Submitted to ANS/NT (2021) LA-UR-20-30042 (v3)

On the Symmetry of Blast Waves

Roy S. Baty (XTD-PRI)*and Scott D. Ramsey (XTD-NTA) Applied Physics Theoretical Design Division Los Alamos National Laboratory

Abstract— This article presents a brief historical review of G. I. Taylor’s solution of the point blast wave problem which was applied to the Trinity test of the first atomic bomb. Lie group symmetry techniques are used to derive Taylor’s famous two-fifths law that relates the position of a blast wave to the time after the explosion and the total energy released. The theory of exterior differential systems is combined with the method of characteristics to demonstrate that the solution of the blast wave problem is directly related to the basic relationships that exist between the geometry (or symmetry) and the physics of wave propagation through the equations of motion. The point blast wave model is cast in terms of two exterior differential systems and both systems are shown to be integrable with local solutions for the velocity, pressure, and density along curves in space and time behind the blast wave. Endorsement: Len G. Margolin (XCP-5)

I. HISTORICAL INTRODUCTION the Trinity test is 24.8 (±1.98) kilotons, which is pre- In 1950, the Proceedings of the Royal Society of Lon- sented in this issue by Selby, et al. [5]. Given the physical don published two papers by Sir Geoffrey Taylor [1] and complexity of the atomic explosion, the fact that Taylor’s [2]: the main purpose of these papers was to develop, blast wave solution accurately bounded and predicted the solve, and apply the results of a so-called point blast wave Trinity yield is remarkable. problem to the determination of the energy released in an The historical details surrounding the genesis of the intense explosion. These papers are of enduring historical intense explosion model are fascinating in their own right, significance for many reasons; and not the least of which though the full story is too lengthy to be reproduced in being, in the second paper, Taylor’s provision of an esti- this short note. A full account of the effort is provided mate of the yield of the explosion of the Trinity atomic by Clark [6]; see also the books by Barenblatt [3] and test conducted July 16, 1945. As noted by Barenblatt [3], [7] for shorter summaries and attendant technical insight into the implications of the work. A discussion of Tay- lor’s technical contributions to the blast wave problem “... Taylor’s prediction of the value ... caused, in his in contrast with parallel analyses is discussed by Deakin words, ‘much embarrassment’ in American government [8]. However, by all accounts, a central question surround- circles ...” ing the investigation involved the potential mechanical effects associated with an intense explosion, namely the as the yield of the Trinity explosion was still, at the time, generation of a powerful blast (or shock) wave. Given the a closely guarded secret; even though the length and time- well-established difficulties associated with solving any stamped high-speed photographic stills of the Trinity fire- set of fluid mechanical equations involving shock waves, ball (owing to Julian Mack, among others) used as an again according to Barenblatt [3], essential component of Taylor’s analysis had otherwise been declassified and released to the public, Fig. 1. “... in the whole of Britain there was only one man able arXiv:2103.05710v1 [physics.hist-ph] 9 Mar 2021 Taylor compared his solution of the evolution of the to solve this problem – Professor G. I. Taylor.” blast wave to the published photographs to estimate the initial energy release as a function of the specific heat Indeed, given his expertise in fluid mechanics honed ratio for an ideal gas, γ. The blast wave analysis produced over the course of nearly three decades, Taylor recognized the following yield estimates: 34 kilotons for γ = 6/5, that a mathematical model of the minimum required fi- 22.9 kilotons for γ = 13/10, 16.8 kilotons for γ = 7/5, delity for characterizing shock wave propagation in air is and 9.5 kilotons for γ = 5/3. The historical yield of the represented by the spherically symmetric compressible Trinity test published by the U.S. Department of Energy Euler equations in an ideal gas, namely is 21 kilotons, [4]. The modern estimate of the yield of ρu * ρ + uρ + ρu + 2 = 0, (1) corresponding author: [email protected] t x x x Submitted to ANS/NT (2021) LA-UR-20-30042 (v3)

Figure 1: Photograph of the Trinity atomic test showing a blast wave.

ρut + ρuux + px = 0, (2) to the shock wave location, and Rt is the shock wave  u speed. p + up + γp u + 2 = 0, (3) t x x x In the 1940s, even Eqs. (1) through (6) – themselves a drastic simplification of vastly more complicated entities where x and t denote the spherical radial position co- such as the multi-dimensional compressible Navier-Stokes ordinate and time, respectively, ρ, u, and p denote the equations – proved impossible to solve in general (either fluid density, radial flow velocity, and isentropic pressure, analytically or numerically) without resorting to a variety respectively, subscripts denote partial derivatives with re- of additional simplifying assumptions. At this point, as spect to the indicated variables, and γ is the specific heat noted by Barenblatt [3], ratio of an ideal gas, further assumed to be a given con- stant for air. By the 1940s it was already well established “... G. I. Taylor, however, was astute. His ability to deal that in the absence of dissipative mechanisms Eqs. (1) to with seemingly unsolvable problems, by apparently minor (3) admit weak or piecewise continuous solutions, with adjustment converting them to problems admitting simple discontinuities that are interpreted as shock waves of in- and effective mathematics, was remarkable ...” finitesimal width. In shock wave solutions of Eqs. (1) to (3), the differen- Indeed, Taylor’s astonishing physical insight allowed tial equations themselves are taken to hold on either side him to formulate the blast wave problem as depicted in of the discontinuity, but not at the location of the discon- Fig. 2. In this scenario, for t ≥ 0 the air exterior to the tinuity itself, x = R(t). To ensure conservation of mass, blast wave is assumed to be quiescent and of constant momentum, and energy across this discontinuity, the ideal density, namely gas Rankine-Hugoniot jump conditions must be enforced

at the boundary of the blast wave (x = R(t)), given by u0 = u(x > R) = 0, (7)   ρ0(u0 − Rt) = ρ1 u1 − Rt , (4) ρ0 = ρ(x > R), (8)

where ρ0 is a fixed positive constant; these conditions 2 2 p0 + ρ0(u0 − Rt) = p1 + ρ1(u1 − Rt) , (5) present little conceptual difficulty. However, perhaps less intuitively obvious are Taylor’s two additional key assump-

2 tions, including for t ≥ 0, γp0 1  + u0 − Rt (γ − 1)ρ0 2 p0 = p(x > R) = 0, (9) 2 γp1 1  = + u1 − Rt , (6) indicating that the ambient counter-pressure p of the (γ − 1)ρ1 2 0 air into which the blast wave propagates is negligible in where the subscripts 0 and 1 denote, respectively, the un- comparison to the pressure in the region bounded by the perturbed and perturbed fluid state immediately adjacent blast wave. Submitted to ANS/NT (2021) LA-UR-20-30042 (v3)

Figure 2: Diagram defining Taylor’s point blast wave problem.

