Ricci

A Mathematica package for doing tensor calculations in dierential geometry Users Manual

Version 1.32

By John M. Lee assisted by Dale Lear, John Roth, Jay Coskey, and Lee Nave

Ricci
A Mathematica package for doing tensor calculations in dierential geometry Users Manual Version 1.32 By John M. Lee assisted by Dale Lear, John Roth, Jay Coskey, and Lee Nave

Copyright c 19921998 John M. Lee All rights reserved Development of this software was supported in part by NSF grants DMS-9101832, DMS-9404107 Mathematica is a registered trademark of Wolfram Research, Inc. This software package and its accompanying documentation are provided as is, without guarantee of support or maintenance. The copyright holder makes no express or implied warranty of any kind with respect to this software, including implied warranties of merchantability or tness for a particular purpose, and is not liable for any damages resulting in any way from its use. Everyone is granted permission to copy, modify and redistribute this software package and its accompanying documentation, provided that: 1. All copies contain this notice in the main program le and in the supporting documentation. 2. All modied copies carry a prominent notice stating who made the last modication and the date of such modication. 3. No charge is made for this software or works derived from it, with the exception of a distribution fee to cover the cost of materials and/or transmission.

John M. Lee Department of Mathematics Box 354350 University of Washington Seattle, WA 98195-4350 E-mail: [email protected] Web: http://www.math.washington.edu/~lee/

CONTENTS

Contents
1 Introduction 1.1 1.2 1.3 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Obtaining and using Ricci . . . . . . . . . . . . . . . . . . . . . . A brief look at Ricci . . . . . . . . . . . . . . . . . . . . . . . . . 6 6 7 7 10 10 12 12 13 16 17 18 19 19 19 20 20 21 21 22 22 22 23 23 24 25 25 25 26 26

2 Ricci Basics 2.1 2.2 2.3 2.4 2.5 2.6 2.7 Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Indices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mathematical functions . . . . . . . . . . . . . . . . . . . . . . . Basic tensor expressions . . . . . . . . . . . . . . . . . . . . . . . Saving your work . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 Products, Contractions, and Symmetrizations 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 TensorProduct . . . . . . . . . . . . . . . . . . . . . . . . . . . . Wedge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . SymmetricProduct . . . . . . . . . . . . . . . . . . . . . . . . . . Dot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Inner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . HodgeInner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Int . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sym . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4 Derivatives 4.1 4.2 4.3 4.4 4.5 4.6 4.7 Del . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CovD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Div . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Grad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Laplacian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Extd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ExtdStar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

CONTENTS 4.8 4.9 LaplaceBeltrami . . . . . . . . . . . . . . . . . . . . . . . . . . . Lie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

26 26 28 28 29 29 29 29 30 30 30 30 30 31 31 32 32 33 33 34 34 34 34 34 36 36 38 40 40 40 41 42

5 Simplifying and Transforming Expressions 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 TensorSimplify . . . . . . . . . . . . . . . . . . . . . . . . . . . . SuperSimplify . . . . . . . . . . . . . . . . . . . . . . . . . . . . . TensorExpand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . AbsorbMetrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . RenameDummy . . . . . . . . . . . . . . . . . . . . . . . . . . . . OrderDummy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CollectConstants . . . . . . . . . . . . . . . . . . . . . . . . . . . FactorConstants . . . . . . . . . . . . . . . . . . . . . . . . . . . SimplifyConstants . . . . . . . . . . . . . . . . . . . . . . . . . .

5.10 BasisExpand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.11 BasisGather . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.12 CovDExpand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.13 ProductExpand . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.14 PowerSimplify . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.15 CorrectAllVariances . . . . . . . . . . . . . . . . . . . . . . . . . 5.16 NewDummy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.17 CommuteCovD . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.18 OrderCovD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.19 CovDSimplify . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.20 LowerAllIndices . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.21 TensorCancel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Dening Relations Between Tensors 6.1 6.2 DeneRelation . . . . . . . . . . . . . . . . . . . . . . . . . . . . DeneRule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

7 Special Features 7.1 7.2 7.3 7.4 One-dimensional bundles . . . . . . . . . . . . . . . . . . . . . . . Riemannian metrics . . . . . . . . . . . . . . . . . . . . . . . . . Matrices and 2-tensors . . . . . . . . . . . . . . . . . . . . . . . . Product tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . .

CONTENTS 7.5 7.6 Connections, torsion, and curvature . . . . . . . . . . . . . . . .

44 46 49 49 87

Non-default connections and metrics . . . . . . . . . . . . . . . .

8 Reference List 8.1 8.2 Ricci commands and functions . . . . . . . . . . . . . . . . . . . Global variables . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 INTRODUCTION

1
1.1

Introduction
Overview

Ricci is a Mathematica package for doing symbolic tensor computations that arise in dierential geometry. It supports: Tensor expressions with and without indices The Einstein summation convention Correct manipulation of dummy indices Mathematical notation with upper and lower indices Automatic calculation of covariant derivatives Automatic application of tensor symmetries Riemannian metrics and curvatures Dierential forms Any number of vector bundles with user-dened characteristics Names of indices indicate which bundles they refer to Complex bundles and tensors Conjugation indicated by barred indices Connections with and without torsion Ricci is named after Gregorio Ricci-Curbastro (1853-1925), who invented the tensor calculus (what M. Spivak calls the debauch of indices). This manual describes the capabilities and functions provided by Ricci. To use Ricci and this manual, you should be familiar with Mathematica, and with the basic objects of dierential geometry (manifolds, vector bundles, tensors, connections, and covariant derivatives). Chapter 8 contains a complete reference list of all the Ricci commands, functions, and global variables. Most of the important Ricci commands are described in some detail in the main text, but there are a few things that are explained only in Chapter 8. I would like to express my gratitude to my collaborators on this project: Dale Lear, who wrote the rst version of the software and contributed uncountably many expert design suggestions; John Roth, and Lee Nave, who reworked some of the most dicult sections of code; Jay Coskey and Pm Weizenbaum, who contributed invaluable editorial assistance with this manual; the National Science Foundation and the University of Washington, who provided generous nancial support for the programming eort; and all those mathematicians who tried out early versions of this software and contributed suggestions for improvement.

1 INTRODUCTION

1.2

Obtaining and using Ricci

Ricci requires Mathematica version 2.0 or greater. It will also run with Mathematica 3.0, although it will not produce formatted StandardForm output. Therefore, when you use Ricci in a Mathematica 3.0 notebook, you should change the default output format to OutputForm. When you rst load Ricci, youll get a message showing how to do this by choosing Default Output Format Type from the Cell menu. The Ricci source le takes approximately 287K bytes of disk storage, including about 49K bytes of on-line documentation. I have tested the package on a DEC Alpha system and a Pentium 100, where it runs reasonably fast and seems to require about 6 or 7 megabytes of memory. I dont have any idea how it will run on other systems, but I expect that Ricci will be very slow on some platforms. The source les for Ricci are available to the public free of charge, either via the World-Wide Web from http://www.math.washington.edu/~lee/Ricci/ or by anonymous ftp from ftp.math.washington.edu, in directory pub/Ricci. Youll need to download the Ricci source le Ricci.m and put it in a directory that is accessible to Mathematica. (You may need to change the value of Mathematicas $Path variable in your initialization le to make sure that Mathematica can nd the Ricci lessee the documentation for the version of Mathematica that youre using.) Once youve successfully transferred all the Ricci les to your own system, start up Mathematica and load the Ricci package by typing <<Ricci.m. Once youve loaded Ricci into Mathematica, you can type ?name for information about any Ricci function or command. If you have questions about using Ricci, suggestions for improvement, or a problem that you think may be caused by a bug in the program, please contact the author <[email protected]>.

