CHAPTER 3 Christoffel Symbols and Covariant Differentiation 3.1 Introduction Physical laws are often described by differential equations and differential ‘equations are concerned with the rate of change of physical quantities with respect to space and time, These physical quantities are vectors, tensors ete, In earlier chapters we have defined and discussed vectors and tensors and their transformation laws. This chapter includes the rate of change of vectors and. tensors and related results. 3.2 Christoffel Symbols From the metric tensor gis and its conjugate g', defined in the previous chapter, we can construct two functions. These functions are not tensors but are used to define the differentiation of tensors. They are known as Christoffel symbols and are defined as (3.1) and “ai ~ Bam (2) a) ‘The symbols given by Eqs. (3.1) and (3.2) are, respectively, called Christoffel symbols of first and second kind. Since gi; is symmetric, so are the Christoffel symbols, ie., Tag =Tags The = Thy (3) Now, replacing k by m in Eq, (3.1), we get 1 (29m, aim — 2oy Ho (Bet Gee ~ ae) 54Christoffel Symbols 35 ‘Multiply this equation by g*™ to get Lem (33 MP mis = sat (St gets = 39" oes 295m a) =r, (Ba) ‘ax! Bam Consider now Eq, (3.2), interchange i and m and mul equation by gim, to get iply the resulting Po%p _ Agix. der (3.5) Equations (3.4) and (3.5) are the relationships between the Christoffel symbols of fist and second kind. Remarks 1. A comparison of Eqs. (8.4) and (3.8) with Eqs. (2.25) and (2.26) suggest thatthe process of going from Christoffel syinbol of first kind to socond kind and vice versa is same as the process of raising and lowering of indies. 2. The Christoffel symbols of first and second kind are also denoted as Uesij] and {5}, respectively 3, It seems worthwhile to mention that when the generalizations of the Riemannian geometry are considered there appears a quantity, known fine connection; the Christoffel symbol of second kind is a particular case of such quantity. The affine connection Ty, given by Eq. (3.2) is exclusively valid in Riemannian geometry. Usually, in non-Riemannian geometry affie connections are non-symmetric in thelr lower indices. A sy 441) Indigent components. "Ths, fr four-dimensional Riemannian space, which is a space of interest in general theory of relativity, each of the Christof symbol hes 40. independent components. If the affine connection is not symmetric, we define the tensor metric connection defined in n-dimensional space has 1 Thay = 3a Thy) whichis skew symmetric inj and &, This tensor is called the torsion of the spacetime ‘The Christoffel symbols of first and second kind have the following properties @ (3.6)36 Christoffel Symbols and Covariant Differentiation Proof Use definition (3.1) and the symmetry of gy o a 1 aya o Te gar 8VI~ pee (37) Proof Differentiate the det g = |gij| and use u1 _ cofactor of gjt in (det gj1) # detgy: togst 0 0 oo ot =a9!(Cijx Tox) from Eq. (3.6) = oC}, +0) = 200, hich leads to Me Note: For g to be negative, Eq. (3.7) reduces to 1, = 2 log V=9 = (38) a= gor lee oh ag 4, oT, ay Joe PT — Phy (a) Proof We know that ana" Differentiate it with respect to 2”, we get 29" D056 1, forl=j i OE + gh OE — 0 a5 6! = Sam TT aan 77 Lo, fortes Multiply this equation by 9? and use property (i), to get ao! og BS + ata (Caim +E jam) = 0‘Transformation Laws for Christoffel Symbols 87 Since gi is symmetric, therefore ant 20 PD Eaym + IPTG =O “ age ya For 9m ~ Thy Now replace k by j and I by q to get required result. 3.3 Transformation Laws for Christoffel Symbols We shall now find the transformation laws for the Christoffel symbols and show that they are not tensors. Since gij is a second rank covariant tensor, thus under coordinate transformation, it transforms as ax Ax) a ano Differentiate it with respect to a’? and use Ags _ Og Ox™ Ox? ~ 8x" Ox'p a to get Oy 0 (Ax x4 an? ~ On oat xk) 4 [pert 2 a Of ~ LaePan on * Bat Barron) * Out Bar® Oa” Bx? Interchanging the dummy suffix i and j inthe first term of the square bracket and using the symmetry property of gi, we get oe =| dat det | _ dat os 4, 28 Ba! 2” 265 (a9) az” ~ avant ax * oxox ax"| 9 * ax" ax Ox Ox™ By cyclic permutation of indices i,j,m and I, k,p we get two more similar equations which on suitable change of dummy indices can be written as 8p _{ at Oat, Rat Das) | Dat Dat Ds” Dom Ox" Bx" Ox® Ox” * Bx" x? Bx oO} Bx" Bx! Dx? Gxt (31) and 2 dat do) tat Aad) |, da! Bx) 929m (5 9) de® ~ |Beox? a0" * Sean Be? | 4 * Ge dF De® Ox38 Christoffel Symbols and Covariant Differentiation Now subtract Eq. (8.10) from the sum of Eqs. (3.11) and (3.12) and divide the resulting equation throughout by 2 and use Eq, (3.1) to get Ox! Ox Ax” x ae Dip = oe oe oy oat 219) E> Dall ak BaP * Healy OPM G13) Equation (3.13) is the transformation law for the Christoffel symbol of first kind. ‘To get the transformation law for the Christoffel symbol of second kind, we us the transformation law of the contravariant tensor of rank two as oq _ 22? Bx"? oy oY Te Oe (G4) Now multiply both sides of Eq. (3.13) by the corresponding side of Eq. (3.14), and after simplification, we get Ox' Ard x! Pat Ox't T= =i, > (3.15) 8 gat an O° * ReTae® Bat G18) which is the transformation law for the Christoffel symbol of second kind. From