Finally, Taylor assumed the initial length scale R0 of cally referred to as the Sedov-Taylor-von Neumann point the rapidly exploding object is negligible in comparison to blast wave (or point explosion) problem; for additional the spherical blast wave radius R evaluated at some later details see Korobeinikov, [11], or Kamm, [12]. time, so that the finite initial blast energy E0 (or yield) is assumed to be instantaneously released at an infinitesimal II. SYMMETRY ANALYSIS point in space and time (x = 0 and t = 0). Among other With implicit relevance to Taylor’s analysis of the blast consequences this assumption results in the formulation wave, Garrett Birkhoff published his classic work Hydro- of the so-called energy integral given by dynamics: A Study in Logic, Fact, and Similitude, [13], also in 1950. With much reference to an earlier seminal Z R  p 1 2 2 E0 = 4π + ρu x dx, (10) text on dimensional analysis by Percy Bridgman [14], one 0 γ − 1 2 of the stated objectives of Birkhoff’s work is to provide indicating the total energy interior to R is the initial blast “... a critical account ... of dimensional analysis. This energy E0, and is conserved for all times. This final assumption also lends itself to the more precise name of is usually invoked in justifying model experiments; it has the model depicted in Fig. 2, Taylor’s point blast wave the advantage of requiring no mathematical background problem. beyond high-school algebra, but has the disadvantage of With Eqs. (1) through (9) and the point blast wave needing additional postulates, whose physical validity assumption, Taylor famously used dimensional analysis must be tested independently. [Birkhoff] give[s] these techniques to reduce and ultimately produce a numerical postulates a group-theoretic formulation, in terms of the solution of Eqs. (1) to (6) for not only the blast wave tra- dimensional group of all changes of fundamental units.” jectory, R(t), but also the space time distributions for the velocity, u, pressure, p, and density, ρ, for the flow field The ultimate success of Birkhoff’s systemization of di- at positions behind the blast wave x < R(t). Consistent mensional analysis and model reduction techniques within with the techniques of dimensional analysis, and among the formalized setting of group theory in turn inspired the many other remarkable characteristics, Taylor’s solution widespread revitalization of 19th century mathematician for the flow field behind the blast wave has the properties Sophus Lie’s group-theoretic techniques as a generalized of being both scale-invariant and self-similar. Figure 3 solution strategy for differential equations (see, for exam- shows the similarity solutions for velocity, pressure, and ple, Ovsyannikov [15], or Olver [16]). The seemingly density computed by Taylor, [1]. curious confluence of all these events initially within the It should also be noted that both John von Neumann, [9] context of hydrodynamics is, according to Andreev et al., and Leonid Sedov, [10] independently produced closed- [17], not to be wondered at: form solutions to the point blast wave problem, as a direct consequence of their explicit identification and use of Eq. “... this analysis was especially fruitful in application to (10). As such, the scenario depicted in Fig. 2 is now typi- the basic equations of mechanics and physics because the Submitted to ANS/NT (2021) LA-UR-20-30042 (v3)

Figure 3: Taylor’s solution of the point blast wave in terms of the similarity variable, η. In the large plot, the dotted curve is the velocity, φ, and the solid curve is the pressure, f, while the inserted plot is the density, ψ.

invariance principles are already involved in their deriva- and tion. It is in no way a coincidence that the equations u˜x =u ˜x(x, u, ux), of hydrodynamics served as the first object for applying leave Eq. (11) invariant. Lie further assumed that deriva- the new ideas and methods of group analysis which were tives of u satisfy the following equation in terms of the developed by L.V. Ovsyannikov and his school ...” total derivative: X du = u dx . (12) Accordingly, the group-theoretic (or symmetry) interpre- xi i tation of dimensional analysis and similarity reduction Equation (12) has the modern interpretation as a co-vector theories has profound consequences for self-similar scal- or differential one-form ing hydrodynamic phenomena, including the solution of X the Taylor point blast wave. ω = du − uxi dxi, (13) To this point, a symmetry of a given equation is defined which may be used to analyze the local geometric proper- to be a coordinate transformation of the equation such that ties of Eq. (11). From a classical perspective, the expres- the equation remains invariant. While the basic goal of sion du of Eq. (12) represents the infinitesimal change of symmetry analysis is to determine the coordinate transfor- u under an infinitesimal change of the point on the under- mations which leave an equation invariant, the practical lying geometric object where the expression is evaluated. utility of symmetry analysis is to understand the transfor- Since geometric infinitesimals are not rigorously defined mations that may be used to solve or simplify the equation. and because total derivatives may be defined along tan- Let gent vectors, du is interpreted as a differential one-form, F (x, u, u(1), ..., u(k)) = 0, (11) Spivak [18]. Differential one-forms are the objects dual to represent a partial differential equation (PDE) of order vector fields defined locally on curves and surfaces. The 1 m use of differential forms is very powerful in the analysis k, where x = (x1, ..., xn) and u = (u , ..., u ) are the of systems of PDEs. By admitting higher order exterior independent and dependent variables respectively, and u(l) represents the totality of derivatives of order l. forms, smooth systems of PDEs are equivalent to exterior Lie studied transformations of the form differential systems (EDSs) which may be analyzed with mathematical techniques from differential and algebraic

(x, u, ux) −→ (˜x, u,˜ u˜x), geometry, where an EDS is a set of equations in terms of differential forms, Krasil’shchik and Vinogradov [19], as symmetries, where and Vinogradov [20]. The one-dimensional Euler equations, Eqs. (1), (2), and x˜ =x ˜(x, u, ux), (3) may be reduced to an EDS by introducing a transfor- mation which explicitly reduces the PDEs to ordinary dif- u˜ =u ˜(x, u, ux), ferential equations (ODEs), or by introducing a universal Submitted to ANS/NT (2021) LA-UR-20-30042 (v3)

