) ' '
stop
endif
c Check for negative pressure

if (pp(i) .lt. 0d0) then
write(*,*) ' '
write(*,*) ' Negative pressure detected.'
write(*,*) ' i = ',i
write(*,*) ' r = ',rp(i)
write(*,*) ' u = ',up(i)
write(*,*) ' p = ',pp(i)
write(*,*) ' z = ',zp(i)
write(*,*) ' ea= ',ea
write(*,*) ' '
write(*,*) ' r-1 = ',rp(i-1)
write(*,*) ' u-1 = ',up(i-1)
write(*,*) ' p-1 = ',pp(i-1)
write(*,*) ' z-1 = ',zp(i-1)
write(*,*) ' '
write(*,*) ' Program STOP'
write(*,*) ' '
stop
endif
c Calculate the speed of sound

derv(i,1) = dpdr
derv(i,2) = dpde

a2 = dpdr + pp(i)*dpde/rp(i)/rp(i)
if (a2 .le. 0d0) then

write(*,*) ' '
write(*,*) ' Negative squared sound speed!'
write(*,*) ' i = ',i
write(*,*) ' dpdr = ',dpdr
write(*,*) ' dpde = ',dpde
write(*,*) ' pp = ',pp(i)
write(*,*) ' rp = ',rp(i)
write(*,*) ' a2 = ',a2
write(*,*) ' '
stop
endif
ap(i) = dsqrt(a2)
89
enddo
c Estimate mixture temperature Hayes-II/JWL EOS only
item = 0
item = 1
dtkmx = 0d0
denmx = 0d0
c First temperature estimate

do i = 1,imax-1
if (zp(i) .gt. ztol2) then
omz = 0d0
else
omz = 1d0 - zp(i)
endif
denm = cvs*omz + cvg*zp(i)

if (denm .gt. denmx) denmx = denm
de1 = 0d0
de2 = 0d0
de3 = 0d0
de4 = 0d0
de5 = 0d0
de6 = 0d0
c if (zp(i) .lt. 0.999d0) then

if (zp(i) .lt. ztol2) then
rs = omz*rp(i)
de1 = t4*(((rs/r0)**alfa) - 1d0)
de2 = beta*(1d0 - r0/rs)
endif
c if (zp(i) .gt. 0.001d0) then
if (zp(i) .gt. ztol1) then
rg = zp(i)*rp(i)
de3 = aj/rh1*dexp(-rh1/rg)
de4 = bj/rh2*dexp(-rh2/rg)
de5 = aj/rh1*dexp(-rh1/r0)
de6 = bj/rh2*dexp(-rh2/r0)
endif
numr = eip(i) - omz*(de1 - de2)

& - zp(i)*(de3 + de4 - de5 - de6 - qdet0
& + e0cr)
dtk(i) = numr/denm
c write(*,*) ' zp = ',zp(i)

c write(*,*) ' de1 = ',de1
c write(*,*) ' de2 = ',de2
c write(*,*) ' de3 = ',de3
c write(*,*) ' de4 = ',de4
c write(*,*) ' de5 = ',de5
90
c write(*,*) ' de6 = ',de6
c write(*,*) ' numr = ',numr,' dtk = ',dtk(i)
c pause
if (dtk(i) .lt. dtkmx) dtkmx = dtk(i)

enddo
c Check the temperature difference (is T < T0?)

if (dtkmx .lt. 0d0) then
item = -1
c Calculate the internal energy correction (fwded to next time level)

e0cr = dtkmx*denmx/eta
c Apply the temperature correction

do i = 1,imax-1
omz = 1d0 - zp(i)
denm = cvs*omz + cvg*zp(i)
dtk(i) = dtk(i) - e0cr/denm
enddo
endif
c Calculate the corrected temperature field

do i = 1,imax-1
tk(i) = tk0 + dtk(i) - dtk(imax-1)
c tk(i) = dtk(i)
enddo
endif
c Update particle properties and positions
do np = 1,npar
c Compute Reynolds number

ra = rp(pcel(np))*zp(pcel(np))
delu = up(pcel(np)) - pu(np)
adelu = dabs(delu)
if (adelu .lt. 1d-10) adelu = 1d-10
rep = dip*ra*adelu/mu
c write(*,*) ' rep = ',rep
c if (rep .le. 0d0) then

c write(*,*) ' '
c write(*,*) ' Rep <= 0!'
c write(*,*) ' cell = ',pcel(np)
c write(*,*) ' rp = ',rp(pcel(np))
c write(*,*) ' zp = ',zp(pcel(np))
c write(*,*) ' ra = ',ra
c write(*,*) ' delu = ',delu
c write(*,*) ' adelu = ',adelu
91
c write(*,*) ' '
c stop
c endif
c Compute particle accelerations