1.3

A brief look at Ricci

To give you a quick idea what typical Ricci input and output look like, here are a couple of examples. Suppose alpha and beta are 1-tensors. Ricci can manipulate tensor products and wedge products: In[4]:= TensorProduct[alpha,beta] Out[4]= alpha (X) beta In[5]:= Wedge[alpha,beta] Out[5]= alpha ^ beta and exterior derivatives:

1 INTRODUCTION

In[6]:= Extd[%] Out[6]= d[alpha] ^ beta - d[beta] ^ alpha To express tensor components with indices, you just type the indices in brackets immediately after the tensor name. Lower and upper indices are typed as L[i] and U[i], respectively; in output form they appear as subscripts and superscripts. Indices that result from covariant dierentiation are typed in a second set of brackets. For example, if alpha is a 1-tensor, you indicate the components of alpha and its rst covariant derivative as follows: In[7]:= alpha [L[i]] Out[7]= alpha i In[8]:= alpha [L[i]] [L[j]] Out[8]= alpha i; j Ricci always uses the Einstein summation convention: any index that appears as both a lower index and an upper index in the same term is considered to be implicitly summed over, and the metric is used to raise and lower indices. For example, if the metric is named g, the components of the inner product of alpha and beta can be expressed in either of the following ways: In[9]:= alpha[L[k]] beta[U[k]] k beta

Out[9]= alpha k

In[10]:= g[U[i],U[j]] alpha[L[i]] beta[L[j]] i j Out[10]= alpha i beta j g

Riccis simplication commands can recognize the equality of two terms, even when their dummy indices have dierent names. For example:

1 INTRODUCTION

In[11]:= %9 + %10 i j Out[11]= alpha i beta j g k + alpha beta k

In[12]:= TensorSimplify[%] i beta

Out[12]= 2 alpha i

You can take a covariant derivative of this last expression just by putting another index in brackets after it: In[13]:= % [L[j]] i beta i beta ;j

Out[13]= 2 (alpha i ;j

+ alpha i

The remainder of this manual will introduce you more thoroughly to the capabilities of Ricci.

2 RICCI BASICS

10

Ricci Basics

There are four kinds of objects used by Ricci: bundles, indices, constants, mathematical functions, and tensors. This chapter describes these objects, and then describes how to construct simple tensor expressions. The last section in the chapter describes how to save your work under Ricci.

2.1

Bundles

In Ricci, the tensors you manipulate are considered to be sections of one or more vector bundles. Before dening tensors, you must dene each bundle you will use by calling the DefineBundle command. When you dene a bundle, you must specify the bundles name, its dimension, the name of its metric, and the index names you will use to refer to the bundle. You may also specify (either explicitly or by default) whether the bundle is real or complex, the name of the tangent bundle of its underlying manifold, and various properties of the bundles default metric, connection, and frame. The simplest case is when there is only one bundle, the tangent bundle of the underlying manifold. For example, the command In[2]:= DefineBundle[ tangent, n, g, {i,j,k} ] denes a real vector bundle called tangent, whose dimension is n, and whose metric is to be called g. The index names i, j, and k refer to this bundle. If any expression requires more than three tangent indices, Ricci generates index names of the form k1, k2, k3, etc. If the bundle is one-dimensional, there can only be one index name. Every bundle must have a metric, which is a nondegenerate inner product on the bers. By default, Ricci assumes the metric is positive-denite, but you can override this assumption by adding the option PositiveDefinite -> False to your DefineBundle call. The metric name you specify in your DefineBundle call is automatically dened by Ricci as a symmetric, covariant 2-tensor associated with this bundle. By default, Ricci assumes all bundles to be real. You may dene a complex bundle by including the option Type -> Complex in your DefineBundle call. For example, suppose you also need a complex two-dimensional bundle called fiber. You can dene it as follows: In[3]:= DefineBundle[ fiber, 2, h, {a,b,c}, Type -> Complex] This species that fibers metric is to be called h, and that the index names a, b, and c refer to fiber. The fth argument in the example above, Type -> Complex, is a typical example of a Ricci option. Like options for built-in Mathematica functions, options

2 RICCI BASICS

11

for Ricci functions are typed after the required arguments, and are always of the form optionname -> value. This option species that fiber is to be a complex bundle. Any time you dene a complex bundle, you also get its conjugate bundle. In this case the conjugate bundle is referred to as Conjugate[fiber]; the barred versions of the indices a, b, and c will refer to this conjugate bundle. Mathematically, the conjugate bundle is dened abstractly as follows. If V is a complex n-dimensional vector bundle, let J: V V denote the complex structure map of V the real-linear endomorphism obtained by multiplying by i = 1. If CV denotes the complexication of V, i.e., the tensor product (over R) of V with the complex numbers, then J extends naturally to a complexlinear endomorphism of CV . CV decomposes into a direct sum CV = V V , where V , V are the i and (i)-eigenspaces of J, respectively. V is naturally isomorphic to V itself. By convention, the conjugate bundle is dened by V = V . The metric on a complex bundle is automatically dened as a 2-tensor with Hermitian symmetries. In Ricci, this is implemented as follows. A Hermitian tensor h on a complex bundle is a real, symmetric, complex-bilinear 2-tensor on the direct sum of the bundle with its conjugate, with the additional property that hab = ha = 0. Thus the only nonzero components of h are those of the b form ha = h = hb . a ba b (It is done this way because Riccis index conventions require that all tensors be considered to be complex-multilinear.) To obtain the usual sesquilinear form on a complex bundle V , you apply the metric to a pair of vectors x, y, where x, y V . In Riccis input form, this inner product would be denoted by Inner[ x, Conjugate[y] ]. For every bundle you dene, you must specify, either explicitly or by default, the name of the tangent bundle of the underlying manifold. Ricci needs this information, for example, when generating indices for covariant derivatives. By default, the rst bundle you dene (or the direct sum of this bundle and its conjugate if the bundle is complex) is assumed to be the underlying tangent bundle for all subsequent bundles. If you want to override this behavior, you can either give a value to the global variable $DefaultTangentBundle before dening bundles, or explicitly include the TangentBundle option in each call to DefineBundle. See the Reference List, Chapter 8, for more details. Another common type of bundle is a tangent bundle with a Riemannian metric. The DefineBundle option MetricType -> Riemannian species that the bundle is a Riemannian tangent bundle. Riemannian metrics are described in Section 7.2. There are other options you can specify in the DefineBundle call, such as FlatConnection, TorsionFree, OrthonormalFrame, CommutingFrame, PositiveDefinite, and ParallelFrame. If you decide to change any of these

2 RICCI BASICS

12

options after the bundle has already been dened, you can call Declare. See the Reference List, Chapter 8, for a complete description of DefineBundle and Declare.

2.2

Indices

The components of tensors are represented by upper and lower indices. Ricci adheres to the Einstein index conventions, as follows. For any vector bundle V , sections of V itself are considered as contravariant 1-tensors, while sections of the dual bundle V are considered as covariant 1-tensors. The components of contravariant tensors have upper indices, while the components of covariant tensors have lower indices. Higher-rank tensors may have both upper and lower indices. Indices are distinguished both by their horizontal positions and by their vertical positions; the altitude of each index indicates whether that index is covariant or contravariant. Any index that appears twice in the same term must appear once as an upper index and once as a lower index, and the term is understood to be summed over that index. (Indices associated with one-dimensional bundles are an exception to this rule; see Section 7.1 on one-dimensional bundles below.) In input form and in Riccis internal representation, indices have the form L[name] for a lower index and U[name] for an upper index. Barred indices (for complex bundles only) are typed in input form as LB[name] and UB[name]; they are represented internally by L[name[Bar]] and U[name[Bar]]. In output form, upper and lower indices are represented by superscripts and subscripts, with or without bars. This special formatting can be turned o by setting $TensorFormatting=False. Every index name must be associated with a bundle. This association is established either by listing the index name in the call to DefineBundle, as in the examples above, or by calling DefineIndex. You can create new names without calling DefineIndex just by appending digits to an already-dened index name. For example, if index j is associated with bundle x, then so are j1, j25, etc.