space of higher dimension which replaces the derivatives Σ ⊂ J 1(R2, R3) via the jet variables with the coordinate in the PDEs with independent variables and then couples cover (x, t, ρ, u, p). the new variables to the original derivatives through differ- Next, consider the space of all differential forms defined ential forms (total derivatives). Both of these approaches on Σ, Λ∗(Σ) ≡ ⊕ Λk(Σ), where Λk(Σ) ≡ Γ(ΛkT ∗Σ) are discussed in this article; the transformation approach is the space of sections of smooth functions defined on to outline solutions of the blast wave problem, and the the bundle of k-order differential forms. A subspace general approach to show how symmetry is coupled to the I ⊂ Λ∗(Σ) is an algebraic ideal if it is a direct sum physics of blast wave propagation. The notation, concepts, of homogeneous subspaces Ik ⊂ Λk(Σ) and it is closed and theorems applied in this article on exterior differential under the wedge product. An algebraic ideal I is a differ- systems follow, Ivey and Landsberg [21]. ential ideal if dI ⊂ I. An exterior differential system on For Eqs. (1), (2), and (3), the following independent a surface Σ is a differential ideal I ⊂ Λ∗(Σ). variables are introduced: The EDS representing the Euler equations is then given by the differential ideal x1 ≡ t, x2 ≡ x, u1 ≡ ρ, u2 ≡ u, u3 ≡ p, n I = α ∧ σ1 + β ∧ σ2 + λ ∧ σ3 + γ ∧ dσ1 and o + δ ∧ dσ2 + φ ∧ dσ3| α, β, λ, γ, δ, φ ∈ Λ∗(Σ) , 1 1 2 u1 ≡ ρt, u2 ≡ ρx, u1 ≡ ut, 2 3 3 defined by the algebraic equations (14), (15), and (16), u2 ≡ ux, u1 ≡ pt, u2 ≡ px. and the differential forms given by Eqs. (17), (18), and The Euler equations are then cast on the 11 dimensional (19). Moreover, for the general system of PDEs the space space R2 × R3 × R6 in terms of (x1, x2) × (u1, u2, u3) × and time variables are assumed to be linearly independent 1 1 2 2 3 3 1 2 (u1, u2, u1, u2, u1, u2) which is called the space of 1-jets, dx ∧ dx 6= 0. and denoted by J 1(R2, R3). Eqs. (1), (2), and (3) then Infinitesimal symmetries have a natural expression for become an EDS. Let I be an EDS on Σ. A vector field V is an infinitesimal symmetry of I if LV Ψ ∈ I for all Ψ ∈ I. u1u2 1 1 2 1 1 2 Here LV is the Lie derivative with respect to the vector F ≡ u1 + u u2 + u u2 + 2 , (14) x2 V . F 2 ≡ u1u2 + u1u2u2 + u3, 1 2 2 (15) III. SYMMETRY ANALYSIS OF THE POINT and BLAST WAVE PROBLEM 2 3 3 2 3 3 2 u As discussed in Section 1, Taylor’s construction of his F ≡ u1 + u u2 + γu (u2 + 2 ); (16) x2 point blast wave solution proceeds with the use of dimen- subject to the differential forms: sional analysis techniques. Alternatively, this solution may also be systematically derived through invariance of 1 1 1 1 1 2 σ = du − u1dx − u2dx and Eqs. (1) through (10) under a Lie group of scaling transfor- dσ1 = −du1 ∧ dx1 − du1 ∧ dx2, (17) mations. Both approaches are summarized in the sections 1 2 to follow. However, ultimately, using group-theoretic tech- niques Taylor’s point blast wave solution may be explicitly σ2 = du2 − u2dx1 − u2dx2 and categorized according to a symmetry group inherent to 1 2 Eqs. (1) through (10). 2 2 1 2 2 dσ = −du1 ∧ dx − du2 ∧ dx , (18) III.A. Dimensional Analysis

3 3 3 1 3 2 The most important result arising from the codification σ = du − u1dx − u2dx and of dimensional analysis techniques in the late 19th and 3 3 1 3 2 dσ = −du1 ∧ dx − du2 ∧ dx , (19) early 20th centuries is the celebrated “Buckingham-π The- orem” (so called due to its explicit statement by Edgar which are used to define the exterior differential system Buckingham in 1914) which, according to Barenblatt [7], representing the Euler equations, where ∧ is the exterior states product for differential forms. The exterior (or wedge) product is a multiplicative operation defined on alternating Theorem A physical relationship between some dimen- multilinear mappings of vector spaces (tensors). The map- sional (generally speaking) quantity and several dimen- ping (ρ, u, p)t : R2 −→ R3 identifies a smooth surface sional governing parameters can be rewritten as a re- Submitted to ANS/NT (2021) LA-UR-20-30042 (v3)

lationship between some dimensionless parameter and A solution of Eqs. (1) to (6) subject to Eqs. (7) through several dimensionless products of the governing parame- (9) is required to determine S. Proceeding as with Eq. (20), ters; the number of dimensionless products is equal to the Taylor assumes the radial flow velocity, pressure, and total number of governing parameters minus the number density interior to the spherical blast wave to depend on of governing parameters with independent dimensions. the parameters: u In his first paper [1], Taylor uses this powerful result = Fu(x, R, u1, γ), (25) to construct the solution of the point blast wave problem. u1 Referring to Fig. 2, in this analysis Taylor formulates the p blast wave position R as depending on the initial energy = Fp(x, R, p1, γ), (26) p1 E , the time t elapsed since the explosion began, and the 0 ρ ambient mass density ρ and ideal gas specific heat ratio = F (x, R, ρ , γ), (27) 0 ρ ρ 1 γ of the surrounding air; that is 1 where the necessary dependences on E0, t, and ρ0 are R = φ0(E0, t, ρ0, γ), (20) implicit in those of R, and ρ1, u1, and p1 follow from an algebraic solution of Eqs. (4), (5), and (6) subject to where φ0 is a function to be determined of the indicated Eqs. (7) to (9). Equations (7) through (9) suggest that arguments. The dimensions of the parameters appearing in addition to E , t, and ρ , the flow profiles inside the 2 −2 0 0 in Eq. (20) are given by: R ∼ L, E0 ∼ ML T , t ∼ T, region bounded by the outgoing blast wave, defined by −3 ρ0 ∼ ML , and γ is dimensionless, where the set of R, also depend on the Rankine-Hugoniot jump conditions fundamental mechanical dimensions of L, T, and M refer immediately behind the shock wave, as indicated by the to units of length, time, and mass, respectively. It is readily subscript 1. verified that the dimensions of E0, t, and ρ0 are linearly Following an identical procedure that resulted in independent; that is, the dimensions of any one of these Eq. (23), Eqs. (25) through (27) become quantities cannot be formed as a linear combination of any