Eqs. (3.13) and (3.15) it may be noted that Ty, and Tif, respectively, Dbehave like rank three tensors as far as the first termis in righthand sides of these equations are concerned, but. the second terms of Eqs. (3.13) and (3.15) introduce the inhomogeneous terms in the transformation laws, Hence Ty tnd T%f do not transform ike tensors. However, in the particular case when ae"oa* ‘quantities I, and Tif in this particular case behave like the components of a tensor, Such tensors are called pseudo tensors. the coordinate transformations are linear, the term, = 0 and thus the 3.4 Equation of a Geodesic What do we mean by a straight line in Euclidean space? One meaning implied by the adjective “straight” is that its direction remains unchanged as ‘we move along it, The other property associated with the straight line is ‘that it represents the path of the shortest distance between any two given points. Here we shall find out what curves are implied by the later definition in a more general space, ie., the Riemannian space. The implication of the former definition shall be considered in Section 3.11. ‘The path (world line) of a particle between two points P and Q is a ‘geodesic and is determined by the condition that the interval between two points P and @ given by J ds be stationary. In other word, geodesic is lp defined by the condition that fra-o (16)Equation of a Geodesic 59 where 5 denotes the variation from the actual path (world line) between the two points P and Q on it, to any other path in the neighbourhood of this line (actual path). Equation (3.16) is the analogue of the condition of shortest distance between two points in Buclidean space ‘To obtain the equation of geodesic, consider the metric of the Riemannian space, i.e, ds? = gijdetde! (a7) and take the é-variation of both sides of Eq. (3.17) so that 5(ds*) 6(gisde'da) leads to 2dsb(ds) = dade bqys + gyda? Sl da) + gigdn'6( da?) = dada 28H 50 + gy 6(dat) + gudz'6ld)) (8.18) where dz* denotes an arbitrary infinitesimal displacement from the actual path. Divide now both si ‘ (=) Loe) (3.19) 1 gy 4600) te Ba + oF a (82) + 95°F Ga (62)| da (8.20) from which condition (3.16) becomes da dod O03 5 y 9 Wd e5rty g 9 ttt _ jee 12908 gh + gy M2 8 bn!) + gy 4 (60*)| do = 0 Changing the dummy suffices in the last two terms, we get da! de! 80 ed [[zewe + (aS +o =) eset) a Consider the second term of the above equation and integrate it by parts, remembering that the é-variation at the end (fixed) points P and Q is zero, we have 1 f fda ded B95 d (dod ast 1 fds de! doy a2) ahd = af era a (ig tou) wee Since the variation d2* is arbitrary, therefore, for the integral to be stationary the coefficient of é2* in the square bracket in the above equation must vanish60 Christoffel Symbols and Covariant Differentiation at all points of the path, i.e, 3 [del dot Bony ds ds dx which may be written as Ide! det Og; 1 dgie dot Bas de Oct Bde ds But dow _ Bain da? dans _ Bony dt ds ~ Oat de’ “ds ~ Oz do G2) Equation (8.22) thus takes the form Multiply this equation by 9, to get Lado! do yy [Bary Oove _ Bons) _ ge Px? 2ds ds Oak Oxt P ds? fet le 38 + Sas Baus] det det as Oa' | ds ds which, from the definition of Christoffel symbol of second kind, takes the form Beaty de ded wee (3.24) @ ‘This is the required condition forthe given integral "ds to be stationary Ip ‘The “straight line” given by Bq. (8.24) is the equation of geodesic [see also Eq. (3:248)]. Remarks 1, For ¢ = 1,2,8,4 Bq. (8.24) gives four equations which determine the geodesic. " Now, since ds is an clement of the world line, we can interpret these equations as the equations of motion of a particle which ‘moves along a world line. When the components of the metrie tensor are constant, then I‘, is zero and Eq. (3.24) in this case reduces to et ds? (8.25) which shows that the particle moves with a uniform velocity in a straight line,Affine Parameter a 2. The occurrance of the Christoffel symbols in mechanics and its relationship with the equation of geodesic is illustrated as follows: Consider evolution of time of a mechanical system for which, in dz! Sip the kinetic energy generalized coordinates 2(t), the velocit ov. 2 oti and F, is the generalized force obtainable from is T = joud'e* and Fy = — 55 is the generalized force obtainable fr ‘a generalized potential V. In analytical dynamics, the metric is Tat? = ds. The Lagrange’s equations of motion are a (aL) aL at Oe) ~ Ont where L = T — V is the Lagrangian, These Lagrange’s equations lead to ppd foam , Poe Pane) sige aa +3 Laat * Bak Oat a ‘Multiply this equation by g*, to get 4 Tigatat = FF (3.26) which clearly indicates the appearance of the Christoffel symbol in the Lagrange equations of motion of a particle. If there are no forces, F; = 0, Eq, (3.26) is just Eq. (3.24) and defines geodesic lines. But in Eq. (3.26), the differentiation is performed with respect to a time parameter ¢ along the trajectory while in Eq. (3.24) the differentiation is carried with respect to the arc length parameter # along the geodesics. Since both # and ¢ are normal parameters, we have ds oc dt. It may be noted that this proportionality holds only on the geodesic lines: along other curves the notion of time of transit and are length are different (see also Remark 2 of Section 3.11). 