2.3

Constants

In the Ricci package, a constant is any expression that is constant with respect to covariant dierentiation in all directions. You can dene a symbol to be a constant by calling DefineConstant: In[10]:= DefineConstant[ c ] This denes c to be a real constant (the default), and ensures that covariant derivatives of c, such as c[L[i]], are interpreted as 0. To dene a complex constant, use

2 RICCI BASICS

13

In[11]:= DefineConstant[ d, Type -> Complex ] If a constant is real, you may specify that it has other attributes by giving a Type option consisting of one or more keywords in a list, such as Type -> {Positive,Real}. The complete list of allowable Type keywords for constants is Complex, Real, Imaginary, Positive, Negative, NonPositive, NonNegative, Integer, Odd, and Even. The Type option controls how the constant behaves with respect to conjugation, exponentiation, and logarithms. After a constant has been dened, you can change its Type option by calling Declare, as in the following example: In[12]:= Declare[ d, Type -> {NonNegative,Real} ] In the Ricci package, any expression that does not contain explicit tensors is assumed to be a constant for the purposes of covariant dierentiation, so in some cases it is not necessary to call DefineConstant. However, if you attempt to insert indices into a symbol c that has not been dened as either a constant or a tensor, you will get output that looks like c[L[i],L[j]], indicating that Ricci does not know how to interpret c. For this reason, as well as to tell Ricci what type of constant c is, it is always a good idea to dene your constants explicitly. Note that constants are not given Mathematicas Constant attribute, for the following reason. Some future version of Ricci may allow tensors depending on parameters such as t; in this case, t will be considered a constant from the point of view of covariant dierentiation, but may well be non-constant in expressions such as D[f[t],t]. You can test whether a tensor expression is constant or not by applying the Ricci function ConstantQ.

2.4

Tensors

The most important objects that Ricci uses in calculations are tensors. Ricci can handle tensors of any rank, and associated with any vector bundle or direct sum or tensor product of vector bundles. Scalar functions are represented by 0-tensors; other objects such as vector elds, dierential forms, metrics, and curvatures are represented by higher-rank tensors. Tensors are created by the DefineTensor command. For example, to create a 1-tensor named alpha you could type: In[6]:= DefineTensor[ alpha, 1 ] This creates a real tensor of rank one, which is associated with the current default tangent bundle, usually the rst bundle you dened. All tensors are assumed by default to be covariant; thus the tensor alpha dened above can be thought of as a covector eld or 1-form. To dene a contravariant 1-tensor (i.e., a vector eld), you could type:

2 RICCI BASICS

14

In[7]:= DefineTensor[ v, 1, Variance -> Contravariant ] To create a scalar function, you simply dene a tensor of rank 0: In[8]:= DefineTensor[ u, 0, Type -> {NonNegative, Real} ] The Type option of DefineTensor controls how the tensor behaves with respect to conjugation, exponentiation, and logarithms. For a rank-0 tensor, its value can be either a single keyword or a list of keywords as in the example above. The complete list of allowable Type keywords for 0-tensors is Complex, Real, Imaginary, Positive, Negative, NonPositive, and NonNegative. For higherrank tensors, the Type option can be Real, Imaginary, or Complex. The default is always Real. If you include one of the keywords indicating positivity or negativity, you may leave out Real, which is assumed. For higher-rank tensors, you can specify tensor symmetries using the Symmetries option. For example, to dene a symmetric covariant 2-tensor named h, type: In[9]:= DefineTensor[ h, 2, Symmetries -> Symmetric ] Other common values for the Symmetries option are Alternating (or, equivalently, Skew) and NoSymmetries (the default). To associate a tensor with a bundle other than the default tangent bundle, use the Bundle option: In[10]:= DefineTensor[ omega, 2, Bundle -> fiber, Type -> Complex, Variance -> {Contravariant,Covariant} ] This species that omega is a complex 2-tensor on the bundle named fiber, which is contravariant in the rst index and covariant in the second. If the value of the Variance option is a list, as in the example above, the length of the list must be equal to the rank of the tensor, and each entry species the variance of the corresponding index position. The Bundle option may also be typed as a list of bundle names: this means that the tensor is associated with the direct sum of all the bundles in the list. Any time you insert an index into a tensor that does not belong to one of the bundles with which the tensor is associated, you get 0. After a tensor has been dened, you can change certain of its options, such as Type and Bundle, by calling Declare. For example, to change the 0-tensor u dened above to be a complex-valued function, you could type: In[11]:= Declare[ u, Type -> Complex ]

2 RICCI BASICS Internally, Ricci represents tensors in the form Tensor[ name, {i,j,...}, {k,l,...} ]

15

where i, j,... are the tensor indices, and k, l,... are indices resulting from covariant dierentiation. In input form, you represent an unindexed tensor just by typing its name. Indices are inserted by typing them in brackets after the tensor name, while dierentiated indices go inside a second set of brackets. A tensor of rank k should be thought of as having k index slots. Each index slot is associated with a bundle or list of bundles (specied in the DefineTensor command); if an inserted index is not associated with one of the bundles for that slot, the result is 0. Once all index slots are full, indices resulting from covariant dierentiation can be typed in a second set of brackets. For 0-tensors, all indices are assumed to result from dierentiation. For example, suppose eta is a 2-tensor and u is a 0-tensor (i.e., a scalar function). Table 1 shows the input, internal, and output forms for various tensor expressions involving eta and u. input eta internal Tensor[eta,{},{}] output eta j eta [L[i],U[j]] Tensor[eta,{L[i],U[j]},{}] eta i j eta [L[i],U[j]] [L[k]] Tensor[eta,{L[i],U[j]},{L[k]}] eta i u Tensor[u,{},{}] u ;k

u [L[a],L[b]]

Tensor[u,{},{L[a],L[b]}]

u ;a b

Table 1: Input, internal, and output forms for tensors The conjugate of a complex tensor without indices is represented in input form by Conjugate[name]; you can type indices in brackets following this expression just as for ordinary tensors, as in Conjugate[name] [L[i],L[j]]. In internal form, a conjugate tensor looks just as in Table 1, except that the name is replaced by the special form name[Bar]. In output form, the conjugate of a tensor is represented by a bar over the name. Indices in any slot can be upper or lower. A metric with lower indices represents the components of the metric on the bundle itself, and a metric with upper indices represents the components of the inverse matrix, which can be interpreted

2 RICCI BASICS

16

invariantly as the components of the metric on the dual bundle. For any other tensor, if an index is inserted that is not at the natural altitude for that slot (based on the Variance option when the tensor was dened), it is assumed to have been raised or lowered by means of the metric. For example, if a is a covariant 1-tensor, then the components of a ordinarily have lower indices. A component with an upper index is interpreted as follows: i i j a = g

a j

As explained above, components of tensors ordinarily have lower indices in covariant slots, and upper indices in contravariant slots. Ricci always assumes there is a default local basis (or moving frame) for each bundle you dene, and components of tensors are understood to be computed with respect to this basis. The default contravariant basis vectors (i.e., basis vectors for the bundle itself) are referred to as Basis[L[i]], and the default covariant basis vectors (representing the dual basis for the dual bundle) as Basis[U[i]]. (You may specify dierent names to be used in input and output forms by using the BasisName and CoBasisName options of DefineBundle.) Note that the index conventions for basis vectors are opposite those for components: this is so that the expansion of tensors in terms of basis vectors will be consistent with the summation convention. The Ricci function BasisExpand can be used to expand any tensor expression in terms of the default basis. For example: In[13]:= BasisExpand[alpha] k3 Basis

Out[13]= alpha k3

2.5

Mathematical functions

If u is a scalar function (0-tensor), you may form other scalar functions by applying real- or complex-valued mathematical functions such as Log or Sin to u. This is indicated in Riccis input form and output form simply by typing Log[u] or Sin[u]. You may also dene your own mathematical functions and use them in tensor expressions. In order for Ricci to correctly interpret an expression such as f[u], it must be told that f is a mathematical function, and what its characteristics are. You do this by calling DefineMathFunction. For example, to dene f as a complex-valued function of one variable, you could type:

2 RICCI BASICS

17

In[14]:= DefineMathFunction[ f, Type->Complex ] The Type option determines how the function behaves with respect to conjugation, exponentiation, and logarithms. The default is Real, which means that the function is always assumed to take real values. Other common options are Type -> {Positive,Real}, Type -> {NonNegative,Real}, and Type -> Automatic, which means that Conjugate[f[x]] = f[Conjugate[x]]. Ricci automatically calls DefineMathFunction for the Mathematica functions Log, Sin, Cos, Tan, Sec, Csc, and Cot. If you wish to use any other built-in mathematical functions in tensor expressions, you must call DefineMathFunction yourself.