of the others. u = u1φ(η, γ), (28) Since R has dimensions of L, the function φ0 must also have dimensions of L; accordingly, Eq. (20) may be p = p1f(η, γ), (29) written as a1 a2 a3 ρ = ρ0ψ(η, γ), (30) R = S(γ)E0 t ρ0 , (21) where where a1, a2, and a3 are constants to be determined, and x η ≡ . (31) where S is an arbitrary function of the dimensionless R quantity, γ. The dimensions of Eq. (21) are given by The independent variable η is a similarity variable, which L = (ML2T−2)a1 (T)a2 (ML−3)a3 . (22) may be used with the functions φ, f, and ψ to reformu- late the PDEs (1) to (3) as three ODEs, while Eqs. (4) Equation (22) produces three linear algebraic equations through (6) combined with Eqs. (7), (8), and (9) may like- for powers of L, T, and M in terms of the exponents a1, wise be reformulated as three associated initial condi- a2, and a3, which implies that a1 = 1/5, a2 = 2/5, and tions. The reduction to a system of ODEs is only possi- a3 = −1/5. With these values for the exponents, Eq. (22) ble since the x and t dependences featured in the defini- becomes tions of φ, f, and ψ appear only in terms of their com- 1 E0  5 2 bination through η. Therefore, a solution of the reduced R = S(γ) t 5 , (23) ρ0 ODEs for φ, f, and ψ is necessarily self-similar: since or, in terms of the equivalent dimensionless π-group, R = R(t), u1 = u1(t), p1 = p(t), and ρ1 = ρ1(t) the spatial dependence in all flow variables is structurally in- 1 2 variant and may be obtained between different times by E0  5 t 5 S(γ) = π0 = constant. (24) the scaling transformations indicated in Eqs. (28) through ρ0 R (31). Equation (23) is the famous result derived by Taylor, in- Finally, for a fixed value of γ, a solution for φ, f, and cluding the proportionality between the spherical blast ψ transformed back to that for u(x, t), p(x, t), and ρ(x, t) wave position R and the two-fifths power of the time t using Eqs. (28) to (31) may be substituted into the energy elapsed since the explosion began. At this point the func- integral given by Eq. (10) together with Eq. (23) to yield tion S remains undetermined. the previously undetermined function S(γ) as appearing Submitted to ANS/NT (2021) LA-UR-20-30042 (v3)

in Eq. (23), and so the point blast wave solution is deter- As such, and in the context of scaling symmetries, all mined in its entirety. That this solution is achievable under results of Taylor’s point blast wave problem as summa- construction of only Eqs. (23) and (28) through (31) is at rized in Section 3.1 may be derived using Lie’s systematic the root of various statements appearing in the literature framework, beginning with the fundamental governing to the effect of: equations given by Eqs. (1) through (6). Proceeding, in the context of Eqs. (1) through (6), the maximal Lie group “... solutions of the first type possess the property that of all potential scaling transformations in all dimensional the ... exponents of t and R in all scales are determined variables may be written as either by dimensional considerations or from the conser- vation laws ...” x˜ = eα1x, t˜= eα2t, ρ˜ = eα3ρ, u˜ = eα4u,

α5 as noted by Zel’dovich and Raizer [22]. Indeed, the point p˜ = e p, (32) blast wave solution is remarkable in the elegance of the where the tildes represent transformed variables,  is purely dimensional arguments that give rise to it. As dis- known as the group parameter, and α , α , α , α , and cussed extensively by Barenblatt [3] and [7], the principal 1 2 3 4 α are constants to be determined. The transformations difficulty of the analysis – and, more remarkable still, Tay- 5 given by Eqs. (32) are referred to a Lie group, as they lor’s genius in overcoming it – is in the selection of the feature an identity element  = 0, the inverse element, and minimum parameter set still of sufficient physical fidelity the closure and associativity properties under composition. for establishing the dependencies of R, ρ, u, and p, as rep- The goal of symmetry analysis is to identify the pre- resented essentially by E , t, ρ , and γ. In turn, the selec- 0 0 cise instantiation of Eq. (32) that leaves a given set of tion of this parameter set demanded of Taylor a significant mathematical relationships invariant; that is, the relevant amount of judgement so as to credibly ignore possible ef- equations assume identical forms in both the original and fects such as variable γ or ρ , higher-dimensional effects, 0 transformed variable sets. Lie’s fundamental achievement and the effects of finite p or R . 0 0 in the execution of this technique for differential equations For this very reason dimensional analysis techniques was the realization of this equivalence on a purely local (or have been subject to much misunderstanding and criticism infinitesimal) level, using a generalized directional deriva- from almost the beginning of their formalization in the late tive operator L known as a Lie derivative. In the case of 19th century. Certain pointed questions surrounding the V Eq. (32), the relevant Lie derivative is defined using the systemization of any such efforts remained largely unan- vector field swered even through the middle of the 20th century, when Lie’s symmetry analysis techniques were essentially re- ∂ ∂ ∂ V = α x + α t + α ρ discovered, and introduced into much of modern physics. 1 ∂x 2 ∂t 3 ∂ρ ∂ ∂ III.B. Lie Group Analysis + α u + α p , (33) 4 ∂u 5 ∂p Classical treatments of dimensional analysis theory (see, for example, Bridgman [14]) include suggestive terminol- that may be constructed using a first-order Taylor expan- ogy such as “change ratios,” “complete equations,” and sion of Eqs. (32) about the identity element. Geometri- “dimensional homogeneity.” Birkhoff [13] first character- cally, V is a tangent vector to a surface invariant under the ized these and related concepts according to their rigorous, transformations given by Eqs. (32), existing in the space common theme: the notion of invariance or symmetry where (x, t, ρ, u, p) are regarded as independent coordi- of a mathematical relationship under a group of scaling nates. Indeed, invariance of a mathematical relationship transformations. under Eq. (33) is then assessed by evaluating In turn, the uses of symmetries of mathematical relation-

ships are based on the geometric interpretation of those LV Ψ ∈ I, (34) same relationships. Put simply, Lie’s analysis technique is a systematic means for identifying symmetric coordinate where each Ψ ∈ I are, for example, the elements of systems under which mathematical relationships assume Eqs. (1) through (10) cast as a differential ideal. If, during simpler forms than otherwise originally cast. In this sense, the course of the analysis of all relevant differential forms, the reduction in complexity ultimately afforded via in- at least one of α1, α2, α3, α4, or α5 is revealed to be non- variance is best interpreted as being connatural with the zero, the entire mathematical model under investigation symmetry set present in a scenario of interest, as opposed is said to possess a scaling symmetry. Moreover, the to being a byproduct of (in the case of scaling) nondimen- infinitesimal form of this symmetry given by Eq. (33) may sionalization principles. then be used to construct a coordinate system in which the Submitted to ANS/NT (2021) LA-UR-20-30042 (v3)