3.5 Affine Parameter The equation of a geodesic given by Eq. (3.24) has been derived in terms of the infinitesimal distance ds along the geodesic lines. This equation can also bo obtained in terms of the other parameters. If we choose the parameter a, then dzt de‘ da =-EE (3.27) w " 4@)-a@8)-2¢8 fat (3.28) da?62 Christoffel Symbols and Covariant Differentiation From Eqs. (3.27) and (3.28), Eq. (3.24) takes the form Pat | pu dat dot __ Pafds? dot dat Ta da ~ (dafao dae (8.29) which is true for an arbitrary parameter a. If we choose the right-hand side of Eq, (3.29) to zero, then Bai | de dat Tete (3.30) ‘This geodesic equation is same as that of Bq. (3:24) except the parameter a replacing the distance parameter s. It may be noted that the right-hand side of Eq, (3.29) is zero only when @a fag which has the solution aaah (3.31) ‘where a and b are two arbitrary real constants, Equation (8.31) shows that the two parameters @ and s are related to each other through a linear ‘transformation ‘A parameter of the geodesic line (here it is a) due to which the geodesic equation retains the standard form of Eq. (3.80) is called an affine parameter. Equation (3.31) shows that any other parameter, say 2, will also be related. to s through Eq. (8:31) and consequently all affine parameters are related to each other by a linear transformation (see also Seetion 3.1). 3.6 Geodesic Coordinate System ‘We have seen earlier that in Euclidean space, when Cartesian coordinates are considered, the components of the metric tensor are constant and consequently the Christoffel symbols vanish at every point of the Euclidean space, But no such coordinates exist in case of Riemannian space. If this is so, then such a coordinate system is called geodesic coordinate system and we have ‘Theorem 3.1 It is always possible to choose a coordinate system in which, all the components of the Christoffel symbols vanish at a given point. Proof Consider a point P in the coordinate system 2 and suppose that the Christoffel symbols do not vanish at P. Now introduce coordinate system 2 by means of the transformation = (ep + Uj ele’ — Cellet —(e")p] (832) where the subscript P denotes the value at the given point P and (2!)p = 0.Geodesic Coordinate System 63 We now calculate, using Eq. (8.15), the Christoffel symbol in the new coordinate system. Differentiating Eq, (3.32) with respect to z!, we get =. FS + Oe ele" — Chel + der Sarl — (2)e] = Fe Leh ndflet — Che] + SC edt le’ — el ant 7 =a+ Shela = (e)p] + MO ele — (2)e) 4+ Cielo? — el + FC ele! — ee] ~ Bat = oe + (Fh )ple! — (@)] (3.33) ie which means that the Jacobian J =| Thus (3.34) [40 and the transformation (3.32) is allowed in, the neighbourhood ia PP taking the ox! 3.93) with 27, we get Eq. (3.33) apr We 8 (32) (Be) ~ (Be) (see) + gener 1 product of = + i)rle! = rl sar (3.35) which leads to ox! . (3) pot (3.36) Now differentiating Eq. (3.35) with respect to 2” and using Eq. (3.36), we eet (gin), --tthor =64 Christoffel Symbols and Covariant Differentiation From Eqs. (3.34), (8.36) and (3.37), Eq. (3.15) thus leads to we (5a), (ae), (Se), * Carat), (3), = HHLIUTS) — Hie) = (ide - Che =0 which completes the proof. Remarks 1. If the point P in Theorem 3.1 is taken as origin, then in a similar way it can be shown that the coordinate system defined by Baa + Lois! (338) is a geodesic coordinate system {here (2) p = 0] 2. The choice of the vanishing of the Christoffel symbols at a point of the spacetime (four-dimensional Riemannian space) has a fascinating and deep physical meaning explained as follows: IF =0 in Eq, (3.26), then #4 Tale <0 which means that the acceleration of the particle, moving in the gravitational field, is proportional to the Christoffel symbol. The vanishing of the Christoffel symbol at a point means that we can choose a coordinate system in which the acceleration of the particle at ‘a given point of the spacetime is zero, that is, the gravitational force ‘acting on the particle vanishes at this point, or, in other words, by ‘means of a coordinate transformation, the gravitational effects can be climinated. But this happens only at a point and not in a finite region of the spacetime, These considerations are related to the principle of equivalence ~ a postulate of general theory of relativity. 3. For other coordinate systems, the reader is referred to Section 6.3, EXAMPLE 3.1. Show that, Tuc +s +n = 3 (3p + Bk — See) Oxk © Ox! ~ Bx) Solution Use the definition of Christoffell symbol of first kind to write the left-hand side and simplify to get the result, EXAMPLE 8.2 Find the components of the Christoffel symbols for the metric ds? = r?d6? +r? sin? aad?Geodesic Coordinate System 65 (This metric represents the surface of a sphere in a two-dimensional Riemannain space.) Solution Here x! = 6,22 ro af ue 0 To ssinta ]) 7 [0 a/rtsinte “sin? @ det ais From Eq. (3.1), we have 1 (Aaj , Pans _ Oasn _ Ton 5 (Fe Get) ik = 1,2 which leads to Qo 5 Po _ Pon (Ge oe oe Agr2 r?sinO cos In a similar way, we have =r? sind cos Tyo =Pans = Pan = 0, Tae Also from the definition of Christoffel symbol of second kind, we have ty = oT nis which leads to Tia = 90 naa =o! Pa + o"Paaa —sin 0 cos Similarly, P}_ = cot 0; while the remaining Pi,66 Christoffel Symbols and Covariant Differentiation EXAMPLE 3.3 Obtain all the Christoffel symbols for the metric ds? = —A(da')? = B(dz*)? — C(de*)? + D(ae!)