Preface This is a book about continuum mechanics, electrodynamics and the mechanics of electrically polarized media, and gravity. It differs from other ‘books on these subjects in thatthe theories are formulated by means of the principle of least action. The chiet advantage of this formulation is that it is simple and easy, since the physical input into the theory consists of choosing only one scalar function of the fields involved—the Lagrangian density. Thus my chiet ‘motivation for writing this book was to make the physical content of the classical subjects discussed available to students of physics in a con- ceentrated, easy to digest form. My second goal has been to explain the general structure of field thoories based on the least action formulation. 1 hope that this discussion of general principles in a purely classical setting will be useful to students fof quantum field theory, since quantum field theory is almost without exception based on the principle of least action Its also my hope that my colleagues in quantum field theory will enjoy secing their everyday methods applied to such subjects as the motion of fluids ‘Although I have borae the applications fo quantum field theory in mind, Thave tried to remain true to the subject of classical field theory. | have therefore avoided interesting diversions that do not bear on real classical phenomena and have omitted such subjects as spin 4 fields, non-Abelian ‘gauge fields, and Poisson brackets Physicists often reyard Hamilton's principle of least action as the start- ing point for particle mechanies. However, modern workers in continuum mechanics almost always prefer Newton's formulation of mechanies, in which one writes the equations of motion directly instead of writing an action and deriving the equations of motion, Their preference is not without good reason, for they seek to describe both mechanics and anProtece thermodynamics—that is, both entropy conserving phenomena, which generally can de described by an action principle, and entropy increasing Phenomena, which generally cannot, In this book, I have followed the usual physicists’ inclination to sweep “friction” under the rug and discuss conservative forces only—at least to begin with, In the final chapter, Aissipation is resurrected and added onto some simple theories. When (a), is the same no matter what orthogonal coordinate system is used to measure the 4, = F454, 2 ray ‘The orthogonality condition (1.1.2) implies that the determinant of the matrix @ is either +1 or —1. The additional requirement det® = +1 eliminates those @ that express the transformation bevween a right handed coordinate system and a left handed coordinate system. The same method shows that the dot product A+ B=3,4,B, between two vectors is invariant under rotations of the coordinate systems. We write out the proof using a simplified notation in which all summation signs are omitted; any latin index i,j,k, ete. that appears twice in a product Rotational Covariance | is assumed to be summed (Be Tensor In the development that which are a simple gene tensor itself that usually 2 particular coordinate ()* components in a four indices, each of which T in two coordinate systems, A first rank tensor is physically as an arrow. physical representation for: ‘mathematically respectable: a rule for assigning the components obey the of visualization, @ tensor physical object, whereas coordinate system depend ‘One can easily give: ‘equation defines a second rank Indeed, higher rank tensors in this way. If Ais. then the equation Tay.» New tensors can also be multiplying by ceal numt then Td), c+ BBLFlelds and Transformation Laws. and A, are the components relation between J, and Ay (1) (hay (11.3) p= |, 6,2™ 83 82)=82,= 35, condition (1.1.2) is that the Pythagorean formula Goordinate system is used to Hu), that the determinant of the Bonal requirement det® = +1 formation between a right coordinate system, fuct A» B=3,4,B, between the coordinate Systems, We ion in which all summation ‘that appears twice in a product, Rotational Covariance 3 is assumed to be summed from | 10 3 BAA ay Bay B = ABD AAs Tensors In the development that follows we need to use objects called tensors, which are a simple generalization of vectors. As with vectors, itis not the tensor itself that usually appears in equations, but rather its components in 4 particular coordinate system. For instance, a “fourth rank” tensor T has (3)* components in a particular coordinate system, which are labeled by four indices, each of which takes the values 1,2,3: T,,,. The components of T in two coordinate systems are related by the simple rule pam By By Bay Hay Tyee (lay |A first rank tensor is the same as @ vector and can be represented physically as an arrow. Unfortunately, there is no intuitively appealing physical representation for a higher rank tensor. One is foreed to the mathematically respectable but physically cffete definition of a tensor T as aa rule for assigning components Ty, to each coordinate system such that the components obey the transformation law (1.1.4), Despite the difficulty of visualization, a tensor 'T, like a vector, is to be thought of as a real physical object, whereas the components T,,, measured in a particular coordinate system depend on which coordinate system was used. ‘One can easily give examples of tensors. If A and B are vectors, then the equation, By defines a second rank tensor, T= AB= (Ry AN Ay B,) 2 Bp Bog Te Indeed, higher rank tensors can always be formed from lower rank tensors in this way. If A is a tensor of rank N and B is a tensor of rank M, then the equation Ty ay.4=Ay,,cB), .» defines a tensor T of rank N+ M. New tensors can also be formed from old tensors of the same rank by multiplying by real numbers and adding: if A and B are tensors of rank then 7, ,= 04, ,+BB, , defines a tensor T=aA+/B.Floids and Transformation Laws Fortunately it is also possible to form lower rank tensors from higher rank tensors, so that we will not be caught in an inevitable population ‘explosion of tensor indices. If T is a tensor of rank 3, say, then AnTy defines a tensor of rank 1, A. The proof is casy, and has an obvious generalization to any rank: Rial yp Bia) Tate = Ry BT hin Fa Ta = eA ‘This method for forming new tensors is called contraction of indices. An important special case is the formation of scalar (that is, an ‘ordinary real number) by contracting the indices of a second rank tensor. For cxample, from (wo vectors A and B we can form the tensor with ‘components 4,8,, then the scalar A+ B= 4,8, There are two special tensors with the remarkable property that their ‘components do not vary from one coordinate system to another. The fitst ’s the “unit tensor” 8, whose components in any coordinate system are 8, the Kronecker delta. To see that this definition obeys the transformation lawe(1.1.4) write iy Hy By = an Re = By The other invariant tensor is called the completely antisymmetric tensor € sand has components fay n= n= +). éar7@n7 n= =H (LL5) All other components are zero. This rule can be summarized by saying that j= +1 and 6 is antisymmetric under interchange of any two indices (80 6, =— ty). To see that this rule obeys the transformation law (1.1.4) we note that the same object 6, appears in the definition of the de- terminant of « matrix: et R= By, By, Bae Ge Lorentz Covariance Thus, using der %, The invariant tensor € the definition of angular notation 8s L,= «RP. ‘The method for ensuring, covariant under rotations and tensors to represent form “(scalar)= (sealar)”, tensor)=(N" rank tensor)” 1.2, LORENTZ COV: Two observers whose other will discover the equations expressing invariance manifest whose reference frames are fone another must also dis investigate how to write manifest. ‘Observers and Inertial The special theory of coordinates x,y,z and the’ space assigns to an event oor, in the canonical, that the observer lays out another and that at the federates’ with clocks. The determined by noting whi him for his coordinates (3 event occurred. We i coordinate system. That