symmetric equations Ψ ∈ I assume simpler forms; in the so that with Eq. (38) and α3 = 0 resulting from Eq. (37), case of scaling symmetry, the accompanying reduction is Eq. (35) finally becomes identical to that resulting from maximal application of the Buckingham-π Theorem. 2 ∂ ∂ 3 ∂ 4 ∂ V = α2x + α2t − α2u − α2p , (39) Proceeding, evaluation of Eq. (34) with each Ψ ∈ I 5 ∂x ∂t 5 ∂u 5 ∂p given by each of Eqs. (1) through (3) yields α = α − α 4 1 2 a version of which is also given by Cantwell [26]. More- and α = α +2α = α −2(α −α ) (see, for example, 5 3 4 3 1 2 over, with Eq. (38), Eq. (36) becomes Eq. (23), which is Ramsey and Baty [23]), so that Eq. (33) reduces to now demonstrated to arise from the presence of a one- parameter (i.e., α as written) scaling symmetry of Eqs. (1) ∂ ∂ ∂ 2 V = α x + α t + α ρ through (10), in turn given by Eq. (39). 1 ∂x 2 ∂t 3 ∂ρ This outcome may be compared with the cases where ∂ ∂ either u 6= 0 or p > 0, or when a finite initial length + (α1 − α2)u + (α3 + 2α1 − 2α2)p , (35) 0 0 ∂u ∂p scale R0 appears in the problem formulation. In any of these cases, satisfaction of the appropriate manifestation indicating that in the absence of initial, boundary, or other- of Eq. (37) (for example) further demands α = 0, thus wise ancillary condition information, Eqs. (1) through (3) 2 entirely nullifying Eq. (39) and rendering the problem for- are dimensionally consistent with the mechanical scalings mulation without scaling symmetry. This result reflects u ∼ x/t and p ∼ ρu2. The presence of three independent the importance of Taylor’s intuition resulting in Eqs. (7), free parameters in Eq. (35) also reflects the presence of (8), (9), and (10). More broadly, it also demonstrates the three independent fundamental dimensions (e.g., L, T, and intuitive phenomenon that one degree of freedom is re- M as appearing in Section 3.1) in the formulation of the moved from the three-parameter mechanical scaling group fluid mechanical system given by Eqs. (1), (2), and (3). of Eqs. (1) to (6) for every fixed dimensional constant (e.g., Furthermore, evaluation of Eq. (34) with each Ψ ∈ ρ and E ) also introduced into a specific problem formu- I given by each of Eqs. (4) through (6), and the initial 0 0 lation. Consequently, for a self-similar scaling solution of condition R| = 0, yields (see, for example, Giron, et t=0 Eqs. (1) to (6) to exist, at least one such degree of freedom al. [24]) must be present in Eq. (39). α1 Under this construction, Eq. (39) may be used to iden- α tR + (α − α )R = 0 ⇒ R = kt α2 , (36) 2 tt 2 1 t tify the symmetric coordinates in which Eqs. (1) through for k = constant, further indicating that scale invariant (10) assume a simpler form and may ultimately be solved. shock wave propagation proceeds according to power-law In particular, the similarity variables introduced in Sec- behavior in time. The precise power appearing in Eq. (36) tion 3.1 via dimensional considerations are now readily shown to actually be the invariant coordinates of the one- remains arbitrary as neither α1 nor α2 is constrained given Eqs. (4), (5), and (6), which are dimensionally equivalent parameter scaling group generated by Eq. (39), determined to Eqs. (1), (2), and (3). according to Indeed, at this point, none of Taylor’s fundamental as- sumptions (i.e., Eqs. (7) to (10)) have been invoked. To rig- V H(x, t, ρ, u, p) = 0 ⇒ orously investigate the consequences of assuming Eqs. (7) dx dt dρ du dp = = = − = − , (40) to (9), evaluating Eq. (34) with each Ψ ∈ I given by each 2 3 4 5 x t 0 5 u 5 p of these conditions yields where H is an otherwise arbitrary invariant function of the

(α1 − α2)u0 = 0, α3ρ0 = 0, indicated arguments. The solution of the characteristic equations appearing in Eq. (40) features four constants (α + 2α − 2α )p = 0, (37) 3 1 2 0 of integration; these constants are identical to the func- tions φ, f, ψ, and η defined by Eqs. (28) through (31). satisfaction of which at least requires α3 = 0, since ρ0 > Following the systematic derivation of these similarity 0 by the physical problem formulation. Otherwise, α1 and variables, the symmetry reduction associated with Eq. (39) α2 are not further constrained since u0 = 0 and p0 = 0 via Eqs. (7) and (9), respectively. is revealed to collapse to ODEs otherwise encountered Continuing, evaluating Eq. (34) with the final Ψ ∈ I in Section 3.1; the complete solution of the self-similar given by Eq. (10) yields (see, for example, Hutchens [25]) Taylor point blast wave problem then proceeds in the same manner as previously disseminated. Through this analysis, 1 E0  5 2 the entire Taylor blast wave solution is thus revealed to be k = S and α1 = α2, (38) a direct manifestation of a one-parameter group of scal- ρ0 5 Submitted to ANS/NT (2021) LA-UR-20-30042 (v3)

ing transformations admitted by the problem formulation, ρ, p, and u. Since the equations are cast in terms of the namely Eqs. (1) through (10). total derivatives of the flow variables, these equations are naturally interpreted as differential one-forms defined on IV. METHOD OF CHARACTERISTICS submanifolds (the Mach lines) associated with the exterior An important theme of using symmetry to analyze a differential system induced by the one-dimensional Euler system of PDEs is that of simplifying the system of equa- equations. tions. For the Taylor point blast wave problem, symmetry The abstract framework for exterior differential systems is applied to reduce the system of PDEs to a system of gives insight into the relationship between characteristics ODEs, which then allows direct numerical integration of and symmetries. To see this relationship, let I ⊂ Λ∗(Σ) the problem. In compressible fluid mechanics, the symme- be a differential ideal representing the Euler Eqs. (1), (2), tries that reduce the PDEs to a system of ODEs are related and (3), where Σ is a smooth surface defined as a subset

to the fundamental physics of wave propagation. For un- of the jet space J 1(R2, R3). A vector field V ∈ Γ(T Σ)

steady, compressible flow problems there are well defined (that is, V is section of the tangent bundle T¬Σ of Σ) is a curves or surfaces called characteristics along which phys- Cauchy characteristic¬ vector field for I if V Ψ ∈ I for ical disturbances such as shock waves propagate. For two all Ψ ∈ I. Here V Ψ is the interior product of the vector or three spatial variables and time, waves propagate along field V with the differential form Ψ. The key result is that characteristic surfaces, while for one spatial variable and if V is a Cauchy characteristic vector field of I, it is also time, waves propagate along characteristic curves. In the an infinitesimal symmetry of I. Hence, the vector fields mathematics literature, characteristic curves and surfaces associated with the Mach lines are Lie symmetries of the are called Cauchy characteristics. Euler equations associated with the propagation of waves. For two-dimensional supersonic flows containing shock waves, the characteristics are also called Mach lines. In IV.A. Characteristics and Compatibility Equa- compressible flows with waves, the equations defined tions throughout the flow simplify to equations specified on The characteristic equations for one-dimensional, un- the characteristic curves (or surfaces) which reduces the steady, compressible flow are given by number of spatial variables defining the problem. Restrict- dx 1 ing the governing PDEs to characteristics produces a new = = u (Pathlines), (41) dt λ set of equations called compatibility equations. Equations and (1), (2), and (3) describing the one-dimensional, unsteady, dx 1 compressible flow of an isentropic gas, are a system of = = u ± a (Mach lines), (42) dt λ first-order PDEs in one space variable and time. There- ± fore, the blast wave problem may be reduced to a system where a is the local speed of sound. The associated com- of equations on the Mach lines governing the propagation patibility equations are of a shock wave. dp − a2dρ = 0 along Pathlines, (43) To find the Mach lines and restrict the equations of motion to these lines, a linear combination of Eqs. (1), (2), and and (3) is formed and recast in terms of total derivatives ρua2 dp ± ρadu + α dt = 0 along Mach lines, (44) of the density, pressure and velocity: x Zucrow and Hoffman, [28]. In Eq. (44), α represents the dρ = ρtdt + ρxdx, coordinate system: α = 0 cartesian coordinates, α = 1