? where A, B,C and D are real positive quantities and functions of all the four coordinates 2! 2.2% and 2. Solution Here ou m5 fori Aj {D, 9 =0 fori xj det 9s From Eqs. (3.1) and (8.4), we have Thy = 9D nje ‘These equations lead to Th=9'Pnn =9'Ti +9" Pan +9 Tan toa =o"Pin (i) pn 2A Det Th = 92m P21 Agu 1 " (Gz) 1 0A Bada In similar manner, the remaining components of the Christoffel symbols can bbe calculated and all the components are given as follows: 104 4, 10a aa’ "2 aqom" M8 za aes ThGeodesic Coordinate System o7 1 0B 1 op 9-1 2B 3, a) show that the differential equation of the geodesic is ar\? o (ary? | 2,2 pape ot ($) tere where Kis a constant. Solution The given metric can be written as do? (341) so that, For i= 1, Eq. (3.24) yields ry de dot wtlhea m7 @r ay (ar)? ay (a +n ($) +(Z) @r 2 (dr\?__9 (do? a wala) ~" a) ~°Geodesic Coordinate System a while for i , Eq. (3.24) leads to #8 yyy ar dd tthe fo, 1) ar da ae 7s +) deds ‘Now separating the variables and integrating, we get where B is a constant, Thus, after simplification, from Eq. (8.41), we have o(8) veer where k? = (B® ~ 1)/B? is a constant. EXAMPLE 3.8 For the metric ds? (da? + dy? + dst) + dt? find the equations of geodesics. Solution For the given metric, the non-vanishing components of Py, are Th=Th =The=Th=Th=Th=* hy = Pha = Ph he For i= 1, Ea, (3.24) leads to ed dz dt a tae? which on integration yilds B nae Similarly, from Eq. (3.24), for i = 2,3,4, respectively, we have dy —2ke ao be dz _og-oe #. #t yaa ag ~ he a? +0? +e? and a,b,c are constants of integration.2 Christoffel Symbols and Covariant Differentiation 3.7 Covariant Differentiation covariant vector (a tensor of rank one). Now, if A is a vector, then what the nature ofits derivative 24°? Will it be a tensor oF not? ‘This and similar types of questions shall be addressed in this section. Consider a contravariant vector A' which transforms as oa! os Differentiating this equation partially with respect to x’*, we get aa" 0 (de yy Bx'* Ax'® \ Ort 8 (a0 y) oer Ox? \ xt ax! ae! aa? Oat | da! aa? Oa? ax® ae A 1 2 emer tn (2) sal eosin ott rm on and right-hand side, But due to presence of the second term on the rig! side of Eq. (3.42), the quantity a does not behave like a tensor, i.e., the outcome of the differentiation of a tensor is not a tensor; and in tensor analysis only those operations are allowed which when performed on a tensor lead to a tensor. Thus, the quantity 247 wil be a tensor only when the quantity SE vanishes, Le if the coordinates 2 are linear function of the coordinates 2! then Eq. (3.42) is the transformation law of a tensor. ‘The. process of obtaining tensors through the process of partial differentiation is known as covariant diferentiation and ls being developed as follows (a) Covariant derivatives of contravariant and covariant vectors ‘Wo know that [ef. Eq. (3.15)] 4 _ Oa Ox! da! ,, | Oat dx’? 1 aa! de® Ox" 9” Bx ax® Ox* eam or Oa! dai Ox’ Ox" Ox" Ort ‘Take its inner product with <7, to get Ont Ox! Ox" 22" Ha + ax! a" 9° Baron Ox Oa ae Bx Be"Covariant Differentiation 3 which gives Be’ Aa" 4 Oat Oat Detoa™ ~ BETH get ATS a Interchanging the x and 2’ coordinates in Eq. (3.43), we get Px oa!” ax!* Bx! Sit _ Bal yy _ oat ont Balas ~ Bea hk Bal Be Gay Substituting this value in Eq. (3.42), we have oA" dx’ Ox? AAI [Oa" 4 Ax! Ax!” ase Oe Da? OAI [Oat pg _ Del" Oe ys] D2 4s Be ~ a5 ant Ber * | Oat" ~ Bar O27 '*| Bee But O0 DA | 1g OE" OE 4y_ da" 2 ry Ox) Ax!® Ox?” ”) Ax4 Aa'® bx® Ox Oz!" a? OAT | Vg Dx" Bx? 45 Oa! Want pa ™ ~ gaint Since 2" and 2’* are independent, therefore, “=; = 6 and the above ep oF equation reduces to Batt _ Be BaP OA oy Ot Os ay 1 Ox* — Oxi Bx! Dx? © 4 Ox? Aal™ “ which leads to OA | pus ga. 28% Da? OA! | py Da" D0? 4g oak Noe Oni art Ox *'P Ort Dai = (22% 227) 24% Ox) Az! } dx? Ou Ox? [O45 4g ter ae [ae + oe) ‘ar Now introduce the notation ’ 44, = 24 sry at (3.46) so that Eq. (3.45) can be written as 1 _ 82! Bx? 5 bk Gal gah (347)ma Christoffel Symbols and Covariant Differentiation which shows that the quantity AJ, defined through Eq. (3.46) transforms like the components of a mixed tensor of rank two. This quantity Aj, is called the covariant derivative of a contravariant vector. We also know that the transformation law for a covariant vector Ap is siven by Differentiate this equat same manner as above, to obtain — 2Ai Lik = Bak Ay -T,4, ‘yep (3.48) which is known as the covariant derivative of a covariant vector. (b) Covariant derivatives of rank two tensors ax Consider a covariant tensor Aj; of rank two and lett! = F~ be a unit tangent vector along the geodesic, then aA Ce si at 4g pet Gyldte) = FA ee + Age + Aye Ags da a (ax! ad (des Pes doe 4 ay ($) B+ Aytd (4) OAs pig Gal 5 seat Felt + Ag Fee + Ash Pak ore 1 4 Aye 4 Aut! CUP + Ag Gee + Ant oar But from Eq, (3.24), we have Using this value and the definition of ¢* in the above equation, we get @ cae) — [240 ‘ Seago) = [248 — hay — “haa tor (3.49) In Eq. (3.49), the left-hand side is a scalar (a tensor of rank zero) and (t't/t?) is a rank three contravariant tensor. It therefore, fellows from the quotient rule that the quantity in the square bracket on the righthand side of Eq (3.49) is a