is, confederates and subjected We will shortly make light is a certain constantletés and Transformation Laws rank (ensors from higher in an inevitable population fof rank 3, say, then is easy, and has an obvious )Tain Tay contraction of indices tion of a scalar (that is, an iees of a second rank tensor. ‘we can form the tensor with warkable property that their system to another. The first any coordinate system are 8, ion obeys the transformation aay letely antisymmetric tensor € +1, (ts) bbe summarized by saying that aerchange of any two indices the transformation faw (1.1.4) jin the definition of the de- Sy Lorentz Covariance 6 Thus, using det ® = 1, f Ry Hy Bor Sy = (+ Dion ‘The invariant tensor ¢ appears in the vector eruss product. For example, the definition of angular momentum, L=R x P, is written in the present notation as 1; «RP, ‘The method for ensuring that equations expressing physical laws are covariant under rotations should now be clear. One uses sealars, vectors, and tensors to represent physical entities and writes physical laws of the form “(scalar)= (sealar)”, “(vector)=(vector)”, or, in general, “(N" rank tensor)=(V" rank tensor)” using the component notation, 1.2. LORENTZ COVARIANCE ‘Two observers whose reference frames are rotated with respect to ench other will discover the same physical laws. We have seen how to write equations expressing physical laws in a way that make this rotational invariance manifest. According to the principle of relativity, (wo observers whose reference frames are moving with constant velocity with respect to fone another must also discover the same physical laws. We will now investigate how to write equations that make this “Lorentz” invariance manifest Observers and Inertial Reference Frames ‘The special theory of relativity begins by focusing attention on the space coordinates x,y,z and the time coordinate that an observer in gravity free space assigns 10 an event (for instance, the collision of two billiard balls, or, in the canonical example, the exploding of a firecracker), We imagine that the observer lays out a grid of meter sticks at right angles to one another and that at the intersections of the meter sticks he places con- federates with clocks. The coordinates (i,x,y,2) of an event are 10 be determined by noting which confederate was next to the event and asking him for his coordinates (x,y,2) and the time ¢ shown on his elock when the event occurred. We imagine that the coordinate sysiem is an “inertia” coordinate system. That is, a body placed at rest in front of any of the confederates and subjected to no forces will remain at rest. ‘We will shortly make use of a remarkable physical law: the speed of light is a certain constant e=3%10" m/sec, independent of any char‘ Flelds and Transformation Laws acteristics of the emitting or detecting devices. As a matter of convenience, we will assume that our observers choose their time and distance scales in such a way that this constant is c=1. (For instance, 107? sec is a convenient unit of length, being about equal to 1 ft). ‘To express this light propagation law in a compact form, let us rename the coordinates of an event x*, 4=0,1,2,3, with ams, x's xx7=y,x° Suppose two events labeled by x}! and xf could be connected by a light signal, That is, a pulse of light emitted at x,,,2, at time f would reach XqYa.Zp at time ¢. Then the fact that the speed of light is | implies that (f~ 4) (xq x) +049, #(Z,— 2)? In @ more compact notae tion this is (x88) Bo" 21) =0. (121 Here we have adopted the “Einstein summation convention” that repeated Greek indices are to be summed over the values 0,1,2,3. The matrix g,,, called the “metric tensor,” is Bo=—h 8u=S= sae +h (1.22) Be wee. Lorentz Transformations Consider now two inertial coordinate systems, @ and ©. The coordinates (6.3.5.2) and (,x,9,2) assigned by the two sets of observers to the same event are related by a set of functions Fl: ¥*= F*(x°,x4,x7,x°), We wish to determine the form of the functions F(x) by imposing the requirement teat both observers find that the speed of light is 1 [FM G2)— FO) gL FG) FO] 20 if and only if (38-1) (38-2) =0. (123) ‘The requirement (1.23) is very restrictive. Let us make the additional reasonable technical assumptions that the functions F°(x) are twice con- "imuously differentiable forall x" and thatthe x? are related to the "by a twice continuously differentiable inverse mapping x" = F*(X). One can show* that the only such functions F* which satisfy (1.2.3) have the simple form RPS PM(x)mAA x tah, ‘The proof is abit involved if one does not assume at the start thatthe transformation & tinea. See Problems 4 and 5 Lorentz Covariance where a* are any four numbem| with the property a The matrix A is called a the structure of Lorentz numbers a” apparently tell the} ‘coordinate system as seen seale factor that arises observers to use the same Tet us now require that all for instance, by standardizing light emitted in a certain scale factor 2 must always be For this purpose, consider event 10 be the origin of ther are related by 3*=AAR x cannot be two observers with matrix Af, but different f these two observers would be: would manifestly not agree om of the coordinate transformasi allowed coordinate systems Furthermore, any other there is nothing special abowt ‘We must now ask whet Lorentz transformation A. FaAAAA AY we find law aM But it is a simple mathemati which satisfies this law is . A) ‘Thus the coordinate tras scale, and leave the origin transformations”: If we allow the origin of,a Jane Tronstormation Laws ‘matter of convenience, and distance seales in 10? sec is a form, let us rename wnt xisixxteyx? bbe connected by a fw? at time f, woul, of light is 1 implies ‘@ more compact nota- (124) tion” that repeated 0,1,2,3. The matrix g,,, (122) aad ©. The coordinates at observers to the same a9 xt,x?,x°), We wish Famposing the requirement (1.23) ws make the additional FAC) are twice con- Fare related to the ¥* bya at= FA(%). One can (1.23) have the simple art thatthe transformation is Lorentz Covariance 7 where a” are any four mumbers, ) is a positive number, and AM, is a matrix with the property APA Bae = Bap (1.24) ‘The matrix A is called a Lorentz transformation matrix; we will discuss the structure of Lorentz transformations in some detail presently. The numbers a” apparently tell the coordinates of the origin x*=0 of the © coordinate system as seen in the © system, The positive constant is a scale factor that arises because nothing we have said so far requires both observers to use the same scale of length Let us now require that all observers choose the same scale of length— for instance, by standardizing their meter sticks against the wavelength of light emitted in a certain atomie transition. Then we can argue that the seale factor A must always be 1. For this purpose, consider the class of observers who agree on a single event to be the origin of their coordinate systems. Any two such observers are related by 3*=AAF,x". Let “me” be one of the observers. There cannot be two observers with coordinates related to mine by the same matrix At, but different factors 2, say 3*=A,.