dp = ptdt + pxdx, cylindrical coordinates, and α = 2 spherical coordinates. and IV.B. Spherically Symmetric Potential Flow du = utdt + uxdx. For a one-dimensional, spherically symmetric flow field, The characteristic curves, denoted by dt/dx = λ, are the the velocity field will be irrotational at each point in the coefficients of the time derivatives for the differentials of fluid. Hence, a potential function, ϕ, exists such that density, pressure, and velocity. A general method to derive u = ϕx. For an unsteady, isentropic, compressible flow, the characteristic curves (or surfaces) as well as the restric- the conservation equations, Eqs. (1) to (3), may be used tion of the equations of motion to the characteristics is to derive a second-order, hyperbolic, partial differential developed in Rusanov [27]. The restriction of the PDEs of equation in terms of the potential function: motion to Mach lines for the blast wave problem reduces a2u (a2 − u2)ϕ − 2u ϕ − ϕ + 2 = 0, (45) the problem to a set of ODEs in terms of the flow variables, xx xt tt x Submitted to ANS/NT (2021) LA-UR-20-30042 (v3)

subject to the compressible Bernoulli equation, The ODEs derived for the blast wave may then be cast as differential one-forms by defining: Z p 1 2 dp ϕt + u + = 0. (46) 0 2 ρ 1 1  1 f 3  p0 θ := dφ − − φ dη, (53) η − φ γ ψ 2 Then restricting the potential equation (45) to the Mach lines (the characteristic curves) given by Eqs. (42), the f(−3η + φ(3 + γ/2) − 2γφ2/η) θ2 := df − dη, (54) compatibility equations for spherically symmetric, isen- ((η − φ)2 − f/ψ) tropic, compressible potential flow become: and φ0 + 2φ/η a2u θ3 := dψ − ψ dη. (55) (a2 − u2)du − (u ± a)dϕ + 2 (u ± a)dt = 0, (47) η − ψ t x Equations (53), (54), and (55) are applied to identify inte- Sauer, [29]. gral curves on the manifold Σ modeling solutions of the The compatibility equations (47) for spherically sym- Euler equations (1), (2), and (3) with the coordinate cover metric potential flow are equivalent to the compatibility (x, t, ρ, u, p). The following theorem (see Ivey and Lands- equations associated with the Euler equations, Eqs. (44). berg [21]) gives conditions for the existence of solution To see the equivalence of the two sets of compatibility curves to the ODEs above restricted to a submanifold of equations, compute the exterior derivative of the com- Σ: pressible Bernoulli equation to produce Theorem Let Σ be a C∞ manifold of dimension m, and dp 1 m−1 1 dϕt = −u du − . (48) let θ , ..., θ ∈ Λ (Σ) be pointwise linearly inde- ρ pendent in some neighborhood U ⊂ Σ. Then through z ∈ U there exists a curve c : −→ U, unique up to Combining Eq. (48) with Eqs. (47) then yields R reparametrization, such that c∗(θa) = 0 for 1 ≤ a ≤ ρua2 m − 1. dp ± ρadu + 2 dt = 0, (49) x The theorem guarantees that the system of ODEs modeling along the Mach lines of Eqs. (44) as claimed. the self-similar motion of the blast wave will have a local solution represented by a curve c = (φ(η), f(η), ψ(η)) V. THE TAYLOR POINT BLAST WAVE AS near a specified point z ∈ (η, φ, f, ψ). The differential EXTERIOR DIFFERENTIAL SYSTEMS forms are zero along solution curves c of the ODEs used to V.A. Similarity Solution of the Blast Wave define the forms. Notice that the similarity transformation Combining the functional forms of Eqs. (28) to (31) restricts the underlying surface (x, t, ρ, u, p) modeling the with expressions for R and dR/dt as well as the equations blast wave to a submanifold described by (η, φ, f, ψ) with of conservation of mass, momentum and energy, it may m = 4 in terms of m − 1 = 3 linearly independent one- shown that the following system of ODEs results modeling forms, Eqs. (53), (54), and (55). the blast wave: V.B. Exterior Forms Defined on the Blast Wave dφ 1  1 f 0 3  Characteristics = − φ , (50) dη η − φ γ ψ 2 In Section 4, the Method of Characteristics was used to reduce the spherically symmetric Euler equations defined df f(−3η + φ(3 + γ/2) − 2γφ2/η) in space and time (x, t) to a system of ODEs (which may = , (51) dη ((η − φ)2 − f/ψ) also be interpreted as an EDS) defined on the pathlines and Mach lines in the flow field. Restriction of the PDEs and to the Mach lines projects the manifold Σ defined by dψ φ0 + 2φ/η = ψ , (52) the coordinates (x, t, ρ, u, p) onto the manifold defined dη η − ψ by (t, ρ, u, p), and reduces the coordinate cover by one Taylor [1]; also see Sachdev [30]. Here the prime notation variable. The compatibility equations and the coordinates represents differentiation with respect to η. Recall that characteristic curves may be solved at the points were the Fig. 3 shows plots of the numerical solutions of Eqs. (50), characteristic curves intersect in space and time. (51), and (52) for the limiting values of the strong shock The solution of the blast wave problem requires three conditions of Eqs. (4) to (6). linearly independent differential one-forms for the flow Submitted to ANS/NT (2021) LA-UR-20-30042 (v3)