tensor of rank three, We denote this quantity as Ay pe SAS De ay — Dh Ae (3.50) A Ox?Covariant Differentiation 5 ‘The covariant rank three tensor Ay.p is known as the covariant derivative of 4 covariant tensor of rank two. In a similar way, the covariant derivatives of contravariant and mized tensors of rank two are defined through the following equations: aa" y Ag = Fy te +A (3.51) and aay Aig = 52 + TAS TAAL (3.51a) (c) Covariant derivative of tensors of higher rank The process of covariant differentiation can be applied to tensors of higher ranks and in general, for a mixed tensor of rank (p-+q), the covariant derivative is defined as aati’ Miudeduik ~ — Oak Tra Ajgecse t Pra Ajigcia + WTA TAS TARE on (3.52) Remarks 1, The covariant differentiation is denoted by a semi-colon (;) while the partial differentiation is denoted by a comma (,). 2. It may be noted that the covariant derivative is an operator which reduces to partial derivative in flat space (where giy are constants) with Cartesian coordinates but transforms as a tensor on an arbitrary manifold. 3. For a vector A¥, the covariant derivative At, for each direetion j, will be given by the partial derivative opeartor =; plus a correction specitied by My 4, From Bags. (3.46), (3.48), (8.50) and (8.51), it may be noted thet through the process of covariant differentiation, we get tensors of higher rank Thus, the rank of a tensor ean be raised by differentiating it covariantly, while the rank of the tensor is lowered by the process of contraction. From Eq, (2.8), we have jazace Gs de = ayuw! def S, i# @ unit tangent vector. The equation of geodesic given by Bq. (3.24) can be expressed as d (de®) 4 dztde™ az (=) Tin Ge aa 9 Here w'76 3. Christoffel Symbols and Covariant Differentiation dk ye tym — sett + Thautu™ = 0 dué — dub de! uk But So = FES = Fou, hence, the above equation can now be written as Out pe ym) af (= + Thaw’ ) =0 which, using the definition of covariant derivative of a contravariant vector [ef., Bq. (3.46)], takes the form uh! = (3.240) ‘This equation is another form of the equation of geodesic when ul is regarded as the unit tangent vector to the curve [see also Eq, (3.24)] 6. Ina Riemannian space Va, the geodesic is also defined as a curve whose first curvature relative to Va is zero at all points. The first curvatur defined as piu where u! is the unit tangent vector. From this definition of geodesic, we again get Eq. (3.24a), 8 Rules for Covariant Differentiation ‘The rules for covariant differentiation are as follows: 1. The covariant derivative of a linear combination of tensors, with constant coefficients, equals to the linear combination of these tensors after the covariant differentiation was performed. ‘Thus, for example, if A’ and Bj are two mixed tensors of rank two and a and 6 are scalars then (aS + DB) 4 = 32 (aAS + bBS) + Vig(aAh + OBL) — Thy (al + 6B) Oat aay ¢ —rt,aj] 0 | °F) 4 rept 1, Be = aAhy, OBS, (3.53) 2. The covariant derivatives of outer and inner products of tensors obey the same rules as that of the usual derivatives.Rules for Covariant Differentiation 7 Thus, for example, using Bq. (3.52), we have a (AB ye )p = ap (A'Byn) + ing A™ Bye = UIA’ Bra — Uh Boon oat OByk Bar Bie BE Tg Bya ATI Byak — ATE A'Bim oat i gm i ( OBik m m = (Fa arian) an — ai (BA 13 B00 Them) = AS By + A'Byiim (351) : Ore ress e (AB) g = LAB) + CyB. PYA'B aA‘ yp , OB" een Fe Bt ATE + Thy AB — TBAB = ALBi + ABs (3.55) Since A‘B; is an invariant, therefore (A'B)5 = Aus) 3. The covariant derivative of an invariant is same as its ordinary derivative. Thus, for example, if is an invariant and A, isa covariant vector, then from the definition we have a ” Bai (OAs) — Tyo’ cc Oxt ag, = gal + OAs (3.56) (@Aj) oA, Ay + ogy — oT A™ Also, from the product rule of covariant differentiation, we have (A) = O45 + OAs (3.57) Since Ay is arbitrary, from Eqs. (3.56) and (3.57), we have a6 Ou oa (3.58)78 Christoffel Symbols and Covariant Differentiation 4. The fundamental tensors are covariantly constant. ‘That is, the covariant derivative of fundamental tensors is zero. For example, first consider gj; and its covariant derivative 2 isn = GEE — Cages — Phan cr ek Dyin — Tig Ba; _ Dans ‘ack ~ ork =o (3.59) Similarly, for 5, we have t= 8 re ge regs Fn = ape t Mah — Thad T-Te (3.60) and moreover, since 949" = df, therefore the covariant derivative of g is also zero, Thus, the covariant derivative of all the three fundamental tensors is zero, Further (a: Amoin = 94mm + 95 Armin = 4 Aran (3.61) Hence, in a covariant differentiation gj; can be regarded as a constant. Since the covariant derivative of the metric tensor vanishes, it may be noted that raising and lowering of indices of tensors is not affected by the operation of covariant differentiation. Thus, for example AY, = (9 Ap) = 9" Ang (3.62) Remarks 1. The covariant derivative of a non-tensor has no meaning. 