\4,x” and TP—AyA* x, for these two observers would be related to each other by #*=(,/A)x* and ‘would manifestly not agree on the standard of length, Thus the parameters, of the coordinate transformations that relate my system to other physically allowed coordinate systems must be related by an equation ANA). Furthermore, any other observer must obtain the same function M(A), since there is nothing special about “me” as an observer. We must now ask whether A(A} can have a nontrivial dependence on the Lorentz transformation A. By comparing T=A(A, A,X, X=A(A)Agx with (A,An)A,Agx we find that the function A(A) obeys a multiplication MAAR) But it is a simple mathematical fact (see Problem 6) that the only function which satisfies this Iaw is (AyA2)- MAJ=1 for all A ‘Thus the coordinate transformations that keep c= 1, preserve the length scale, and leave the origin of coordinates unchanged are the “Lorentz, transformations”: BPs AM x" (1.25) If we allow the origin of coordinates to be changed we have a “PoincaréLorentz Covariance terms of the second Fre Meat, ‘We now turn our attention (0 the structure of the matrices A, The Structure of Lorentz. Transformations Since Lorentz invariance is an important tool in this book, it pays us to investigate what kind of matrices A have the property (1.2.4). This property is very similar to the defining property of a rotation matrix: Ba Hy By Bey moving with a constant to the second. ‘One might expect on the! Lorentz transformation cam! first a rotation of the axes off with the direction of mo another rotation. Thus amy i the form (127) In fact, it is clear from comparing (1.2.7) and (1.2.4) that rotations are a subset of all Lorentz transformations, That is, if @ is a 3%3 rotation matrix, and the 4x4 matrix A(R) has the form . {1 0 oO a Gy Fr Foy where ACG) and ACG?) ane AQ Fy Ay Aas" (29) [0% By Hay ‘An explicit construction: most general transh ‘cause we have not eo then A is a Lorentz transformation. Another important special example of, Lorentz transformations, called a “boost” in the z-direction, is Ay cosh 0 0 sinho » [0° 10 o A(w)*,= 1.2.9) lo ot 0 i The transformation Ay sinho 0-0 coshe ordinate system to those im ansformation. The & run forward to a system im time reversal transformati (124) is either one of the Ay Ay oF Ap where the “angle” w can be any real number. Imagine two coordinate systems related by @ boost A(w). ‘The path through space-time of the first observer, who sits at the origin of his space coordinates and watches his, clock run, is given by (x,y,2)=(0,0,0), f=anything. This path is given in@ Flelde and Transformation Laws, . (1.2.6) of the matrices A. tool in this book, it pays us to property (1.2.4). This property rotation matrix: (27) Jand (1.2.4) that rotations are a is, if @ is a 3X3 rotation o has By ay (1.28) important special example of the z-direction, is sinhw ) 0 0 coshes (1.29) ber. Imagine two coordinate through space-time of the first coordinates and watches his anything. This path is given in Lorentz Covariance ® terms of the second coordinate system by or F==0, 707 where p= tanh w. Thus the first coordinate system is moving with a constant velocity tanh © in the z-direction with respect, to the second. ‘One might expect on the basis of physical intuition that the most general Lorentz transformation can be built from three successive transformations first a rotation of the axes of the © system so that the new z-axis is aligned with the direction of motion, then a boost in the (new) z-direction, then another rotation. Thus any Lorentz transformation matrix A should have the form A= MR)", (6) ACB)", (12.10) where A(R) and ACG’) ate rotations as in (1.2.8) and A(w) is a boost as in (1.2.9), ‘An explicit construction (see Problem 8) shows that (1.2.10) is almost the most general transformation satisfying (1.2.4). The exception arises be~ cause we have not considered the transformations 1 0 0 oO o -1 0 Aly 0-18 ose oaeiott and -1 000 eal ete aga] 2108 a2 : ooo1 The transformation Ap relates the coordinates in’a right hanied co- ‘ordinate system to those in a left handed system, and is called the parity transformation, The transformation A, relates a system in which clocks run forward to a system in which clocks run backward, and is ealled the time reversal transformation. The most general transformation satisfying (1.24) is either one of the type (1.2.10) or else one of this type followed by Ap, Ap OF Ap10 Flelds and Transformation Laws. We will often refer 10 the transformations (1.2.10) made up from rotations and boosts as “Lorentz transformations.” A more precise (but somewhat stuffy) name is “proper orthochronous Lorentz transforma tions," to distinguish them from transformations containing A, or A, or both. The proper orthochronous transformations can be distinguished by the requirements ea det A= +1, A> 1 (1.2.12) in addition to the requirements that A",A"pSq, Bag. (See Problem 7.) We will demand that physical laws be invariant under (proper ortho- chronous) Lorentz transformation of the coordinates. The reason that we Go not at the outset require invariance under Ap and) Ay as wel is that the real world does not have this extra invariance im the so called “weak interactions” of nuclear physics. Nevertheless, the theories we discuss in all Dut the last chapter of this book turn out to be invariant under A and A. (in the last chapter we introduce irreversible processes like friction, and thus lose invarianee under A;-) Four-Vectors We can write Lorentz covariant equations by generalizing the familiar methods discussed in the preceding section for writing. rotationally covariant equations. We begin by defining a four-vector V 10 be a rule which assigns to each inertial reference frame four numbers °, V', V2, V3, called the components of V in that reference frame; the components P* in one reference frame are to be related to the components * in another frame by the transformation Iqw ri ASV", (12.13) where A is the Lorentz transformation matrix relating the two reference frames: F*=A",x’+a". Thus the archetype four-vector is the vector whose components are the coordinate differences x*—y? between two eveats. The fundamental property (1.24) of Lorentz transformations implies tat the product u- Ue VP of two four-vectors is a scalar (ie., independent of reference frame). It is often convenient to write such products as U+ V= UY, where Ven bel (1.2414) Lorentz Covariance The components Vo, V1, Vana covariant components of Ws called the contravariant om ‘covariant components of a ‘The components V* cam ‘multiplying by the inverse same numerical values: Thus ‘The notation is arranged sa’ always an upper index and Let us mention here t moving point partic, its ¥8 particle is conveniently deses path in space time x*¥=x* clearly a four-veetor, Since is convenient to define a [-why =I ut(oy| ‘The vector w* is called the: three-velocity of the particle %a ‘Flelde and Tranetormation Laws (1.2.19) made up from jons.” A more precise (but mous Lorentz transforma containing A, or Ay oF ns can be distinguished by (1212) = og: (See Problem 7.) ‘Ay and A, as well is that the in the s0 called “weak ‘the theories we discuss in all invariant under A, and Ay. processes like friction, and by generalizing the familiar jon for writing rotationally ‘a four-vector V to be a rule four numbers V°, V1, V2.3, frame; the components V* in components F* in another 1.2113) relating the two reference four-vector is the vector ences x*—y" between two sntz transformations implies lent of reference frame). It is U-+V=UtY, where (1.2414) Lorentz Covariance " ‘The components Mo, Vy,V2,V3 of V defined in this way are often called the covariant components of V: the original components V°V',V%;V9 are called the contravariant components. The transformation law for the covariant components of a vector is FAWN Vn (1213) ‘The components 7* can be recovered from the components V, by multiplying by the inverse matrix to g,,, whieh is called g* and has the same numerical values: (1.2.16) ‘Thus Vem gry, (1247) ‘The notation is arranged so that when two indices are summed over, one is always an upper index and the other is a lower index Let us mention here two important four-vectors associated with a moving point particle, its four-velocity and its momentum, The motion of a particle is conveniently described by giving a parametric equation for its path in space time x"=.xM(a). The tangent vector dxF/ da to the path is clearly a four-veetor. Since the length of this vector has no signifieance, it is convenient to define a normalized tangent vector u* with length [-wy 221 (1.2.18) The vector u* is called the four-velocity of the particle. The ordinary three-velocity of the particle is dich dst ut oes (1.2.19)(1.220) If we multiply w* by the mass of the particle we obtain a vector P= mut called the four-momentum of the partcle. P° is identified as the energy of the particle and P',P%,P? are the components of its three ‘momentum. There are two main reasons for this indentifieation. First, if the velocity of the particle is small compared to the speed of light we recover the nonrelativistic expressions for momentum and energy, except for a constant term B,= me? in the energy’ Pha mot = mot, Vie mt Lm vi-¥ Second, it is found from experiment that when fast moving particles collide the total four-momentum P*= Pf+ P+... of the system is conserved. Given the development so far, the definition of four-tensors will come as no surprise. A third rank tensor T, for instance, has components 7** that transform according to AR ATT (1221) ‘Any of the components of T can be lowered using the metric tensor, as in T*,!