if (idrg .eq. 0) then
c Spray drag law

if (rep .lt. 1d-10) then
cdp = 0d0
else if (rep .le. 1d3) then
cdp = 24d0/rep*(1d0 + (rep**c23)/6d0)
else
cdp = 0.44d0
endif
pa(np) = c316*mu*cdp*rep/rop/rdp/rdp*delu
else if (idrg .eq. 1) then
c Rocket drag law

if (rep .lt. 1d-10) then
cd1 = 0d0
cd2 = 0d0
else
cd1 = 24d0/rep + 4.4d0/dsqrt(rep) + 0.42d0
cd2 = c43*(1.75d0 + 150d0*alf21/rep)/alf1
endif
if (alf2 .le. 0.08d0) then
cd0 = cd1
else if (0.08d0 .lt. alf2 .and. alf2 .lt. 0.45d0) then
cd0 = (0.45d0-alf2)*cd1 + (alf2-0.08d0)*cd2
cd0 = cd0/0.37d0
else if (alf2 .gt. 0.45d0) then
cd0 = cd2
endif
c Mach correction
if (imach .eq. 1) then
mach = (adelu/ap(i))**4.63d0
cdp = cd0*(1d0 + dexp(-0.427d0/mach))
else
cdp = cd0
endif
pa(np) = c18*pi*dip*dip*cdp*ra*adelu*delu/p0mas
else
write(*,*) ' '
write(*,*) ' Unknown drag law.'
write(*,*) ' '
stop
endif
c write(*,*) ' cdp = ',cdp
c write(*,*) ' pa = ',pa(np)
92
c write(*,*) ' '
c Compute particle velocity

pup(np) = pu(np) + dt*pa(np)
c Compute particle position

pxp(np) = px(np) + dt*pup(np)
c Set default particle temperature

ptkp(np) = tk0
c Update particle heat transfer and temperature

c Compute the Nusselt number based on particle Reynolds number

if (rep .le. 2d2) then
nup = 2d0 + 0.106d0*rep*crppr
else
nup = 2.274d0 + 0.6d0*(rep**0.76d0)*crppr
endif
c Compute the heat transfer coefficient

hp = tcon*nup/dip
c Compute the heat transfer coupling term

pq(np) = hp*pi*dip*dip*(tk(pcel(np)) - ptk(np))
c Compute the particle temperature change

dtp = dt*pq(np)
ptkp(np) = ptk(np) + dtp
endif
c Check particle bounds

c if (pxp(np) .lt. x1) then
c pxp(np) = x1
c write(*,*) ' Particle ',np,' out of bounds.'
c stop
c endif
c if (pxp(np) .gt. x2) then
c pxp(np) = x2
c write(*,*) ' Particle ',np,' out of bounds.'
c stop
c endif
enddo
endif
c Update time and iteration number
n = n + 1
time = time + dt
write(*,*) nstart+n,' ',dt,' ',time,' ',item

write(*,*) 'pum = ',pum

93
c Solution and restart file output
if (mod(n,ndmp) .eq. 0) then
nfil = nfil + 1
c Solution file
90 format('sol_',i3.3,'.data')
write(filex,90) nfil
open(22,file=filex,form='formatted')
write(22,*) '# ',time
do i = 1,imax-1
xc = c12*(x(i) + x(i+1))
write(22,72) xc,rp(i),up(i),pp(i),zp(i),eip(i),ap(i),
& rxr(i),tk(i)
enddo
close(22)
c Particle file
91 format('par_',i3.3,'.data')
write(parex,91) nfil
open(22,file=parex,form='formatted')
do np = 1,npar
write(22,*) pxp(np),' ',pup(np),' ',ptkp(np)
enddo
close(22)
endif
c Derivatives file
open(22,file='deriv.data',form='formatted')
do i = 1,imax-1
write(22,*) i,' ',derv(i,1),' ',derv(i,2)
enddo
close(22)
c L/R Z files
c open(22,file='zlzr.data',form='formatted')
c do i = 1,imax
c write(22,*) i,' ',zzl(i),' ',zzr(i)
c enddo
c close(22)
c Restart file
open(40,file='restart.data',form='unformatted')
write(40) nstart+n
write(40) nfil
write(40) time
do i = 1,imax-1
write(40) rp(i),pp(i),up(i),zp(i)
enddo
close(40)
endif
c Reset arrays
94
do i = 1,imax-1
r(i) = rp(i)
u(i) = up(i)
z(i) = zp(i)
ei(i) = eip(i)
p(i) = pp(i)
a(i) = ap(i)
enddo
92 format(2x,d15.9,2x,d15.9,2x,d15.9,2x,d15.9,2x,i5)

do np = 1,npar
px(np) = pxp(np)
pu(np) = pup(np)
ptk(np) = ptkp(np)
if (idbgp .eq. 1) write(110+np,92) time,pxp(np),