2. The laws of physics must be valid in all coordinate systems and thus they must be expressible through tensor equations. This is known as the principle of covariance—one of the basic postulate of general theory of relativity. So, whenever a tensor equation involves the derivative of a tensor quantity, it must be a covariant derivative; and the field equations of physics must be written in such a way that the ordinary deriv bbe replaced by covariant derivatives.Some Useful Formulas 79 EXAMPLE 3.9 From the rules of the covariant derivatives of contravariant, covariant and mixed vectors, prove that Ais Aun = Be — hedy—PhyAu a 8 ya rat a, = ot ral That Uk = gpk + PAs — Ty Solution Let P* and Q, be contravariant and covariant vectors, and if we express Ai as the product of P! and Q; then from the rules of the covariant differentiation of contravariant and covariant vectors, we have Abn = (P'Q)x = PQs + PQie ~ (Fa tte) a5 4 (FS - ra) oP* ri] =, « 8p oat + TPQ; Thy PQ = PelP'9y) + PyP'Q, 1h, Pres OAS rt al rh = age tT hay - Thal In a similar way, other two formulas can be proved. 3.9 Some Useful Formulas Here we shall derive some useful formulas which are related to the concept of covariant differentiation (a) Divergence of a vector field Let A! be a contravariant vector and A, be its covariant derivative, then A¥,, ‘unique scalar (invariant), is called the divergence of a vector field From the definition of covariant differentiation, we have div AY Ay oat rea80 Christoffel Symbols and Covariant Differentiation 2A 1 OI 4s bet Goer 1 aoe = pata (3.63) Consider the metric of a three-dimensional Buclidean space da? = de? + dy? + de? Here gir = 922 = 93s = 1 and det. gj = 1. Equation (3.63) in this case reduces to aat | ba? at a which is the well known expression for divergence in ordinary vector calculus. Moreover, equation divat = 1a gure is called the equation of continuity satisfied by the contravariant vector A’ (b) Gradient of a scalar and the Laplacian ‘The gradient of a scalar ¢ is a covariant vector defined as 28 gradp = Vd = 55 = 64 (3.64) ‘Therefore, the covariant and contravariant vectors Ay and A! associated with, 64 are given by (3.65) and (3.66) From Eq. (3.63), we have 10 ae Ava 1 diva; = diva' = 4 z va ‘Thus div grad 6 = div¥g = V4 aes (va055) (3.67) which is the Laplacian of a scalar field. If g is negative, then in the above equation /@ must be replaced by /=9.Some Useful Formulas 81 (c) Curl of a vector Consider the covariant vector and use Eq. (3.48) so that BAC pm. BAS pm. dne= (tt -rptn) = (24th) = Pee ic eae sent eer iene amano Aj. The curl operation is not applicable to contravariant vectors and tensors of higher rank, for example a (O48 ore OAs py oat = 4 ($5 a tpyde) - (SBF athe) 2 38 From Eq. (3.68) it may be noted that we can construct a skew symmetric tensor field from a vector field. Moreover, Bq. (3.68) leads to the following necessary and sufficient condition, Ais ie Theorem 3.2 The first covariant derivative of a covariant vector Ay is symmetric, if and only if the vector itself is gradient of a scalar ¢. Proof Let A; be a covariant vector and given that Ajj = Aji which gives OA PA, oo Ait = Migs Oe ~ Bat 8 Des dz so that after integration, we have [erm Be f Ave! serad 6, where f Ayde! = 6 a6 Conversely, given that Ay = 26 then a a6 syd (2), @ a -r524]- [2 (%)-m2e on ) Tags x (3 Tipe - Fo OxtOxI which completes the proof. (4) Divergence of a tensor field Let AY be a second rank contravariant tensor and A‘) be its covariant82 Christoffel Symbols and Covariant Differentiation derivative then the vector Aif associated with AY is called the divergence of the tensor field AY and we have 4 PAY , a= 8 ran ena ‘A contrnetion ofthis aqution wit jad ons o a a= SAD ram era 24 ou got y 1 OVE ay = Bria + ane 1 =~ Vpaer a Va) + TiyAr® (3.69) Ina similar way, wo also have A= Jeane) 4+1i,a" (370) If the tensor AY is symmetric then the two divergences given by Eqs. (3.69) and (8.70) are same, otherwise not. Moreover, for a symmetric tensor Ajj, We have yt py gt pa ii = Bes + PAE — Pyar AL 1 OVI ep as Se Eat rat oat, 1 VT yp gs ~ as yaar TyAt 1 ~ Spa va Thal on) Using the definition of Pi, we have Thal = Sol*Pay Af 1 te gj (Ogee — Og | Oasn gn At (* Bek + Bar 1 yh 204k (as 445 symmetric’ = 5A BE (as Mis symmetsic) Equation (3.71) now becomes Ay = Japs tiv ~ 5a (3.72)Some Useful Formulas 83 Now, differentiating both sides of gijg*” = 67", we get 25 hm 4. gy, 298m tat + on xt oaat oe xm 2085 5. 89k PO — yy Be (373) and we have 2 yrs 29ik _ 1 gamit a, Ok 34 Oat ~ 39 ‘go Ama BaF 1 PS = mio! (-0 9) {from Eq, (8.73)] ‘Thus, Eq. (3.72) can be expressed as 1a 1 agi" hs = Faas AD + 5am Gor (aa) ‘The divergence and eut! of tensors, as discussed above, can be used to express the thoerems of Green and Stokes in tensorial form as follows. Tensorial form of divergence theorem Let At be tensor field of rank one and Ug denote the outward drawn unit normal to any point of a closed surface $ bounding a volume V. ‘Then the statement of divergence theorem (Green’s theorem in space) is [sar [ atvas v $ ‘The invariant AY, is the divergence of A‘ and the invariant A'U; is the scalar product of A¥ and U; which is analogous to Af in vector notation, Tensorial form of Stokes’ theorem 1 se ai tangent vr tthe dal cue Cand Ute paste tn fonnal to thse 3 whch © a andar, thn Stas haan fata J PAWSst Christoffel Symbols and Covariant Differentiation which can also be written as 4 lf (ad _ OAs if suas! = i (sE- Sh) ase $ EXAMPLE 8.10 For the metric ds? = d2® + 2%dy?, show that div Ay ifthe veetor A; has the components (eos 2y, —2? sin 2y). Solution Here _fioe fre ‘dct on = glo el 2 0 a2 fr aetou=9 and Al = gMAy=g!A, = xcos2y, AP = Ay = 9g? Ay = —sin2y ‘Thus + rm EXAMPLE 3.11 If and ¢ are scalars, then show that div(6¥4) = V7U + 06,5 Solution We have div (6),) = 9 (603) 4 =o (x05 + OV5.4) =o Ong + POV 5x = Ov + oh dby EXAMPLE 3.12, For a scalar 4, show that curl (grad ) =Some Useful Formulas 85 cut (rao) = eur (32) EXAMPLE 3.13 Evaluate V*v in spherical polar coordinates (r, 0, é). Solution Here the metric is ds? = dr? + 120? +r? sin? adg? and 10 0 p=] 0 1? 