2.6

Basic tensor expressions

Tensor expressions are built up from tensors and constants, using the arithmetic operations Plus, Times, and Power, the Mathematica operators Conjugate, Re, and Im, scalar mathematical functions, and the Ricci product, contraction, symmetrization, and dierentiation operators described below. Tensor expressions can be generally divided into three classes: pure tensor expressions, without any indices, in which tensor names are combined using arithmetic operations and tensor operators such as TensorProduct and Extd; component expressions, in which all index slots are lled, derivatives are indicated by components of covariant derivatives, and tensor components are combined using only arithmetic operations; and mixed tensor expressions, in which some but not all index slots are lled, and which usually arise when expressions of the preceding two types are combined. Component expressions are considered to have rank 0 when they are acted upon by Riccis tensor operators. In most cases, Ricci handles all three kinds of expressions in a uniform way, but there are some operations, such as computing components of covariant derivatives, that require component expressions. You may insert indices into any pure or mixed tensor expression, thereby converting it to a component expression, by typing the indices in brackets after the expression, as in x[L[i],U[j]], which is a shorthand input form for the explicit Ricci function InsertIndices[ x, {L[i],L[j]} ]. For most tensor expressions, the number of indices inserted must be equal to the rank of the tensor expression. (The only exception is product tensors, described in Section 3 below.) For example, if alpha and beta are 1-tensors, their tensor product, indicated by TensorProduct[alpha,beta], is a 2-tensor, so you can insert two indices:

2 RICCI BASICS

18

In[12]:= TensorProduct[alpha,beta] [L[i],L[j]] Out[12]= alpha i beta j

If x is a scalar (rank 0) expression, it can be converted to a component expression by inserting zero indices, as in InsertIndices[x,{}] or x[], or by typing BasisExpand[x]. For example, Inner[alpha,beta] represents the inner product of alpha and beta: In[13]:= Inner[alpha,beta]//BasisExpand k1 Out[13]= alpha

beta k1

If you type indices in brackets after a component expression (one in which all index slots are already lled), you get components of covariant derivatives. Thus: In[14]:= %13 [L[j]] k1 Out[14]= alpha k1 + alpha beta k1 ;j

beta ;j k1

If you type indices in brackets after a scalar (rank-0) expression that is not a component expression (i.e., has unlled index slots), it is rst converted to a component expression before covariant derivatives are taken. The Ricci functions that can be used to construct tensor expressions are described in later chapters.

2.7

Saving your work

Ricci provides a function called RicciSave for saving all the denitions you have made during your current Mathematica session. The command RicciSave["filename"] writes your denitions in the le called filename in Mathematica input form. RicciSave also writes the current denitions of Riccis predened tensors Basis, Curv, Tor, Conn, and Kronecker (in case you have dened any relations involving them), and of its global variables (those whose names start with $). Any

2 RICCI BASICS

19

previous contents of filename are erased. You can read the denitions back in by typing <<filename.

3 PRODUCTS, CONTRACTIONS, AND SYMMETRIZATIONS

20

Products, Contractions, and Symmetrizations

This chapter describes the most important Ricci functions for specifying various kinds of products, contractions, and symmetrizations of tensor expressions. Although Ricci performs some automatic simplication on most of these functions, such as expanding linear combinations of arguments, for the most part these functions are simply maintained unevaluated until you insert indices.

3.1

TensorProduct

The basic product operation for tensors is TensorProduct. In input form, this can be abbreviated TProd. In[15]:= TProd[alpha, beta] Out[15]= alpha (X) beta As you can see here, Riccis output form for tensor products uses an approximation of the tensor product symbol . Ricci automatically expands tensor products of sums and scalar multiples.

3.2

Wedge

Wedge products (exterior products) of dierential forms are indicated by the Ricci operator Wedge. In output form, wedge products are indicated by a caret: In[16]:= Wedge[alpha, beta] Out[16]= alpha ^ beta The arguments to Wedge must be alternating tensors. Ricci automatically expands wedge products of sums and scalar multiples, and arranges the factors in lexical order, inserting signs as appropriate. There are two common conventions for relating wedge products to tensor products. Ricci supports both conventions; which one is in use at any given time is controlled by the global variable $WedgeConvention. You should pick one convention and set $WedgeConvention at the beginning of your calculation; if you change it, you may get inconsistent results. In Ricci, the two wedge product conventions are referred to as Det and Alt. Under the Alt convention (the default), the wedge product is dened by = Alt( ), where Alt is the projection onto the alternating part of the tensor. This convention is used, for example, in the dierential geometry texts by Bishop/Goldberg,

3 PRODUCTS, CONTRACTIONS, AND SYMMETRIZATIONS

21

Kobayashi/Nomizu, and Helgason. Under this convention, on an n-dimensional bundle, the wedge product of all n basis elements is 1/n! times the determinant function. Thus, if e1 , . . . , en are basis covectors, e1 . . . en = 1 det . n!

The other wedge product convention, called Det, is dened by = (p + q)! Alt( ). p!q!

where is a p-form and a q-form. This convention is used, for example, in the texts by Spivak, Boothby, and Warner. The name Det comes from the fact that, under this convention, if i are 1-forms and Xj are vectors, 1 . . . k (X1 , . . . , Xk ) = det( i (Xj )). In particular, e1 . . . en = det .

3.3

SymmetricProduct

In Ricci, symmetric products are represented by ordinary multiplication. Thus, if a and b are 1-tensors, then a b represents their symmetric product a b, a 2-tensor. Similarly, a^2 represents a a. In input form, you can type a symmetric product just as you would an ordinary product. If you prefer, for clarity, you may indicate symmetric products explicitly using the SymmetricProduct function: In[17]:= SymmetricProduct[alpha,beta] Out[17]= alpha * beta In output form, Ricci inserts explicit asterisks in products of tensors of rank greater than 0, to remind you that multiplication is being interpreted as a symmetric product. The symmetric product is dened by a b = Sym(a b), where Sym is the projection onto the symmetric part of a tensor.

3.4

Dot

Dot is a Mathematica function that is given an additional interpretation by Ricci. If a and b are tensor expressions, then Dot[a,b] or a.b is the tensor

3 PRODUCTS, CONTRACTIONS, AND SYMMETRIZATIONS

22

formed by contracting the last index of a with the rst index of b. It is easiest to see what this means by inserting indices. For example, suppose a is a 2-tensor and e is a 3-tensor: In[18]:= (a.e) [L[i],L[j],L[k]] k4 Out[18]= a i k4 e j k

Dot can be applied to three or more tensor expressions, as long as all of the tensors except possibly the rst and last have rank two or more. Ricci expands dot products of sums and scalar multiples automatically. Dot products of 1tensors are automatically converted to inner products.