= gy TF". The components 7*," transform according to Tegra As (A) gAU TY. (1.2.22) We can form higher rank tensors by multiplying lower rank tensors together, as in T= 4°R?, just as we did with three-tensors. We can also form lower rank tensors from higher rank tensors by multiplying by g,, and summing, as in S*= 7%, Such equations are usually written in the more compact form $#=7%,"or $*= 74,5. Up to this point, we have been careful to distinguish between a tensor T ‘and its components T",T*,,7,, or T,’ in some coordinate system. In the sequel, however, we will Cconomize’ on words by writing “the tensor ‘Seatar, Vector, and Tensor Flee TP =A"B? instead of coordinate system are related =A*B"”. The reward in Invariant Tensors, ‘There are two important formations (ie, they have the’ ‘The first is the metric tensor (sy is the defining property of a La ‘The second invariant tensor which is defined by cP" ig antisymmetric Using the definition of the: AAA GAP, ‘The determinant of a Lorentz A does not include a parity components of € are transformations, One must covariant components ¢ array of numbers, but éraa= There are no unexpected Every invariant tensor cam <", For example, the most where «,8,y,8 are arbitrary 13 SCALAR, VECTOR, AND] ‘A function that associates a called a sealar field. Am« Flelds and Transtormation Laws O (1.220) particle we obtain a vector partcle. P° is identified as the the components of its three- for this indentification. First, if red to the speed of light we ‘momentum and energy, except fast moving particles collide . of the system is conserved. of four-tensors will come as ee, has components T*” that re (1221) using the metric tensor, as in form according to PTY. (1222) ihiplying lower rank tensors ith three-tensors. We can also tensors by multiplying by 4, jons are usually written in the distinguish between a tensor T fsome coordinate system. In the words by writing “the tensor ‘Seatar, Vecior, and Tensor Fields 2 T= A*B?” instead of “the tensor T whose components in each coordinate system are related to the components of A and B by T¥ = A*B™, The reward in simplicity of language is considerable. Invariant Tensors ‘There are two important tensors that are invariant under Lorentz trans formations (ic. they have the same components in all coordinate systems). ‘The first is the metric tensor g,,. Indeed, the equation (ANY Be at is the defining property of a Lorentz transformation. ‘The second invariant tensor is the completely antisymmetric tensor ¢*™, which is defined by OMe 1, (1.2.23) ct" is antisymmetric under interchange of any pair of indices. Using the definition of the determinant of a matrix we find that AMA? g AP, APgc%P8 = (det Alem”, (1.2.24) ‘The determinant of a Lorentz transformation matria A is + | provided that ‘A does not include 2 parity of time reversal transformation. Thus the ‘components of ¢ are unchanged under proper orthochronous Lorentz transformations. One must be careful about the sign of ¢: clearly the covariant components «,,, of € also form a completely antisymmetric array of numbers, but eras — 1 ‘There are mo unexpected new invariant tensors waiting to surprise us. Every invariant tensor can be formed from sums of products of g,, and ‘ | 8 Show that every Lorentz the form A= AAA Ay boost in the z-direction, amd Show that a fourth rank scalar field: Tyo)= Bee") To verify that this formula defines a tensor field one must check the transformation law (1.3.3) using the chain rule for partial differentiation: TL FOYT) Apparently this result can be generalized. If 7°” is a tensor field, then SEP GT =(G/dx*V8/Ax")--- (8/x*)T is a tensor field. The gradient operator is often abbreviated 3/8x"=8,, so that the field Ss?"37 in the example above is written 9,3,-+° 37%.‘Folds and Transiormation Laws frame, giving @ function ings slightly differently. He 5. EF=AM,x’, of this point tly the two functions are 25,2223), In a more for a scalar field is written 34) T with cach event x is second rank tensor field. An x the components 7" of T™ (x"), Another observer “At,x’, the components 7 frame. Thus the functions frames are related by the Ke"). 32) ariant components, giving Tap"). (133) rank 1) is the-gradient of a * ficld one must check the for partial differentiation: ox") rAx*). ~*~ is a tensor field, then 2 tensor field. The gradient that the field 32°37 in the Problems 6 PROBLEMS: A. Show that €y¢ um = 5,Stn— BS Use the result of Problem 1 to show that A x (BX C)=(A+OB— (A-BIC, Prove that any rotationally invariant second rank tensor must equal Ad, for some number A. 4, Suppose that two coordinate systems x" and 3 are related by a linear transformation 3*= 4",x” and that €g,,X"=0 if and only if x4g,,x” =0, Show that A*,=AA*, where Ais a Constant and A, is @ Lorentz transformation matrix: AMA". = Bop '5. Suppose thai two coordinate systems x* and 3 are related by a transformation %+= F¥(x) which is twice continuously differentiable and invertable, at least in some open region containing x*=0. Suppose also that the transformation preserves light cones in the sense of (1.2.3), Show that the transformation has the form, att x . “TS tbexs be where ) is a constant, A*, is a Lorentz transformation matrix and a* and b" are fixed vectors, (1f A=], \*,—g",, and a” =0 this transfor rmation is called a “special conformal transformation”; if only a*=0 itis called simply a “conformal transformation.") Argue that the transformation relating two inertial coordinate systems must have b*=0 because otherwise (a) the transformation becomes singular along the surface 14+2bx+béx?=0 and (4) par- ‘icles which move with no acceleration in the x* system appear 10 be accelerated in the 3 system. 6. Prove that the only one dimensional representation of the proper orthochronous Lorentz group is the identity representation. That is, if (A) is a function of proper orthochronous Lorentz transformations ‘Af, such that MA)MA,)=MA,A,), then MA)= I. Show also that the conly one dimensional representations of the full Lorentz group are formed by combining MA)=1 for a proper orthochronous A. with MAp)= +1 or —L and A(A;)=+ 1 or =I. : \ow that every Lorentz transformation matrix A satisfies [det |= I and [A°|> 8. Show that every Lorentz transformation matrix AA, can be written in the form A=AyAgA,Ay where Ay and Ag. are‘rotations, A, is a boost in the z-direction, and A, is one of 1, Ap, Ay, ApA,. 9. Show that a fourth rank three-tensor 7“ that is invariant underFlelds and Transformation Laws ations must have the form T/=a0,5,++ fiydy-+ v8,d,. (Hint: Using the rules for combining angular momenta in quantum mechanics, show that the tensor product of four d= irreducible representations of the rotation group contains the J=0 representation exactly three times.) Suppose that TP isa tensor that is invariant under proper ortho- ehronous Lorentz transformations. Show that has the form. 7™° magtgh fgtg™ + yelett+ set, The Prine Stationan In the chapters that follow we principle of stationary action. it applies to the mechanics off generally, to systems with a & chapter we outline very and mention some of ils “F= ma” formulation of show how a formulation of ation can arise as a limiting with a finite number of 2.1 LAGRANGIAN Consider a system consisting: positions of the particles xy. the kinetic energy of the Th, Assuming that the motion of the particles can be deci ‘The form of V will depend.