0 00 1/r8sin?@ ans (7e)| 1a Oy 1 ay sind 90 (655) + aint OG is the expression for the Laplacian in spherical polar coordinates. Thus, by a method of tensor analysis the well-known formula of vector calculus has been obtained in a straightforward manner.86 Christoffel Symbols and Covariant Differentiation EXAMPLE 8.14 If F,, is a rank two skew symmetric tensor representing the curl of the covariant vector Ai, then show that Frigae + Fyne + Fag = Further show that on, OF, OF _ Oxk ox! * Oz (see also Chapter 6). Solution From the definition of the covariant derivative of a second rank covariant tensor, we have Paya + Fyn + Faay = 288 — pF — PYF + Ek Tek — TRF oF + FE PB FP Fim OFy | OF in | OF xk © Oak xd (8.75) because the terms containing Christoffel symbols cancel each other due to skeve symmetry of Fy ‘Also given that Fy is the curl of As, that is aA 0A, PS oni Oat which when used in Eq. (3.75) leads to Fraga + Byaa + Pris = 0 (3.76) which of course is equivalent to OFy | OF . Oxt * Ont EXAMPLE 3.15 Find the covariant derivative of the Levi-Civita contravariant tensor density of weight W = +1 and show that it is zero, Solution From the definit n of covariant differentiation, we have gat OCF a mitt PH gin 4g PE lint 4 PL gti pm QUE = Game + Tam a ‘hm ‘an saIntrinsic Derivative: Parallel Transport 87 Since e'! (from definition) is constant, the first term on the right-hand side of this equation is zero. To find the sum of the remaining five terms we calculate 64334 and we have 2H ph omy PZ, cine pS) lO pt gt2Im_ pm glast = Tipe 4 TE et PS e284 PH el24 pa, ett = The + Tb + T8m +0 bm — Ti =0 Similarly, it can be shown that the other components of Levi-Civita tensor density e'! gives the same result and thus ei"! = 0, 3.10 Intrinsic Derivative Parallel Transport ‘The intrinsic or absolute derivative of the vector A, along a curve 2! = 24(t) i end ste ine produto the ovariantdevatve of Ay and 2.1 ident yg en oa ot dei ae OA peg) dat ~ (2814) DAide , , dx? on ae Oe aay th Day dei > (a7) Similarly, we have 6A day ade oS nae (378) Moving a vector/tensor from one point to another while keeping it constant is called the parallel transport. In a Euclidean space we can move a veetor from one point to another without changing it but we cannot earry out this operation in curved space. In curved space the result of parallel transport will depend on the path taken. The vectors A; or A¥ are said to move parallely o parallely transported along a curve if their intrinsic derivatives along the curve are zero, that is, from Eqs. (3.77) and (3.78) for the parallel displacement of vectors Ay and At, we have day dt dx “dt hae (3.79)88 Christoffel Symbols and Covariant Differentiation and gece TA (3.708) ae Equation (3.79) is Imown as the equation of parallel transport. The intrinsic (absolute) derivative of an invariant is bo dz} _ 36 dx) _ do ie °F de ~ Bal ae Also, the absolute derivatives of Ai and A/S, are, respectively, given by (3.80) 6A, gy dhe det iy ed Re Mew ae = ae ee ge t+ eA Ge (3.81) and Sinn gpa at Be = As Gp data pp aie a2 po git dot See ah — PA at ast ast are apt 2 ry, apt oe age, ThAlme ae + PeAlmn gy + TtpAtin ar (3.82) Moreover, Sng ath 50% _ U0) Tae GEO REO Gs) ‘We now prove Theorem 3.3. For a Riemannian space there exists only one connection (Christoffel symbol of second kind) with respect to which the parallel displacement preserves scalar product, Proof Let A‘ and BY be two unit vectors defined along a curve in @ Riemannian space. ‘These vectors are parallely displaced only when their intrinsic derivatives vanish, that is «ded dt ‘The scalar product of A* and Bt is defined as gi) A‘B?, where gi; is the metric of the Riemannian space. The intrinsic derivative of this scalar product is o . . . yr AD + 9A + gyal [from above equations and Eq. (8.83)] which shows that the scalar product is preserved under parallel displacement. Note. This theorem is often referred to as the fundamental theorem of Riemannian geometry 0Null Geodesies 89 Remark With the concept of absolute differentiation, we can formulate Newton's second law in tensorial form as follows: If a particle moves along the curve 2* = 2'(t), where ¢ is the time parameter, tin v= i the wbcty of the parts and io 6 dvi Bat contravariant tensor of rank one. ‘The quantity "= js, in general, not a tensor and thus cannot represent the physical quantity “acceleration” in all coordinate systems, We define acceleration as the intrinsic derivative be the mass (an invariant and independent of time) of the particle, then ran Re ea tan a ant al oe P= mat = me (3.84) We also have aut a dv! da! ae pe ptt dt +1 dt @et dal da? - ae wae ae oe Since for a Euclidean space 91; = 0 and Ij, = 0, Ba. (3.85), therefore, leads ex! to the usual acceleration af = u c = 3.11 Null Geodesics In Section 3.4, we have mentioned the answer to the