3.5

Inner

The Mathematica function Inner is given an additional interpretation by Ricci. If a and b are tensor expressions of the same rank, Inner[a,b] is their inner product, taken with respect to the default metric(s) on the bundle(s) with which a and b are associated. In output form, Inner[a,b] appears as <a, b>. For example: In[19]:= Inner[a,b] Out[19]= <a, b> In[20]:= % // BasisExpand k5 k6 Out[20]= a b k5 k6

3.6

HodgeInner

HodgeInner is the inner product ordinarily used for dierential forms. This diers from the usual inner product by a constant factor, determined by the global variable $WedgeConvention. It is dened in such a way that, if ei are orthonormal 1-forms, then the k-forms ei1 . . . eik , are orthonormal. <<a, b>>. i1 < . . . < ik

In output form, HodgeInner[alpha,beta] appears as

3 PRODUCTS, CONTRACTIONS, AND SYMMETRIZATIONS

23

3.7

Int

If v is a vector eld and alpha is a dierential form, Int[v,alpha] represents interior multiplication of v into alpha (usually denoted iv or v ), with a normalization factor depending on $WedgeConvention, chosen so that Int[ v, ] is an antiderivation. More generally, if alpha and beta are any alternating tensor expressions with Rank[alpha] Rank[beta], then Int[alpha,beta] represents the generalized interior product of alpha into beta (sometimes denoted ). Int[ alpha, ] is dened as the adjoint (with respect to the Hodge inner product) of wedging with alpha on the left: if alpha, beta, and gamma are alternating tensors of ranks a, b, and a + b, respectively, then <<alpha ^ beta, gamma>> = <<beta, Int [gamma]>>. alpha

3.8

Alt

If x is any tensor expression, Alt[x] is the alternating part of x. If x is already alternating, then Alt[x] = x. Note that Alt expects its argument to be a pure or mixed tensor expression. This means that you must apply Alt before you insert indices. If Alt is applied to a component expression, it does nothing, because a component expression is considered to have rank 0.

3.9

Sym

If x is any tensor expression, Sym[x] is the symmetric part of x. If x is already symmetric, then Sym[x] = x. Note that Sym expects its argument to be a pure or mixed tensor expression. This means that you must apply Sym before you insert indices. If Sym is applied to a component expression, it does nothing, because a component expression is considered to have rank 0.

4 DERIVATIVES

24

Derivatives

This chapter describes the dierentiation operators provided by Ricci. The operators Del, Div, Grad, Extd, ExtdStar, and Lie below are Ricci primitives; after some simplication, they are maintained unevaluated until you insert indices. The rest of the operators described here are immediately transformed into expressions involving the primitive operators.

4.1

Del

The basic dierentiation operator in Ricci is Del. If x is a tensor expression, Del[x] represents the total covariant derivative of x. (Del stands for the mathematical symbol .) If x is a tensor of rank k, Del[x] is a tensor of rank k + 1. If x is a section of bundle b, Del[x] is a section of the tensor product of b with its underlying (co)tangent bundle. The extra index slot resulting from dierentiation is always the last one, and is always covariant. When you insert indices into Del[x], you get components of the covariant derivative of x. For example, suppose a is a 2-tensor: In[21]:= Del[a] Out[21]= Del[a] In[22]:= % [L[i],L[j],L[k]] Out[22]= a i j ;k The operator Del can also be used to take the covariant derivative of a tensor with respect to a specic vector. If v is a 1-tensor expression, Del[v,x] represents the covariant derivative of x in the direction v; this is the contraction of v with the last index of Del[x]: In[23]:= Del[v,a] Out[23]= Del [a] v In[24]:= % [L[i],L[j]] k5 Out[24]= a i j ;k5 v

4 DERIVATIVES

25

Here v can be a contravariant or covariant 1-tensor; if it is covariant, it is converted to a vector eld by means of the metric. If u is a scalar function (i.e., a 0-tensor), then Del[v,u] is just the ordinary directional derivative of u in the direction v. If v is a contravariant basis vector such as Basis[L[i]], Ricci uses a special shorthand output form for Del[v,a]: In[25]:= Del[ Basis[L[i]], a ] Out[25]= Del [a] i WARNING: Some authors use a notation such as i ajk to refer to the (j, k, i)component of the total covariant derivative a. This is not consistent with Riccis convention; to Ricci, the expression Del [a ] i j k is the directional derivative of the component ajk in the i-direction. It is not the component of a tensor; it diers from the component ajk;i by terms involving the connection coecients.

4.2

CovD

If x is a component expression, the L[i] component of the covariant derivative of x is represented in input form either by CovD[x,{L[i]}] or by the abbreviated form x[L[i]]. Multiple covariant derivatives are indicated by CovD[x,{L[i],L[j],...}] or x[L[i],L[j],...]. If x is a scalar expression with some unlled index slots, Ricci automatically inserts indices to convert it to a component expression before taking covariant derivatives. Ricci automatically expands covariant derivatives of sums, products, and powers. For example:

4 DERIVATIVES

26

In[26]:= (1 + a[L[i]] b[U[i]]) ^ 2 i Out[26]= (1 + a i In[27]:= % [L[j]] i Out[27]= 2 (a i ;j b + a i i b k6 ) (1 + a b ) ;j k6 b ) 2

4.3

Div

Div[x] represents the divergence of the tensor expression x. This is just the covariant derivative of x contracted on its last two indices. If x is a k-tensor, Div[x] is a (k 1)-tensor. Ricci assumes that the last index of x is associated with the underlying tangent bundle of xs bundle. For example, suppose a is a 2-tensor: In[28]:= Div[a] [L[i]] k8 Out[28]= a i k8 ; Div is the formal adjoint of -Del.

4.4

Grad

If x is any tensor expression, Grad[x] is the gradient of x. It is the same as Del[x], except that the last index is converted to contravariant by means of the metric. If x is a scalar function (0-tensor), then Grad[x] is a vector eld.

4.5

Laplacian

Laplacian[x] is the covariant Laplacian of the tensor expression x. There are two common conventions in use for the Laplacian, and both are supported by Ricci. Which convention is in eect is controlled by the value of the global variable $LaplacianConvention. When $LaplacianConvention = DivGrad (the default), Laplacian[x] is automatically replaced by Div[Grad[x]]. If

4 DERIVATIVES

27

$LaplacianConvention is set to PositiveSpectrum, then Laplacian[x] is replaced by -Div[Grad[x]].

4.6

Extd

Extd[x] represents the exterior derivative of the dierential form x. In output form, Extd[x] appears as d[x]. Ricci automatically expands exterior derivatives of sums, scalar multiples, powers, and wedge products. WARNING: It is tempting to type d[x] in input form. Because the symbol d has not been dened as a Ricci function, Ricci simply returns d[x] unevaluated. You may not be able to tell immediately from Riccis output that it is not interpreting d[x] as exterior dierentiation, unless you happen to notice that it has not expanded derivatives of sums or products that appear in the argument to d.

4.7

ExtdStar

ExtdStar is the formal adjoint of Extd with respect to the Hodge inner product. It is used in constructing the Laplace-Beltrami operator, described in Section 4.8 below. If x is a dierential k-form, ExtdStar[x] is a (k 1)-form. In output form, ExtdStar appears as shown here: In[29]:= ExtdStar[alpha] * Out[29]= d [alpha]

4.8

LaplaceBeltrami

LaplaceBeltrami[x] is the Laplace-Beltrami operator = dd + d d applied to x, which is assumed to be a dierential form. It is automatically replaced by Extd[ExtdStar[x]] + ExtdStar[Extd[x]].

4.9

Lie

If v is a vector eld (i.e., a contravariant 1-tensor) and x is any tensor expression, Lie[v,x] represents the Lie derivative of x with respect to v. When v and x are both vector elds, Lie[v,x] is their Lie bracket, and Ricci puts v and x in lexical order using the skew-symmetry of the Lie bracket. Lie derivatives of dierential forms are not automatically expanded in terms of exterior derivatives and interior multiplication, because there may be some

4 DERIVATIVES

28

situations in which you do not want this transformation to take place. You can cause them to be expanded by applying the rule LieRule: In[13]:= Lie[v,beta] Out[13]= Lie [beta] v In[14]:= % /. LieRule Out[14]= d[Int [beta]] + Int [d[beta]] v v

5 SIMPLIFYING AND TRANSFORMING EXPRESSIONS

29

Simplifying and Transforming Expressions

This chapter describes commands that can be used to simplify tensor expressions, or to perform various transformations such as expanding covariant derivatives in terms of ordinary directional derivatives or vice versa. The commands in this chapter that take only one argument can be applied only to selected terms of an expression. For example, TensorSimplify[x,n] simplies only term n, leaving the rest of the expression unchanged. TensorSimplify[x,{n1,n2,n3,...}] simplies terms n1, n2, n3, . . . , combining terms if possible, but leaving the rest of x unchanged. The Reference List, Chapter 8, describes which commands accept this syntax.