(CHAPTER 2 The Principle of Stationary Action In the chapters that follow we formulate various field theories using the principle of stationary action. Most readers will have seen this principle as 1 applies to the mechanics of a finite number of point particles or, more ‘generally, to systems with a finite number of degrees of freedom. In this ‘chapter we outline very briefly how the method works in this familiar case and mention some of its advantages and disadvantages as compared to the “P= ma" formulation of mechanics. Then we use a simple example to show how s formulation of a field theory via the principle of stationary action can arise as a limiting case from a similar formulation for a system with a finite number of degrees of freedom. 2.1 LAGRANGIAN MECHANICS OF POINT PARTICLES Consider a system consisting of V point particles that lie on a line, Call the positions of the particles xy, Xay-vwcy and the masses my, Myy.yMy. Th the kinetic energy of the partiles is ® Tk qerky= D imee Qu Assuming that the motion of the particles conserves energy, the forces on the particles can be derived from a potentiil energy fanction Vy) 2.1.2) ‘The form of V will depend on the physical situation at hand, and is left ”a 6 ‘The Principle of Stetionary Action arbitrary here. The equations of motion for this system are F,= mi, or ar gre 2.13) ‘The equations of motion (2.1.3) are equivalent to a variational principle called the principle of stationary action or Hamilton’s principle. To state Hamilton's principle, one defines a funetion L, called the Lagrangian, which depends on the coordinates x, and the velocities §,: Linx) =7R)~ VR) @14) AAs the system moves from an initial configuration x,=a, at time 0 10 a final configeration x/= >, at time 7, it will follow a certain path x,= X,(), as determined by the equations of motion. However, there are many other paths x;(0) which the system might follow to get from 4, 10 bTo each suck path we associate a quantity called its action, a= [aray.nor 15) Why is the physical path X,(¢) different from all other paths? It is because the physical path is a stationary point of the action. That is, a small variation 8x,(i) = «é,(0) of the path away from the physical path produces a change 64 of the action whieh is zero to first order in To show that the equations of motion (2.1.3) are equivalent to the principle of stationary action just enunciated, consider a suitably differen tiable variation of the path, x,(0)=<&,(). Since both the physical path %,(2) and the varied path x,(1)= Xj()+ «&(¢) must go from a, t0 b, we rust require that &(0)=&(7)=0. By differentiating under the integral, we calculate the variation of the action: tnef"e 3 (0 FH.) so ELa.x0)) Since &(2) vanishes at the endpoints, we ean integrate by parts in the first term to obtain 8 bane fd soo e | (2.16)« Principle of Stallonary Action system are Fim mx, of 213) to a variational principle ilton’s principle. To state L, called the Lagrangian, ities: ) (2.14) tion x, =a, at time 0 to a a certain path x,=X,(), there are many other from a, to 6,. To each such ). (1.5) other paths? It is because action. That is, a small physical path produces 2 ler ine ) are equivalent to the sider a suitably diffesen- bboth the physical path must g0 from a, to b, we ing under the integral, we X(9) eo.x.0)| rate by parts in the first 2.1.6) Lagrangian Mechanics of Polat Particles 19 Itis clear from this expression that 84 will be zera for every variation §(1) if, and only if, the quantities in brackets are all identically zero: eae LON) Fe EO Q.7 ‘The derivation is now complete, for this is precisely the equation of motion @13) We have seen that the two equations, F= mi and 84 =0, are equivalent Is there any reason to believe that one formulation of mechanics is ether more corvenient or more fundamental than the other? ‘There is no defini- tive answer to this question, but itis worthwhile (0 mention a few points that may help explain the wide use of the principle of stationary action in advanced classical mechanics ‘One advantage of the variational formulation is that the properties of the system are compactly summarized in oné function, the Lagrangian L(x,3). Ths gives the action principle a certain elegance, and it aso leads to important computational advantages. For instance, itis etsy {0 make approximations in the Lagrangian before going to the equations of motion. Its also easy in a given problem to make a change of coordinates from x, to any different set »1=9,(xy.-.-%yi0. {tis often 4 much more tedious task to change coordinates in each equation of motion. This ease of ‘making coordinate changes makes the variational formulation ideal for problems involving constraints (for example, a roller coaster that is con strained to remain on the track); one uses coordinates y, in which the constraint condition C(x)=0 becomes yy =0. (See Problems 3 and 4.) ‘Two more fundamental properties separate the variational and approaches to mechanics, First, as we shal see later, there is a direct connection between invariances of the Lagrangien and constants of the motion. For instance, ifthe Lagrangian is invariant under rotations then angular momentum is conserved. Second, there is « close relation between the Lagrangian formulation of classical particle mechanics and quantum mechanics. (There is likewise a close relation between the Lagrangian formulation of classical field theory and quantum field theory. This rela- tion is explored in books on quantum field theory, so we will not pursue it here.) ‘Although the principle of stationary action is @ powerful. tool whenever it can be used, itdoes not apply to systems subject to frictional forces.* To discuss such systems one must return to an F=ma approach. Likewise we will ave to go beyond Hamilton's priniple in Chapter 13 when we discuss field theories that include dissipative processes like viscosity, heat flow, and the flow of electrie current through a resistor “This point is turther discussed in Section 9.72» ‘The Principle of Stationary Action 2.2. THE CONTINUUM LIMIT FOR MASS POINTS ON SPRINGS ‘Consider the following simple system with N degrees of freedom: N mass points are constrained to lie on a line; each mass is connected to the next fone by a spring; all of the masses and springs are identical, (See Figure 2.1.) The Lagrangian for this system is ba DEG DAKO 0) 24 coordinate of i* mass point ‘mass of each mass point, K-=spring constant of each spring, mnstretched length of each spring. We may imagine that this is « model for a crystal in a one dimensional world. If we are solid state physicists in this world, we will want to use the full Lagrangian (2.2.1) to extract all the information we can about the behavior ofthe crystal. The model isso simple that itis, in fact, possible to solve the equations of motion exactly without t00 much wouble. The equations of motion derived from the Lagrangian (2.2.1) are Mim (5,284.84) Mi,=K(x,—2,-a), 222) My = K (xy 1 Xy ta). ‘The reader will be able to verify that the general solution of this set of equations is ama j—f]txot or +S en cos(kyalJ~ eos nyt 4,) 23) Where xo) 0 Oy Si. Gyr eee AFE constants chosen to fit the initial encom) Figure 2.1 Mass points on springs. ‘The Continuum Limit for Mass conditions, and the wave m2 ‘This prediction of our model ‘experiment by measuring the: ‘waves with various frequencies, plotted in Figure 2.2. Figure 22. Dispersion ‘An enormous amount of i tained in Figure 22. But there Which most of this informa might want to know how much’ are subjected to a certain foros, frequency of a crystal of lengal is that the distances character much larger than the in a macroscopic model that cam ‘of the material. The m that we will Gn this case) contain only those parameters To obtain a macroscopic continuous index o which x(0 are replaced by a en(i—})a by (U—Yan= sum 3,)M(3,? representing Sdo}(M /aXax/ de), Similarly, can be approximated by Je ‘The Principe of Stationary Action POINTS ON SPRINGS N degrees of freedom: N mass ‘mass is connected to the next springs