question “what do we mean by a straight line in Buclidean space?” The frst definition of a straight line is that its direction remain unchanged as we move along it. What is the implication of this definition in a more general space—the Riemannian space? We will answer this question in this section (using the second definition of the straight line we have already obtained the equation of a “straight line” in the Riemannian space (cf. Eq. (3.24)]); and thus find the condition for null geodesic. Let the coordinate of a point P on the curve 1 be 2‘(A), the curve 77 is given by real parameter A. In terms of this parameter \ the tangent vector dz! #f at the point P has the components t# =. Now, if the curve m does not90 Christoffel Symbols and Covariant Differentiation change its direction as we move along it, then the rate of change of t! along tis zero, ie., te=0 (386) which from the definition of t# and covariant differentiation yield (soe also Remark 5 of Section 3.7) @at , de det ae Tax ax (a7) ‘To know the nature of the parametr A, we multiply Eq. (3.87) by Daye © that ang ttt BE yy Adela aah a ax de TW ax ax Pa Using the definition of Mand symmetry and skew symmetry with respect, to k and p in the second term of the above equation, we get x! da? pn ck de) da? 20 RE ax Ox) dX dN ad Since gi = gkiy the first integral of this equation is| da! da? a Ge = constant (3.88) ‘Tims, A can atmost be replaced by aX +6, where a(# 0) and 6 are arbitrary real constants, ‘The “straight line” given by Eq. (3.87) is called the geodesic. If the constant in Eq, (3.88) is zero, positive or, negative, the geodesic is known as null, time like or space like. In the Intter two cases we can choose the constant to be +1 or ~1 and replace 2 by a, the distance measured along the geodesic. ‘Thus, for a non-null geodesic, we have zt de* ded wea ae (389) and eae one (8.80) while for mll geodesic, we have eat fy 0 (6.90) along with de‘ dxc* ky (3.90a)Null Geodesies a1 ‘The parameter in terms of which the geodesic equation can take the form of Eq, (3.87) is the affine parameter (see also Section 3.5). It may be noted that Eq. (3.89) and (3.89a) is same as of Eq. (3.24) and thus a geodesic may be looked upon both as a line of stationary distance and as a line of unchanging direction. ‘The geodesic Eq. (3.87) expresses the parallel transport of a vector and de SG of & geodesic remains parallel to states that the tangent vector ¢' itself Remarks 1. In the context of general theory of relativity, the equation of geodesic [ct., Eq. (3.24) and (3.87)] describes the motion of a test particle in the gravitational field, while the null geodesic Ea. [(3.90) and (3.90a)[ describes the propagation of zero-rest- mass particles (for example, the propagation of electromagnetic waves). 2. ‘The fact that the vector a‘ does not change its length when parallely transported can be expressed by the equation (see also Remark 2 of Section 3.4) Late — T = 5oni'#* = constant which is simply the principle of conservation of energy. see = [ VinFat, we hve # = VET och eters tur decue! in Ren of ton ive of a vector A‘ along a curve vanishes at every point of the curve, show that the magnitude of the vector is constant along the curve. Solution From the definition of intrinsic derivative, we have 5 (43) — Seas, FA) = Had) = (AAD ae “aE (a 4 at = 4a ala de Given that the intrinsic derivative of At is zero, ie, ALLE 0. Also SF AL = gn" and guy = 0 which loads to 9a AS = 0, Thorofore, from Bq, (8.91) 5 yy 4 cg2 bu - 4a)92 Christoffel Symbols and Covariant Differentiation and A? being the magnitude of A’ is constant, as A? is an invariant [ef., Eq, (3.80)] EXAMPLE 3.17 If 4! and BY are two unit vectors defined along a curve such that their intrinsic derivatives along the curve are zero, show that the angle between the vectors A‘ and B' is constant. Solution Given that 6A, det a AD ae a and since Ay = gijA? and gin = 0, we have a az! Ais = 0, Bis (3.93) If @ is the angle between A‘ and B' then. aoe AB J AA\(B;B) which on differentiation leads to [using Eq. (3.80)] dd dt a0 ” " ~ sin OF = (ABS + AaB From Eqs. (3.92) and (3.93), this equation yields a qn ‘Thus, # is a constant EXAMPLE 3.18 Show that geodesic that is time-like (space-like, mull) ‘ata given point remains time-like (space-like, null) along its path through the space Solution Here we shall consider the parallel transport of a vector A; and Ais parallely transported then from bAg 6At a tA =o 5 at alta =A’ ‘The path element dr' along a geodesic itself is being parallely transported sts) =O natn (22) (2) eine st nd ne dx‘dz, continues to be time-like (space-like, null) if it is started as time-like GoiAlternative Derivation of Equation of Geodesic 93 EXAMPLE 3.19 Show that the curve whose parametric representation is ane f reosteos ds yee f reostsinds see f rsinads te fra ‘mull curve in a space whose metrie is ds? = ds? — dy? — de? + Bde? Solution Here ds? = —dz? — dy? — de? + dt? = —(cr 008 00s 6)? — (cr cos 0sin d)? — (er sin)? + \(ax?)? 1 2 Fey" (ii) ds? = f(x)dz? + dy? + de? + (lil) do? = —de? = dy? — de? + f(e,ys2)d? 8.8 For the metric de? = dr® + r2d02, show that 20 ae)? a (a0)? (08 = n= coma (22)' 4 (2) ~ art itp of the geodesic equation. (Gi) From (i), find a first-order differential equation for r(0).