5.1

TensorSimplify

The most general simplication command provided by Ricci is TensorSimplify. TensorSimplify[x] attempts to put an indexed tensor expression x into a canonical form. The rules applied by TensorSimplify may not always result in the simplest-looking expression, but two component expressions that are mathematically equal will usually be identical after applying TensorSimplify. TensorSimplify expands products and powers of tensors, uses metrics to raise and lower indices, tries to rename all dummy indices in a canonical order, and collects all terms containing the same tensor factors but dierent constant factors. When there are two or more dummy index pairs in a single term, TensorSimplify tries exchanging names of dummy indices in pairs, and chooses the lexically smallest expression that results. This algorithm may not always recognize terms that are equal after more complicated rearranging of dummy index names; an alternative command, SuperSimplify, works harder to get a canonical expression, at the cost of much slower execution. TensorSimplify makes only those simplicatons that arise from the routine changing of dummy index names and the application of symmetries and algebraic rules that the user could easily check. More complicated simplications, such as those that arise from the Bianchi identities or from commuting covariant derivatives, are done only when you request them. The basic reason for this distinction is so that you will not be confronted by a mystifyingly drastic and unxplainable simplication. To reorder covariant derivatives, that is, the ones that come after ;, use OrderCovD or CovDSimplify. To apply Bianchi identities, apply one of the rules BianchiRules, FirstBianchiRule, SecondBianchiRule, or ContractedBianchiRules. TensorSimplify calls CorrectAllVariances, TensorExpand, AbsorbMetrics, PowerSimplify, RenameDummy, OrderDummy, and CollectConstants. Any of these commands can be used individually to provide more control over the simplication process.

5 SIMPLIFYING AND TRANSFORMING EXPRESSIONS

30

5.2

SuperSimplify

SuperSimplify[x] does the same job as TensorSimplify[x], but works harder at renaming dummy index pairs to nd the lexically smallest version of the expression. If there are two or more dummy index pairs in any term of x that refer to the same bundle, SuperSimplify tries renaming the dummy indices in all possible permutations, and chooses the lexically smallest result. If there are k dummy index pairs per term, the time taken by SuperSimplify is proportional to k!, while the time taken by TensorSimplify is proportional to k 2 . Thus SuperSimplify can be very slow, especially when there are more than about 4 dummy pairs. This command should be used sparingly, only when you suspect that some terms are equal but TensorSimplify has not made them look the same.

5.3

TensorExpand

TensorExpand[x] expands products and positive integral powers in x, just as Expand does, but maintains correct dummy index conventions and does not expand constant factors. TensorExpand is called automatically by TensorSimplify.

5.4

AbsorbMetrics

AbsorbMetrics[x] causes any metric components that appear contracted with other tensors in x to be used to raise or lower indices; the metric components themselves are eliminated from the expression. For example: j k g i j In[29]:= AbsorbMetrics[%] k Out[30]= a i AbsorbMetrics is called automatically by TensorSimplify.

Out[29]= a

5.5

RenameDummy

RenameDummy[x] changes the names of dummy indices in x to standard names, choosing index names in alphabetical order from the list associated with the

5 SIMPLIFYING AND TRANSFORMING EXPRESSIONS

31

appropriate bundle, and skipping those names that already appear in x as free indices. When the list of index names is exhausted, computer-generated names of the form kn are used, where k is the last index name in the list and n is an integer. RenameDummy does not try to make the best choice of index names for the expression; use OrderDummy to do that. RenameDummy is called automatically by TensorSimplify.

5.6

OrderDummy

OrderDummy[x] attempts to put the dummy indices occurring in the tensor expression x in a canonical form. All pairs of dummy indices are ordered so that the lower member appears rst whenever possible. Then OrderDummy tries various permutations of the dummy index names in each term of x, searching for the lexically smallest version of the expression among all equivalent versions. OrderDummy has an option called Method, which controls how hard it works to nd the best arrangement of dummy index names. See the Reference List, Chapter 8, for details. OrderDummy is called automatically by TensorSimplify.

5.7

CollectConstants

CollectConstants[x] groups together terms in the tensor expression x having the same tensor factors but dierent constant factors, and performs some simplication on the constant factors. CollectConstants is called automatically by TensorSimplify.

5.8

FactorConstants

FactorConstants[x] applies the Mathematica function Factor to the constant factor in each term of x.

5.9

SimplifyConstants

SimplifyConstants[x] applies the Mathematica function Simplify to the constant factor in each term of x.

5.10

BasisExpand

If x is any tensor expression, BasisExpand[x] expands all tensors in x into component expressions multiplied by the default basis vectors and covectors. For example, suppose a is a 2-tensor:

5 SIMPLIFYING AND TRANSFORMING EXPRESSIONS

32

In[34]:= BasisExpand[2 a] k1 k2 Basis (X) Basis k1 k2 Because of the limitations of the index notation, there is no provision for expanding tensors automatically with respect to any basis other than the default one; you can usually accomplish the same thing by using DefineRule to create transformation rules between the default basis and other bases. If x is a scalar (rank 0) expression, BasisExpand[x] causes all index slots to be lled, thus converting x to a component expression.

Out[34]= 2 a

5.11

BasisGather

Sometimes you may want to convert a component expression for a tensor back to pure tensor form. BasisGather allows you to do this. For example, we can undo the eect of BasisExpand in Out[34] above: In[35]:= BasisGather[%34,a] Out[35]= 2 a The second argument to BasisGather can in principle be any tensor expression, although BasisGather may not recognize the basis expansion of complicated expressions. It works best if you apply TensorExpand (or TensorSimplify) before applying BasisGather. You can gather several tensor expressions at once by putting a list of expressions in the second argument.

5.12

CovDExpand

CovDExpand[x] expands all covariant derivatives in x into ordinary directional derivatives and connection components. Directional derivatives of components of tensors are represented by expressions like Del[ Basis[L[i]], a[L[j],L[k]] ]. For example:

5 SIMPLIFYING AND TRANSFORMING EXPRESSIONS

33

Out[35]= a j k ;i In[36]:= CovDExpand[%] k3 Out[36]= Del [a ] - a i j k k3 k Conn j i - a j k4 Conn k k4 i

Here Conn represents the coecients of the default connection with respect to the default basis. See Section 7.5, Connections, torsion, and curvature, for more information.

5.13

ProductExpand

ProductExpand[x] expands symmetric products and wedge products of 1tensors that occur in x, and rewrites them in terms of tensor products. For example, assuming $WedgeConvention = Alt, you would get the following result: i Out[5]= Basis j ^ Basis

In[6]:= %//ProductExpand 1 i Out[6]= - Basis 2 j (X) Basis 1 j - - Basis 2 i (X) Basis

You can reverse the eect of ProductExpand by applying Alt to an alternating tensor expression, or Sym to a symmetric expression; these operators replace tensor products by wedge or symmetric products.

5.14

PowerSimplify

PowerSimplify[x] performs various simplications on powers that appear in x, such as transforming a^p b^p to (a b)^p, and (a^b)^c to a^(b c) when possible, and expanding and collecting constants in expressions that appear as base or exponent of a power. PowerSimplify is called automatically by TensorSimplify.