are identical. (See Figure Esa) 2.1) B for a crystal in a one dimensional this world, we will want to use the information we can about the ple that itis, in fact, possible to without too much trouble, The sgrangian (2.2.1) are 222) the general solution of this set of 3 J) os out +4) 023) yastants chosen to fit the initial ‘The Continuum Lim for Mass Points on Springs. a conditions, and the wave numbers k,, and frequencies «,, are =m ky Fa? 08 =2£[1 -cos(tpa))- (224) ‘This prediction of our model for the crystal could be compared with ‘experiment by measuring the speed of sound, w/&, in the erystal for sound waves with various frequencies, and comparing with the theoretical curve plotted in Figure 2.2. ¥ Figure 2.2. Dispersion cuve for mass pints on sprint. ‘An enornioti’ amount of information about the model erystal is con- tained in Figure 2.2. But there is a wide class of physical problems for ‘which most of this information is irrelevant. In our simple example, we might want to know how much the length of the crystal changes if the ends are subjected to a certain force, or we might ask for the lowest vibration frequency of a erystal of length L. The common feature of these problems is that the distances characteristic of the phenomena being investigated are much larger than the interatomic spacing a. It is desirable to have at hand a macroscopic model that ean account for this “long wavelength” behavior of the material. The macroscopic model will contain certain parameters that we will (in this case) derive from the microscopic model, but it will contain only those parameters that are relevant ta macroscopic problems. ‘To obisin a macroscopic model we replace the discrete index i by a continuous index o which varies over the range 0%)- IF 2 does depend directly on x4, the equations of motion still retain the form IEC. 9,0404) 92 (Om 9s 6 i a(e0) with the same derivation as before. + 32 CONSERVED CURRENTS Let us take a closer look at the model introduced in (3.1.6), the scalar field with an external source $(x). Consider the quantities &* (x) defined by 5%x) SHa)= 0194) §[(@09)?+ (30)(39) + m767] +45, (B21) (Recall that, by convention, the repeated Latin indices in (8,9 a9) are to be summed over j= 1,2,3.) As we will see shortly, it is appropriate to call 5° the energy density in the model. If we perform the exercise of computing §, 5" = 3)6°+ 3,5/ we find 8,5" =(896)[(— 3,9" + m}6+ S] +9(205). (32.2) When ¢ obeys its equation of motion (3.1.7) the first term vanishes, leaving a,6%=9(38). Suppose now that the external source $(x) is independent of the time 2°. Then ‘ (323) ‘The implication of this is that the “energy” defined by Em f dxE%,x) (G24)e 20 ‘Some General Festures of Classical Feld Theory is conserved. Indeed, use of 8, =0 and Gauss’ theorem gives = faa =o, 25) provided only that $(x,1)alls off rapidly as fx} becomes large. To recapitulate, when the external source S(x) is constant in time, then the energy E defined by (3.2.1) and (3.24) is a constant of the motion Since the “energy” is the integral over space of 5%), its reasonable to interpret 6° as the density of energy. The quantities 6/ can then be interpreted as the energy current; that is & +dA is the amount of energy crossing an infinitesimal surface area dA per unit time. If the equation 8,5" =0 is integrated over a small volume V bounded by a surface OV, it reads 4/ sm--[ §- Sf Stax= [,, 8 24a. 3.2.6) Thus, 0,6" =0 is a focal statement of energy conservation: the rate of change Of the amount of energy contained in the volume ¥ equals the net rate of energy flow into the volume through its surface aV. ‘Apparently this is an example of a general phenomenon. Whenever we can construct out of the basic fields of a theory a set of fields J%(x), J'(x), J), P(A) which obey 2,1" =0, then the quantity Q¢0)= f dxJ°(x) is a constant of the motion, and J° can be interpreted as the local density of “Q." Another example is Q=electric charge, J°=charge density, J= clectrie current ‘Transformation Laws The discussion above applies in one particular reference frame, and applies equally if the quantities J* are the components of a vector field, or some particular components of a tensor field (as in J*= 7"), or have no particular transformation properties at all. However, the transformation law of J* determines the transformation law for the associated conserved quantity Q, as we will now see. Suppose, to begin with, that the quantities J*(x) are the components of a vector field, and that @,/"=0. We will consider the “charge” Q caleu- lated in a particular coordinate system o- feo) ear) Conserved Currents Let us analyze this simple i fashion. The integral is a space-time, We have used the surface, but any other set a, a, a o- fray The Jacobean 9(x)/@ (a) cam be’ ae) 7a) oa Now the integral takes on a where dS, = — Sa Ik should be apparent that ome any three dimensional surface &. form x*=x"(a'aa*). The crthoggnal to the surface at T= Ayaxt/aa!. The reader 1 integral like (3.2.8) is ind parameterize the surface. Since dS, is inanifestly a reference frame used to original reference frame still which we integrated, x"=0. In have to show that we get the ‘The original surface was xy the time axis of the originalI Features of Classical Fleld Theory Gauss theorem gives ~ f 050. 625) as |x| becomes large. sce S(x) is constant in time, then ) is a constant of the motion. space of 6%x), itis reasonable to The quantities &/ can then be is 6 +dA is the amount of energy JA per unit time. If the equation 1e V bounded by a surface AV, it Baa. 626) lay energy conservation: the rate of in the volume V equals the net igh its surface @V. eral phenomenon. Whenever we theory a set of fields J%2), J"(x), the quantity (= f dxJ%x) is a interpreted as the deal density of charge, J= charge density, cular reference frame, and applies ents of a vector field, or some Ad (as in J*= 7"), or have no all. However, the transformation law for the associated conserved ties J*(x) are the components of consider the “charge” Q caleue (0, G27) Conserved Current 3 Let us analyze this simple integral in what may at first seem a perverse fashion. The integral is a surface integral over the surface x°=0 in space-time. We have used the coordinates x', x°, x? to parameterize the surface, but any other set a!, a, a? would do just as well: 292 s0¢¢8(a!,a,0 O= frase? (x*(al.a?,a%)), ‘The Jacobean 4(x)/@ (a) can be written as 3) gee 284] 3) ae aa) [| ax! de! at meg 3a? da? Ax’ ax" ax® Tot Da) Qa? da ‘Now the integral takes on a covariant looking form, o= fas 628) where ax” ax* anf aS,=—-<6, So da'da*da’, (3.2.9) Sone Gar ST pet 629) Ik should be apparent that one can define a surface integral like (32.8) for any three dimensional surface 5 given parametrically by equations of the form x*=x*(a',a?,a%). The surface area differential dS, is a vector orthogonal to the surfave at a, since T#dS,=0 for any tangent vector T= YA de"/Aa. The reader can verify that the value of @ surface integral like (3.28) is independent of what coordinates a'ga® are Used t0 parameterize the surface. Since aS, is manifestly a vector, the “charge” Q is independeat of the reference ffame used to calculate the integral. However, the ghost of the original reference frame sill lives on in the choice of the surface over which we integrated, x°=0, In order to show that Q is really a scalar, we have to show that we get the same result no matter what surface we use. ‘The original surface was x%,=0, where c*=(1,0.0,0) is a vector along the time axis of the original coordinate system. Another observer will use2 ‘Some General Features of Classica! Fold Theory the surface x, =0, where h is a vector along his time axis. We want to show that the corresponding charges, om fas 6210) om fo" ean are equal. We will show that O.