& pup(np),pa(np),pcel(np)
enddo
endif
c pause
c End of solver loop
enddo
c Termination codes
if ( time .gt. tend) then
write(*,*) ' '
write(*,*) ' TIME > TEND.'
else if (n .ge. nstp) then
write(*,*) ' '
write(*,*) ' N > NSTP.'
else
write(*,*) ' UNKNOWN TERMINATION CRITERIA.'
endif
c End of main program

stop
end

95
DISTRIBUTION LIST
Defense Technical Information Center 1 Electronic Copy (1 file, 1 format)

Attn: Acquisition (OCA)
8725 John J. Kingman Road, Ste 0944
Ft Belvoir, VA 22060-6218
EGLIN AFB OFFICES:
AFRL/RWOC (STINFO Tech Library Copy) 1 Copy

AFRL/RW CA-N Notice of publication only
AFRL/RWG - 1 Copy
AFRL/RWM - 1 Copy
AFRL/RWA - 1 Copy

ADA553075

Uploaded by

Copyright:

Available Formats

ADA553075

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

ADA553075

Uploaded by

Copyright:

Available Formats

AFRL-RW-EG-TR-2011-159

BASIC DETONATION PHYSICS ALGORITHMS

DISTRIBUTION A. Approved for public release, distribution unlimited. 96th ABW/PA

AIR FORCE RESEARCH LABORATORY

AFRL-RW-EG-TR-2011- 159 HAS BEEN REVIEWED AND IS APPROVED FOR PUBLICATION

FOR THE DIRECTOR:

ORIGINAL SIGNED ORIGINAL SIGNED

9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR'S ACRONYM(S)

13. SUPPLEMENTARY NOTES

15. SUBJECT TERMS

1.0 Numerical Detonation Physics 1

2.1 The Reactive Euler Equations 4

3.1 Flux Jacobian Matrices 10

4 BUILDING THE NUMERICAL SCHEME 18

4.1 Pressure Derivatives 18

5.1 Coupling Terms 28

6.1 Simple Plane Wave Detonation 32

Distribution A. Approved for public release, distribution unlimited. (96ABW-2011-0548)

APP. A SOURCE CODE 42

Distribution A. Approved for public release, distribution unlimited. (96ABW-2011-0548)

2 Problem 1 Detonation Field Density, Time 3.0 33

3 Problem 1 Detonation Field Velocity, Time 3.0 33

4 Problem 1 Detonation Field Pressure, Time 3.0 34

5 Problem 1 Detonation Field Reaction Progress Variable, Time 3.0 34

6 Numerical Detonation Solution Hayes-I/JWL in HMX at 3 μs. Horizontal Axis

7 Numerical Detonation Solution Hayes-II/JWL in HMX at 3 μs. Horizontal Axis

8 Radial Locations for Steel Particles Embedded in a Mass of Detonating HMX 37

9 Radial Velocities for Steel Particles Embedded in a Detonating Mass of HMX 38

Distribution A. Approved for public release, distribution unlimited. (96ABW-2011-0548)

1.0 Numerical Detonation Physics

Numerical Detonation Physics applies many of the same computational

Distribution A. Approved for public release, distribution unlimited. (96ABW-2011-0548)

The detonation of a condensed explosive is most often modeled as a shock-driven

1.1 A Map for this Report

This report is intended to assist in the process of transitioning detonation physics

The report is organized as follows. In Section 2, we describe the governing

Distribution A. Approved for public release, distribution unlimited. (96ABW-2011-0548)

To address the detonation problem, we follow a body of research documented in

2.1 The Reactive Euler Equations

is the vector of conserved variables, and

is the flux vector. Also,

S Rx  [0, 0, 0,  r]T (2.1.5)

S P  [0, Fs , Q s , 0]T (2.1.6)

We may also write the total energy per unit volume as

where  is the reaction progress variable.

Vectors S G , S Rx and S P contain source terms; as we have shown, these

2.2 Mixture Equations of State

Distribution A. Approved for public release, distribution unlimited. (96ABW-2011-0548)

with  g and  s given by (2.2.2) and (2.2.3).

2.3 Solid Explosive Equations of State

Distribution A. Approved for public release, distribution unlimited. (96ABW-2011-0548)

In equations (2.3.1) through (2.3.4), P0 , T0 and  s 0 are the ambient pressure,

Table 1 - Hayes EOS Data for HMX

Distribution A. Approved for public release, distribution unlimited. (96ABW-2011-0548)

2.4 Detonation Products Equation of State

Table 2 - JWL Coefficients for HMX

Distribution A. Approved for public release, distribution unlimited. (96ABW-2011-0548)

3.1 Flux Jacobian Matrices

Capturing the structure of the detonation wave constitutes a difficult numerical

 F1 F1 F1 F1 