field properties (ρ, u, p). Along pathlines and Mach lines are equivalent to

in the flow field, the governing one-forms are given by the j compatibility equations (43) and (44): Ω ∧ dθ = 0 for j = 1, ..., m − n, (60) θ1 = dp − a2dρ, (56) where Ω = θ1 ∧ θ2 ∧ θ3 6= 0. (61) 2 2 2ρua θ = dp + ρadu + dt, (57) Moreover, it is shown in Flanders [31] that Eqs. (59) assure x the existence functions f i and gj for i, j = 1, ..., m − n 2ρua2 j θ3 = dp − ρadu + dt. (58) such that x i i 1 i m−n The analysis of the EDS defined by Eqs. (56) to (58) does θ = f1 dg + ··· + fm−n dg . (62) not use Eq. (41) for the pathline explicitly, because the pathline equation may be represented as a linear com- Equations (62) imply that integrating factors exist which bination of the Mach lines of Eqs. (42). If this system allow a EDS to be represented in integral form. of exterior differential forms is thought of a system of The Frobenius theorem is satisfied for the EDS model- equations in (t, ρ, u, p), then the theorem of Section 5.1 ing the blast wave if the one-forms of Eqs. (56) to (58) are guarantees that a local solution exists to the system of linearly independent ODEs associated with Eqs. (56), (57), and (58) at points θ1 ∧ θ2 ∧ θ3 6= 0, (63) where the equations are not singular and where the char- acteristic curves intersect. and if On the other hand, if the EDS defined by Eqs. (56), 1 2 3 i (57), and (58) is thought of as a system of differential one- θ ∧ θ ∧ θ ∧ dθ = 0, for i = 1, 2, 3. (x, t, ρ, u, p) forms defined on and the equations for the are satisfied using Eqs (56) to (58). Direct calculation of Mach lines are not used initially to restrict the number of the wedge product produces: independent variables to one, the theorem of Section 5.1 does not apply. For this case, the problem coordinates are ρ2ua3 (x, t, ρ, u, p) θ1∧θ2∧θ3 = −4 dt∧du∧dp−2ρa3 dρ∧du∧dp and an integration theorem must be applied x that holds for more than one independent variable. ρ2ua5 The Frobenius theorem of differential geometry gives − 4 dt ∧ dρ ∧ du 6= 0. (64) x integrability conditions for a system of differential one- forms on a surface. This theorem extends the concept of Moreover, calculation of the exterior derivative of θ1, θ2, integrating factors from a single equation to a system of and θ3 gives: equations. The Frobenius theorem has at least two nat- dθ1 = −2a da ∧ dρ, (65) ural forms: one which gives a functional representation of differential one-forms, and one which expresses an ua2 ρa2 dθ2 = a dρ∧du+ρ da∧du−2 dt∧dρ−2 dt∧du equivalent concept in terms of the closure properties of a x x differential ideal defined by differential one-forms. Fol- ρua ρua2 lowing Ivey and Landsberg [21], the Frobenius theorem − 4 dt ∧ da − 2 dx ∧ dt, (66) x x2 in terms of differential one-forms is: and Theorem Let Σ be a C∞ manifold of dimension m, and 1 m−n 1 2 2 let θ , ..., θ ∈ Λ (Σ) be pointwise linearly indepen- 3 ua ρa i 1 dθ = a dρ∧du−ρ da∧du−2 dt∧dρ−2 dt∧du dent. If there exist one-forms αj ∈ Λ (Σ) such that x x j j i ρua ρua2 dθ = αi ∧ θ for all j, then through each point z ∈ Σ − 4 dt ∧ da − 2 dx ∧ dt. (67) there exists a maximal connected n-dimensional manifold x x2 j j i i : N,→ Σ such that dθ = αi ∧ θ for 1 ≤ j ≤ m − n. To simplify Eqs. (65) to (67), recall that for the speed of This manifold is unique, in the sense that any other such sound a2 = γp/ρ, so that connected submanifold through z is a subset of i(N). γ γp 2a da = dp − dρ. (68) The Frobenius theorem will be applied to show that the ρ ρ2 EDS defined by Eqs. (56) to (58) is integrable. Notice that Then combining Eq. (68) with Eqs. (65) to (67) yields: the conditions 1 γ j j i dθ = dρ ∧ dp, (69) dθ = αi ∧ θ for i, j = 1, ..., m − n, (59) ρ Submitted to ANS/NT (2021) LA-UR-20-30042 (v3)

γ γp dθ2 = a dρ ∧ du − du ∧ dp − dρ ∧ du The key points of this study may be summarized as: 2a 2aρ 1. Lie group methods were applied to derive Taylor’s fa- ua2 ρa2 γu − 2 dt ∧ dρ − 2 dt ∧ du − 2 dt ∧ dp mous two-fifths dimensional result relating the position x x x of the blast wave R to the time t elapsed in the explosion γup ρua2 + 2 dt ∧ dρ − 2 dx ∧ dt, (70) as well as the total energy released. The use of Lie group ρx x2 methods demonstrates that a basic relationship exists be- tween the geometry (or symmetry) and physics of wave propagation through the equations of motion. γ γp dθ3 = a dρ ∧ du + du ∧ dp + dρ ∧ du 2. The set of self-similar ordinary differential equations 2a 2aρ derived by Taylor to model the point blast wave were ua2 ρa2 γu − 2 dt ∧ dρ − 2 dt ∧ du − 2 dt ∧ dp shown to be equivalent to an exterior differential system x x x that has a local solution for the velocity, pressure, and γup ρua2 + 2 dt ∧ dρ − 2 dx ∧ dt. (71) density along a curve representing the motion of the blast ρx x2 wave. Combining Eq. (64) with the wedge product of Eqs. (69), 3. The method of characteristics for one-dimensional, (70), and (71) implies unsteady, compressible flow was shown to be equivalent to the method of characteristics for one-dimensional, un- θ1 ∧ θ2 ∧ θ3 ∧ dθ1 = 0, (72) steady, compressible, potential flow. The equations of the method of characteristics for spherically symmetric, and unsteady, compressible flow were then shown to yield a natural exterior differential system modeling the point θ1 ∧ θ2 ∧ θ3 ∧ dθi = blast wave along Mach lines. ρ2ua5 4. The exterior differential system modeling the point 4 dx ∧ dt ∧ dρ ∧ du ∧ dp, for i = 2, 3. (73) x2 blast wave derived from the method of characteristics was shown to be locally integrable, that is, to have local Finally, restricting the above differential form to the Mach solutions for velocity, pressure, and density at the points lines, dx = (u ± a)dt, implies that the right-hand-side of in space and time where the characteristic curves intersect. Eqs. (73) are linearly dependent, so it follows that: This article has shown that the solution of the point blast wave problem is directly related to the symmetry prop- 1 2 3 i θ ∧ θ ∧ θ ∧ dθ = erties of the motion of the underlining shock wave. The ρ2ua5(u ± a) Lie group symmetry techniques generalize to the study of 4 dt ∧ dt ∧ dρ ∧ du ∧ dp = 0 x2 exterior differential systems which provide a theoretical for i = 2, 3. (74) framework for analyzing shock wave propagation prob- lems in multiple spatial dimensions with the method of Hence, the conditions of the Frobenius theorem are satis- characteristics. Historically, the method of characteris- fied for the exterior differential system defined by Eqs. (56) tics has been used mainly to compute compressible flow to (58) along the Mach lines and the compatibility equa- problems in one space dimension and time because the tions modeling the blast wave are integrable at the nonsin- characteristics reduce to planar curves and the equations gular points where the characteristic curves intersect. of motion reduce to ODEs. The geometry (or symmetry) and physics of complex VI. SUMMARY AND DISCUSSION shock wave phenomena may be studied in detail by ap- This article presented a brief historical overview of plying the methods of EDS to guide and interpret numeri- G. I. Taylor’s solution of the point blast wave problem and cal simulations that have higher geometric precision and its applications to the analysis of the Trinity atomic test in greater physical fidelity than analytic models. For ex- 1945. The blast wave solution formulated by Taylor, which ample, an inviscid flow field with low levels of vorticity used both dimensional analysis and similarity techniques may be approximated as a potential flow, which reduces to simplify the equations of motion, resulted in a powerful a system of first-order PDEs for the velocity vector to mathematical model that yielded fundamental insight into a single second-order PDE for a velocity potential. Re- the physics of blast waves. This article further showed call that the point blast wave may be modeled with the that the solution methods applied by Taylor are examples potential equation (45), which is equivalent to the Euler of the practical manifestation of the symmetry properties equations (1), (2), and (3) for spherically symmetric, isen- associated with the physics of shock wave propagation. tropic, compressible flow. Bryant, et al. [32] have used Submitted to ANS/NT (2021) LA-UR-20-30042 (v3)