—Q, by using 4,/*=0 and Gauss’ theorem. To this end we consider the four dimensional volume bounded by the surfaces xe=0, x-b=0, and x'=R? (see Figure 3.1). This volume consists of two disjoint regions, labeled I and I]. We will apply Gauss’ theorem to the integral o- fateay In general, Gauss’ theorem tells us that Satsasn G21 [atragern f asyer e213) tay’ where a5, is the outward pointing normal vector to the surface 3V-which bounds F, (Note that this four-dimensional Gauss’ thorem is proved in exactly the same way as the corresponding theorem in three dimensional Euclidean space; the metric tensor does not enter the proof at all.) In this particular case we have : of jason fase [ase Now let the radius R become infinite. As long as J* fails off sufficiently sapidly at large distances, we obtain Laser aser* 0214) as desired, We have seen thut the conserved charge @ associated with 2 conserved vector current J* is scalar, By making very slight modifications in the proof, the reader can verify the generalization of this theorem: if J" "*(x) isan N rank tensor field which satisfies the conservation equation ase =0, 8.2.15) Conservation of Momentum amd Figure 3.1. Surfaces used for of the normal vectors dS, chown plotted the covariant components {ravariant components dS" would then the corresponding is a tensor of rank N—1. In the next section we tensor of rank 1, the total our present results, we momentum results from the a,7*=0. We will find that 33. CONSERVATION OF In the example of a scalar when the source (x?) is quantity E, constructed from simple and important Consider a theory of@ Features of Classical Field Theory along his time axis. We want to J" (3.2.10) J" G21 by using @J*=0 and Gauss’ dimensional Volume bounded by ? (see Figure 3.1). This volume Land 11. We will apply Gauss’ farsoe 212) ase", 82.1 [,,“. 62.3) vector to the surface 3¥ which al Gauss’ theorem is proved in ing, theorem in three dimensional not enter the proof at all) In this as,i"4 f dsm ie long as J* falls off sufficiently as,J" 3.2.14) @ associated with a conserved very slight modifications in the tion of this theorem: if J" (x) the conservation equation =0, (3.215) Conservation of Momentum and Energy a Figure 3.1 Surfaces used for proving that Q) = Q,. The smal arrows represent the direction ‘of the normal vectors 4S, chosen for the surface integrals. To avoid confusion, we have Plotted the covariant components AS, of the norral vectors; a igure showing the con teavariant components dS would look quite different. then the corresponding conserved charge orm fdas ,x) (32.16) is a tensor of rank N~1. In the next section we discuss an important example of a conserved tensor of rank 1, the total four-momentum of the system P*. According to four, present results, we may expect that the conservation of four- ‘momentum results from the existence of a tensor field, T*(x) which obeys 3,7* 0. We will find that this is indeed the case. 33. CONSERVATION OF MOMENTUM AND ENERGY . In the example of a sealar field with an external source we found that when the source 5(x*) is independent of the time x°, then a certain quantity £, constructed from the fields, is conserved. Here we derive a simple and important generalization of this result. Consider a theory of field components (x), O09 ...094(0). Theu Some General Features of Clssieal Field Theory Lagrangian density ©(q.9,%4.2) can depend on the fields and their derivatives, and also can depend explicitly on x if we include an external source. Suppose, however, that © does not depend explicitly on one of xxt.32, oF x say 2°: (Px: BP) Oe - Of course © still depends implicitly on this coordinate x* via its depen- dence on 62x) and 0,6(%): a, . 8,80 32) As long as the fields obey the equations of motion, the quantities in square brackets are identically zero, so that Dt=0. 33) Tus if the Lagrangian i indopendent of the eoordinat tbe quan P= fares) 036) is conserved. ‘The quantity P°=~ P, which is conserved when £ is independent of time is called the energy of the system. The quantity P'=P, which is conserved when © is independent of x" is called the one-component of the momentur of the system, and similarly for P? and P®. There is apparently fa close connection between these ideas in classical field theory, the Conservation of Momenturn and corresponding theorem in the the theorem in quantum Hamiltonian is independent of Hamiltonian does not depend system, On a less abstract and momentum on the ot as calculated from (3.3.3) and: basis of the identification P+ slowly moving systems. We did not make use of Chapter 7 we use the cor here in discussing nonrelat independent of x and is a be a tensor. One can give a illuminating to see how it the last section, when 7,* —the “four-momentum” of particle, PROBLEMS 1. Calculate 7," for the free: the equation of motion: 2 Consider the Lagrangian ‘ where » is @ parameter. wave equation for Lagrangian Consider the Lagrangiam: independent of time, the be conserved. Argue that souree is 3,64 (2). Find a source function § (x)= power produced by the (Pred (Note that the instant infinite,)« Featuras of Classical Fleld Theory depend on the fields and their ily on x if we include an external not depend explicitly on one of *) 0, 331) this coordinate x* via its depen- duet 5 ota Baagy Bite: G32) te for @ conserved current ae rt 3 9T he) 633) (2.3,00) ae DS Tore) } J | : (34) of motion, the quantities in square 0. 335) of the coordinate x*, the quantity TM) G36) ved when & is independent of 1. The quantity P'= P, which is is called the one-component of the for P? and P?, There is apparently is in classical field theory, the Conservation of Momentum and Energy % corresponding theorem in the classical mechanics of point particles, and the theorem in quantum mechanics that energy i conserved when the Hamiltonian is independent of time and momentum is conserved when the Hamiltonian does not depend on the position of the center of mass of the system. On a Tess abstract level, we base the identification of P, as energy and momentum on the observation to be made in the next chapter that P, as caleulated from (3.3.3) and (3.3.6) is just what one would expect on the basis of the identification P= Mv and E={Mv?+ (internal energy) for slowly moving systems We did not make use of Lorentz invariance in this section. Indeed, in Chapter 7 we use the conserved energy and momentum we have found here in discussing nonrelativistic systems. However, if the Lagrangian is independent of x and is a Lorentz scalar then 7,"(x) defined by (3.3.3) wil be a tensor. One can give a simple formal proof of this; but it will be more illuminating to see how it works in examples. According to the results of the last section, when 7," is a tensor the quantities P* form a four-vector —the “four-momentum” of the system—just as they do for a point particle PROBLEMS 1. Calculate 7,* for the free scalar field, (3.1.2), and verify directly from the equation of motion (3.1.5) that 3,7, 2 Consider the Lagrangian for 2 sealar field 9: P= 4] (@oe)*—0* Z (ae)}- where o is a parameter, Show that the equation of motion for 6 is the wave equation for waves with velocity v. Calculate Z,* for this Lagrangian. \Consider the Lagrangian (3.1.6). If this external source S(x) is not ‘independent of time, the energy defined in (3.2.1) and (3.2.4) will not be conserved. Argue that the power per unit volume produced by the source is 8,5* (x). Find the outgoing wave solution g(x) prodaced by 4 source function $ (x)= Sqcosw!8(x) and calculate the time averaged ower produced by the souree in two ways: cmp ah B=0 as long as detM>0, Thus R is indeed a rotation matrix. In summary, Gyy contains precisely that information about the local displacement of the material which is not affected by a final rotation If we imagine that the material is made of “atoms” connected by “springs,” the matrix G tells us how much the springs have been stretched. The distance between an atom with coordinates R, and a neighboring atom with coordinates R,+4R, is determined direetly in terms of the inverse G~" of G: =(G"") ROR, (6.15) For this reason, G is often called the deformation matrix. (The reader should be warned that several closely related matrices go under the name ‘of “deformation matrix” or “deformation tensor” in the literature. The present matrix G was introduced