5 SIMPLIFYING AND TRANSFORMING EXPRESSIONS

34

5.15

CorrectAllVariances

CorrectAllVariances[x] changes the variances (upper to lower or lower to upper) of all indices occurring in x whose variances are not correct for their positions, by inserting appropriate metric coecients. For example, suppose v is a vector eld (i.e., a contravariant 1-tensor): j Out[41]= v i ; k

In[42]:= CorrectAllVariances[%] j k4 Out[42]= g i k3 g v ;k4 k k3

5.16

NewDummy

NewDummy[x] replaces all dummy index pairs in x with computer-generated dummy index names of the form kn, where k is the last name in the list of indices associated with the appropriate bundle, and n is a unique integer. This is sometimes useful when you want to insert an index with the same name as a dummy name that appears in x. For example, suppose a is a 2-tensor. After applying TensorSimplify, the output from BasisExpand[a] becomes: i Basis i j If you want the (i, k)-component of this expression, you cannot just type %[L[i],L[k]], because i would then appear twice as a lower indexonce in the output expression Out[31], and once in the inserted index. The correct procedure is to use NewDummy: In[32]:= NewDummy[%] [L[i],L[k]] Out[33]= a i k j (X) Basis

Out[31]= a

5 SIMPLIFYING AND TRANSFORMING EXPRESSIONS

35

5.17

CommuteCovD

CommuteCovD[ x, L[i], L[j] ] changes all adjacent occurrences in the tensor expression x of indices L[i],L[j] (in the given order) after the ; to L[j],L[i] by adding appropriate curvature and torsion terms. For example, suppose a is a 2-tensor on a Riemannian tangent bundle (assuming $RiemannConvention = SecondUp): Out[35]= a i j ;k l In[36]:= CommuteCovD[ %, L[k], L[l] ] l1 Out[36]= a i j ;l k + a l1 j Rm i + a k l i l2 Rm j l2 k l

5.18

OrderCovD

OrderCovD[x] puts all derivative indices (those appearing after ;) in the tensor expression x into lexical order, by adding appropriate curvature and torsion terms. OrderCovD is used by CovDSimplify.

5.19

CovDSimplify

CovDSimplify[x] attempts to simplify x as much as possible, putting all dummy indices in lexical order, including those that result from dierentiation. CovDSimplify calls TensorSimplify, then OrderCovD, then TensorSimplify again. For very complicated expressions, it is probably more ecient to use the other simplication commands individually.

5.20

LowerAllIndices

LowerAllIndices[x] converts all upper indices in x (except those appearing on Basis covectors or metrics) to lower indices by inserting metrics as needed.

5.21

TensorCancel

TensorCancel[x] attempts to simplify products and quotients of tensor expressions by combining and canceling factors that are equal even though they have dierent dummy index names. For example:

5 SIMPLIFYING AND TRANSFORMING EXPRESSIONS

36

i Out[37]= (1 + u ;i u ; )

3 (1 + u ;j u

j ) ;

-1

In[38]:= TensorCancel[%] i Out[38]= (1 + u ;i u ; ) 2

6 DEFINING RELATIONS BETWEEN TENSORS

37

Dening Relations Between Tensors

Because the internal and external forms of tensors dier, you should never use the name of a tensor on the left-hand side of an assignment or rule, such as a=b, a:=b, a->b, or a:>b. Ricci provides two commands, DefineRelation and DefineRule, to create such assignments and rules. This is a common source of error. To emphasize: Never use the name of a tensor on the left-hand side of an assignment or rule; use DefineRelation or DefineRule instead.

6.1

DeneRelation

If you want to cause one tensor always to be replaced by some other tensor expression, or to be replaced only when certain indices are inserted, you can use DefineRelation. The basic syntax is shown by the following example: In[39]:= DefineRelation[ a, 2 b ] Relation defined. In[40]:= a[L[i],L[j]] Out[40]= 2 b i j This use of DefineRelation species that every occurrence of the tensor a should be replaced by 2 b. The replacement specied in such a call to DefineRelation is made whether or not indices are inserted into a. The rst argument to DefineRelation must be a single tensor, with or without indices; it cannot be a more complicated expression. (The principal reason is that Mathematica requires every relation to be associated with a specic symbol, and Riccis philosophy is to associate your denitions only with your own symbols, so that they can all be saved by Ricci. If you tried to dene a relation involving a complicated tensor expression, it would have to be associated with a Ricci or system function.) You can use DefineRule, described below, to dene transformation rules whose left-hand sides are arbitrary tensor expressions. You can cancel any relation created by DefineRelation by calling UndefineRelation. The single argument to UndefineRelation must be exactly the same as the rst argument in the corresponding call to DefineRelation. For example: In[41]:= UndefineRelation[a]

6 DEFINING RELATIONS BETWEEN TENSORS

38

Relation undefined. You can also use DefineRelation to replace a tensor only when it appears with certain indices. For example, suppose e is a 3-tensor: In[42]:= DefineRelation[ e[L[i],U[i],L[j]], v[L[j]] ] Relation defined. This tells Ricci that the indexed version of e is to be replaced by v when the rst two indices of e are contracted with each other. (Note that you do not use underscores in the dummy index names for DefineRelation, as you would in dening a relation directly with Mathematica.) The replacement takes place regardless of the actual names of the indices or whether they are upper or lower, provided the corresponding indices refer to the same bundles as i and j. For example: In[43]:= e[U[j],L[j],L[k]] Out[43]= v k In general, the relation created by DefineRelation recognizes any equivalent relations resulting from raising and lowering indices, changing index names, conjugation, and inserting extra indices, provided that all indices refer to the same bundles as the corresponding indices that were used in the call to DefineRelation. It does not recognize relations resulting from symmetries, howeveryou must dene all such equivalent relations explicitly. You can specify that the relation should be applied only under certain conditions by giving DefineRelation a third argument, which is a True/False condition that must be satised before the relation is applied. The condition may refer to the index names used in the rst argument. For example, suppose u is a scalar function on a Riemannian tangent bundle, and you would like the rst and second covariant derivatives of u always to be placed in lexical order (this is not done automatically by Ricci). You could dene the following conditional relation:

6 DEFINING RELATIONS BETWEEN TENSORS

39

In[44]:= DefineRelation[u[L[i],L[j]], u[L[j],L[i]], !IndexOrderedQ[{L[i],L[j]}]] Relation defined. In[45]:= u[L[j],L[i]] Out[45]= u ;i j

6.2

DeneRule

Sometimes you may want to dene a transformation rule that is not applied automatically, but only when you request it. Such transformations are done in Mathematica via rules. The Ricci function DefineRule allows you to dene rules whose left-hand sides involve tensors. For example, heres how to dene rules that replace the tensor a by 2 b and vice versa: In[52]:= DefineRule[ atob, a, 2 b] Rule defined. In[53]:= DefineRule[ btoa, b, a/2] Rule defined. In[54]:= a[L[i],L[j]] Out[54]= a i j In[55]:= % /. atob Out[55]= 2 b i j

In[56]:= % /. btoa Out[56]= a i j

6 DEFINING RELATIONS BETWEEN TENSORS

40

DefineRule has another advantage over DefineRelation: the left-hand side of the transformation rule dened by DefineRule can be any tensor expression, not just a single tensor with or without indices. Thus, suppose you want to convert the inner product of a and b to 0 wherever it occurs. You could dene the following rules: In[57]:= DefineRule[ ortho, Inner[a,b], 0] Rule defined. In[58]:= DefineRule[ ortho, a[L[i],L[j]] b[U[i],U[j]], 0] Rule defined. The symbol ortho is now a list containing these two transformation rules. The rst rule transforms the pure-tensor expression for <a, b> to 0, while the second rule takes care of the component expression. Heres how it works: In[60]:= a[L[j],L[k]] b[U[j],U[k]] j k b j k In[61]:= % /. ortho Out[61]= 0 Each time you call DefineRule, the new transformation rule you give is appended to a list containing ones already dened under the same name. To start over from scratch, you can include the option NewRule -> True (or just execute the assignment ortho=.). The rule created by DefineRule recognizes any equivalent substitutions resulting from raising and lowering indices and changing index names. In addition, if the second argument is a single tensor name with or without indices, it recognizes equivalent substitutions resulting from conjugation and inserting extra indices.