the theory of exterior differential systems to analyze rigor- 8. Deakin, M. A. B., “G.I. Taylor and the Trinity Test,” ously the integrability of a general class of second-order International Journal of Mathematical Education in PDEs which includes the equations for compressible po- Science and Technology, 42, No. 8, 2011. tential flow. A detailed understanding of compressible potential flow would provide insight into the basic physics 9. Bethe, H. A., Fuchs, K., Hirschfelder, J. O., Magee, J. of shock wave motion and help explain and quantify more L., and Neumann, J. von, “Blast Wave,” Los Alamos complex physical effects found in numerical simulations National Laboratory, 1958. such as the interaction of shock waves and the generation 10. Sedov, L. I., Similarity and Dimensional Methods in of vorticity. Mechanics, CRC Press, 1993.

ACKNOWLEDGEMENTS 11. Korobeinikov, V. P., Problems of Point Blast Theory, The authors would like to thank Richard Moore of Springer, 1991. the Atomic Weapons Establishment, as well as Ralph Menikoff, Christopher Triola, Joe Schmidt, Cory Ahrens, 12. Kamm, J. R., “Evaluation of the Sedov – von Neu- Len Margolin, and Mark Chadwick of the Los Alamos mann – Taylor Blast Wave Solution,” Los Alamos National Laboratory for their many helpful suggestions in National Laboratory, LA-UR-00-6055, 2000. the development of this article. This work was supported 13. Birkhoff, G., Hydrodynamics: A Study in Logic, Fact, by the US Department of Energy through the Los Alamos and Similitude, Princeton University Press, 1950. National Laboratory. Los Alamos National Laboratory is operated by Triad National Security, LLC, for the National 14. Bridgman, P. W., Dimensional Analysis, Yale Univer- Nuclear Security Administration of the US Department of sity Press, 1931. Energy under Contract No. 89233218CNA000001. 15. Ovsiannikov, L. V., Group Analysis of Differential REFERENCES Equations, Academic Press, 1982. 1. Taylor, G. I., “The Formation of a Blast Wave by a 16. Olver, P. J., Applications of Lie Groups to Differential Very Intense Explosion I. Theoretical Discussion,” Equations, Springer-Verlag, 1986. Proceedings of the Royal Society of London, A201, pp. 159–174, 1950. 17. Andreev, V. K., Kaptsov, O. V., Pukhnachev, V. V., and Rodionov, A. A., Applications of Group- 2. Taylor, G. I., “The Formation of a Blast Wave by a Theoretical Methods in Hydrodynamics, Springer, Very Intense Explosion II. The Atomic Explosion of 1998. 1945,” Proceedings of the Royal Society of London, A201, pp. 175–186, 1950. 18. Spivak, M., A Comprehensive Introduction to Differ- ential Geometry, Volume One, Second Edition, Pub- 3. Barenblatt, G. I., Scaling, Cambridge University lish or Perish, Inc., 1979. Press, 2003. 19. Krasil’shchik, I. S. (editor), Vinogradov, A. M. (ed- 4. United States Nuclear Tests, July 1945 through itor), et al., Symmetries and Conservation Laws September 1992, U.S. Department of Energy, Nevada for Differential Equations of Mathematical Physics, Operations Office, DOE/NV-209-REV 15, December American Mathematical Society, 1999. 2000. 20. Vinogradov, A. M., Cohomological Analysis of Par- tial Differential Equations and Secondary Calculus, 5. Selby, H. D., et al., A New Yield Statement for the Trin- American Mathematical Society, 2001. ity Nuclear Test, 75 Years Later, Los Alamos National Laboratory, LA-UR-20-30144, October 2020. 21. Ivey, T. A., and Landsberg, J. M., Cartan for Be- ginners: Differential Geometry via Moving Frames 6. Clark, R. W., The Birth of the Bomb: The Untold and Exterior Differential Systems, Second Edition, Story of Britain’s Part in the Weapon that Changed American Mathematical Society, 2016. the World, Phoenix House, 1961. 22. Zel’dovich, Ya. B., and Raizer, Yu. P., Physics of 7. Barenblatt, G. I., Scaling, Self-Similarity, and Inter- Shock Waves and High-Temperature Hydrodynamic mediate Asymptotics, Cambridge University Press, Phenomena, Academic Press, 1966 (Volume I) 1967 1996. (Volume II). Submitted to ANS/NT (2021) LA-UR-20-30042 (v3)

23. Ramsey, S. D., and Baty, R. S., “Symmetries of the Gas Dynamics Equations Using the Differential Form Method,” Journal of Mathematical Physics, 58, No. 11, 2017. 24. Giron, J. F., Ramsey, S. D., and Baty, R. S., “Scale Invariance of the Hometropic Inviscid Euler Equa- tions with Application to the Noh Problem,” Physical Review E, 101, No. 5, 2020. 25. Hutchens, G. J., “Finite–Strength Shock Propagation for Alternative Equations of State,” Ph.D. Thesis, Uni- versity of Illinois, 1991. 26. Cantwell, B., Introduction to Symmetry Analysis, Cambridge University Press, 2002. 27. Rusanov, V. V., “Characteristics of the General Equa- tions of Gas Dynamics,” Zhurnal Vychislitel’noi Matematiki i Matematicheskoi Fiziki, (3), No. 3, 1963.

28. Zucrow, M. J., and Hoffman, J. D., Gas Dynamics, Volume II: Multidimensional Flow, John Wiley & Sons, 1977. 29. Sauer, R. Anfangswertprobleme bei Partiellen Differ- entialgleichungen, Zweite Auflage, Springer-Verlag, 1958. 30. Sachdev, P. L., Shock Waves and Explosions, Chap- man & Hall/CRC, 2004. 31. Flanders, H., Differential Forms, Academic Press, 1963.

32. Bryant, R. L., Chern, S.S., Gardner R. B., Gold- schmidt, H. L., and Griffiths, P.A, Exterior Differ- ential Systems, Springer-Verlag, 1991.