by G. Piola* at the early date of 1833, Dut its inverse, called Green's deformation tensor, is more popular now. Readers familiar with Riemanian geometry will recognize that the matrix elements Gy are the contravariant components of the metric tensor in the material coordinate system.) ‘The Deformation Matrix and the Density ‘We can connect this discussion of the deformation matrix with our previous discussion of fluids by noting that the density of the material in a local rest frame, p= V—J,J* is simply related to the determinant of G. In OR, AR, oof) p= nVaiG « (5.16) Since both sides of this equation are scalars, the equality holds in any 5G. Pols, 1833, La mecounce dori nturalamente ene tatata col caleoo dele earasionis Opuse a. fis dover tory, Vol. 1 (Giusy, Milano, 1833). pp. 201-236, ‘The Lagrant frame. Thus the fluid discussed imt the class of materials discussed and R,, we recover the case of Some Examples We can get a better feeling for the} the calculation of G for some ais solid, 1. Uniform compression: 2. Squeeze along R, axis with‘The Mechanics of Elastic Solids VG 'GvG as long as detM>0. Thas Ris t information about the local affected by a final rotation, le of “atoms” connected by the springs have been stretched. rdinates R, and a neighboring fined directly in terms of the 6.15) deformation matrix. (The reader ited matrices go under the name tensor” in the literature. The Piola® at the early date of 1833, Sn tensor, is more popular now. will recognize that the matrix ents of the metric tensor in the deformation matrix with our tthe density of the material in a related (o the determinant of G.In aR, aR, let ax! ax! C (6.16) scalars, the equality holds in any ‘The Lagranglan ond the Deformation Matix 50 frame. Thus the fluid discussed in the preceding chapter is a special case of the class of materials discussed here: if £(Gy,,) depends only on detG and R,, we recover the case of the fluid. ‘Some Examples ‘We can get a better feeling for the deformation matrix by working through the calculation of G for some simple deformations of a two dimensional solid 1. Uniform compression: Og eG lehey e-(5~ 2.) “2, Squeeze along R, axis with no compression: acy The Mechanice of Elastic Solids 3. Squeeze along axis inclined at an angle ©, with no compression: Pee anea Bee (e)-(e 25s" Sey ales) eee 4. Shear along the R, axis: (2)-(6 93) c(t 9). cot Note that det G=1 in Example 4, indicating that the density of the material is not changed when itis sheared, The strain matrix produced by this shear can also be produced by a properly chosen compressionless squeeze as in Example 3. (To be precise, the proper choice is cot2#= sinh =«/2), The functions R,(x) are different in these two cases, even though the deformation matrices are the same. In Example 3, the matrix relating R and x is symmetric; itis the square root of G, The matrix relating R and. x in Example 4is related to VG by a final rotation, ‘The Momentum Tensor 52. THE MOMENTUM TENSOR: We come now to the dynamic of equations of motion cor ci aR, (Here and in the following, we Symmetric matrix G is written £/0G,) {As we found when we dis physies of energy-momentum physies of the Euler-Lagrange fum conservation equations @ equivalent t0 the three Euler Same as that given in Section 4 on the energy-momentum tensor The expression we oblain for the form (Gay Rad! whore p= nVEKG =\=J,J" i forward matter to calculate 7” @,R)9 2 /3(,R,) (Gee Section Tm pUuha + where u*=*/p is the four- In a local rest frame of the ‘can be identified as the internal ‘or momentum density in the local ‘One can think of this as a material coordinate system, “1m books on casticity theory oF ‘8%, which is identical to the momentama ‘The Mechanica of Elalic Solids 8, with no compression: 2 Miers )(e) 120 sin 26 ing that the density of the |. The strain matrix produced by roperly chosen compressioniess proper choice is cot20=sinhe in these wo cases, even though Example 3, the matrix relating R G. The matrix relating R and x rotation. ‘The Momentum Tensor ot 52. THE MOMENTUM TENSOR We come now to the dynamics of the elastic material. The Euler-Lagrange equations of motion corresponding to a Lagrangian of the form &(G,4.2,) OR. -n] azoa]-0 ean (Here and in the following, we assume that the dependence of 2 on the symmetric matrix G is written in a symmetric fashion, so that 92 /06,, =8£/ 86.) ‘As we found when we discussed the special case of fluid dynamics, the physics of energy-momentum conservation is more transparent than the physics of the Euler-Lagrange equations. Furthermore, the three momen- tum conservation equations 87!=0 in any fixed reference frame are ‘equivalent to the three Euler-Lagrange equations (5.2.1). (The proof is the same as that given in Section 4.5.) Therefore we concentrate our attention ‘on the energy-momentum tensor T*. ‘The expression we obtain for 7 is easiest to interpret if we write © in the form (Gps R,)= — PU (Gays By Je (622) where px nVEEE =A in the cet fame density tin a stright forward mater to calculate T™ from the general forma GRE TIGL) Gee Sern 43) One Find Pee ptere29a%R,y9"%,) BL (523 where "= J*/p is the four-velocity of the material In a local rest frame of the material the energy density is pU. Thus U can be identified as the internal enerey per atom. There is no energy flow ‘or momentum density in the local rest frame. The momentum flow is* Pe aZanyxaa) itime, —624 One can think of this as a momentum flow tensor Fig =2p0U//AGy, in the material coordinate system, transformed to the laboratory coordinates by books on elasticity heory ar continuum mechanis, ne encounters the “ress tensor” bic i deatcal to dhe suomentam flow tensor 7 in material at rest except for signa ‘The Mechanics of Elastic Solids ‘means of the matrix R,/@x‘, but this interpretation need be only sugges tive. What is important is that the stress in the material is proportional to the rate of variation of the internal energy as the deformation Gy is changed. As an example, we recall that the pressure ina fluid is — aU/8:v Energy Conservation The fact that the rest frame momentum flow tensor is proportional to 8U/ AG,» is directly related to conservation of energy. Suppose that T™ had the form T= 9U (Gaye Ry H+ (DR, MAR, ey. (625) but that T., were unknown, We might then consider the equation 4,0,7"" =0, whieh expresses energy conservation in local rest frame. Using the form (6.2.5) and the kinematic results @,(02")=0, 1,3, and 1,3°R,=0, this equation reads Om = pu", U+u*(9,2,R, Ry Tas ~ pu’, U+ 3(W?,Guy Tay (52.6) Using the chain rule to differentiate U, and remembering that w°2,, we would have au, 0- [samp Jee (622) ‘Thus we viould be led to identity Z, with 2p8U/@G,, on the basis of energy conservation. $3 BOUNDARY CONDITIONS ‘The action principle presented above and in the preceding chapter is applicable to materials that fil the whole of space. It is a simple matter'to, extend the formalism to cover the mechanics of finite material bodies. In this section we consider bodies that are free to float through space and bodies whose boundary surfaces are rigidly fixed or are constrained to slide over another surface. The relevant boundary conditions really need no éerivation, since they can be written down immediately on the basis of “common sense” or “physical intuition.” Nevertheless it is interesting to see how the correct boundary conditions arise from the principle of stationary action. Boundary Conditions Free Boundary Consider first a solid body that dimensional example in Sectiom the body is described by an defining the inside of the body. such a body consists of the usual interior of the body only: An fatxo(rt Here @(2) is the unit step fur ‘When we calculate the variation 4R,(x) of the fields, variation of the region of int integration by parts. These surface! using the relation a ah where (2) is the Dirac delta parts, b4=- f axony{ ~ f asec The first term leads to the interior of the body and need mot 8(F), the second term contains F(R,(x))=0 in space-time swept little algebra, it ean be written tage fe[aen