Out[60]= a

7 SPECIAL FEATURES

41

7
7.1

Special Features
One-dimensional bundles

One-dimensional bundles receive special treatment in Ricci. Since there is only one basis element, Ricci uses only one index name for all indices referring to the bundle. This is the only case in which an index can legitimately appear more than once as a lower or upper index in the same term. When you dene a one-dimensional bundle, you may only give one index name. When Ricci generates dummy indices for the bundle, it uses only this index name instead of generating unique names. If you dene a one-dimensional bundle with the option OrthonormalFrame -> True, the metric coecient with respect to the default basis is always equal to 1. TensorSimplify then puts all indices associated with this bundle at their natural altitude, as specied in the corresponding call to DefineTensor.

7.2

Riemannian metrics

You specify that a bundle is a tangent bundle with a Riemannian metric by including the option MetricType -> Riemannian in your call to DefineBundle. This automatically causes the torsion of the default connection to be set to 0, and denes the Riemannian, Ricci, and scalar curvature tensors. By default, these are named Rm, Rc, and Sc; you may give them any names you wish by including the DefineBundle options RiemannTensor -> name, RicciTensor -> name, and ScalarCurv -> name. If you dene more than one Riemannian tangent bundle, you must give dierent names to the curvatures on dierent bundles. On a Riemannian tangent bundle, curvature components are automatically converted to components of the Riemannian curvature tensor, and contractions of the curvature tensor are automatically converted to the Ricci or scalar curvature as appropriate. The rst and second Bianchi identities, however, are not automatically applied; you can specically apply them using the Ricci rules FirstBianchiRule, SecondBianchiRule, ContractedBianchiRules, and BianchiRules, described in the Reference List, Chapter 8. There are two common sign conventions in use for indices of the Riemannian curvature tensor. Ricci supports both; you can control which one is used for any particular Riemannian tangent bundle either by setting the global variable $RiemannConvention or by including the RiemannConvention option in the call to DefineBundle. See DefineBundle in the Reference List, Chapter 8, for details. You may convert components of the Riemannian, Ricci, and scalar curvature tensors of a Riemannian tangent bundle to connection components (Christoffel symbols) and their derivatives by applying the Ricci rule CurvToConnRule.

7 SPECIAL FEATURES

42

Connection components, in turn, can be converted to expressions involving directional derivatives of the metric components by applying ConnToMetricRule. (Note, however, that the latter rule applies only to bundles that have been given the option CommutingFrame -> True, since this is the only case in which the connection components can be expressed purely in terms of derivatives of the metric.)

7.3

Matrices and 2-tensors

Because every bundle has a default metric, any 2-tensor on a given bundle can be interpreted as a linear endomorphism of the bundle. Ricci supports various matrix computations with such tensors. The basic operation is matrix multiplication, which is indicated by Dot. For example, if a and b are 2-tensors, then a.b or Dot[a,b] represents their matrix product. If v is a 1-tensor, a.v represents the matrix a acting on the vector v. Because the metric is used automatically to raise and lower indices, no distinction is made between row and column vectors, or between covariant and contravariant tensors, for this purpose. The identity matrix on any bundle can be represented by the tensor Kronecker, whose components are the Kronecker delta symbols. Ricci automatically recognizes some basic relations involving multiplication by Kronecker, such as a.Kronecker = Kronecker.a = a. The inverse of any 2-tensor expression a is represented by Inverse[a]. When you insert indices into an inverse tensor, Ricci in eect creates a new tensor whose name is Inverse[a]. For example: In[66]:= Inverse[a] -1 Out[66]= a In[67]:= % [U[i],U[j]] -1 i j Out[67]= (a ) Ricci recognizes that multiplication by the inverse of a tensor yields the identity tensor, whether as components or as pure tensors:

7 SPECIAL FEATURES

43

In[68]:= a . b . Inverse[b] Out[68]= a In[69]:= a[L[i],L[j]] Inverse[a] [U[j],U[k]] k Out[69]= Kronecker i You can express the trace of a 2-tensor (with respect to the default metric) by Tr[a], its transpose by Transpose[a], and its determinant by Det[a]. When you insert indices into Tr[a] or Transpose[a], Ricci computes the components explicitly in terms of the components of a. For Det, however, there is no straighforward way to express the determinant of a matrix using the index conventions, so Det[a] is maintained unevaluated. Riccis dierentiation commands compute derivatives of Det[a] in terms of derivatives of components of a and Inverse[a].

7.4

Product tensors

It often happens that a tensor is associated with the tensor product of two or more bundles. In Ricci, such a tensor is called a product tensor. A product tensor is dened by giving a list of positive integers as the rank of the tensor in the DefineTensor command. The rank of the tensor is the sum of all the numbers in the list; each number in the list corresponds to a group of indices, which can have its own symmetries and can be associated to a separate bundle or direct sum of bundles. Indices can be inserted into a product tensor one group at a time: for example, if t is a product tensor whose rank is specied as {2,3}, you can insert either two or ve indices. Once the rst two indices are inserted, the resulting expression is considered to be a 3-tensor, with the symmetries and bundles associated with the second group of indices. The Bundle and Symmetries options of DefineTensor have special interpretations for product tensors. Bundle -> {bundle1,bundle2,...}: the list of bundles must be the same length as the list of ranks. Each bundle in the list is associated with the corresponding set of index slots. Other forms are Bundle -> bundle, which means that all indices are associated with the same bundle, and Bundle -> {{b1,b2},{b3,b4},...}, which means that each set of index slots is associated with a direct sum of bundles. Symmetries -> {sym1,sym2,...}: again, the list of symmetries must be the same length as the list of ranks. Each symmetry applies to the corre-

7 SPECIAL FEATURES

44

sponding set of index slots. The only other acceptable form for product tensors is Symmetries -> NoSymmetries, which is the default. One common use of product tensors is in dening arbitrary connections and their curvatures. In fact, the built-in tensors Conn and Curv are dened as product tensors. For example, the following command denes a rank-4 tensor whose rst two index slots refer to the sum of fiber and its conjugate, and whose last two refer to tangent. It is skew-symmetric in the last two indices, with no symmetries in the rst two. It is contravariant in the second index and covariant in the other three. In[62]:= DefineTensor[ omega, {2,2}, Bundle -> { {fiber,Conjugate[fiber]}, tangent }, Symmetries -> {NoSymmetries,Skew}, Variance -> {Co,Con,Co,Co} ] When you insert indices into this tensor, you may either insert the rst two indices only, or insert all four at once. (There is no provision for inserting only the last two indices.) When the rst two indices are inserted, the result is a skew-symmetric covariant 2-tensor associated with the tangent bundle. For example, assuming index names a, b, c are associated with fiber: In[5]:= omega[L[a],U[b]] b Out[5]= omega a In[6]:= %//BasisExpand b Out[6]= omega a k6 k7 k6 Basis k7 ^ Basis

Another way to use a product tensor is to represent a section of Hom(a, b), where a and b are vector bundles. Suppose you dene a tensor hom as follows: In[35]:= DefineTensor[hom, {1,1}, Bundle -> {b,a}, Variance -> {Con,Co}] Then hom is a section of b a , or equivalently a homomorphism from a to b.

7 SPECIAL FEATURES

45

7.5

Connections, torsion, and curvature

Ricci assumes that every bundle comes with a default connection that is compatible with its metric. (This assumption is forced by the limitations of the index convention, since the notation for components of covariant derivatives gives no indication what connection is being used.) The name for the default connection on any bundle is Conn. Conn is a product tensor of rank {2,1}, whose rst two indices can come from any bundle, and whose third index is associated with the underlying tangent bundle of the rst two. When you insert the rst two indices, you get an entry of the matrix of connection 1-forms corresponding to the default basis, dened by the relation: Basis i = Basis j Conn i j . When you insert the third index, you get the connection coecients (Christoel symbols) for the default connection with respect to the default basis. These are dened by: Basis k Basisi = Conn i j k Basis j . For example, suppose that indices a, b, c refer to the bundle fiber, and i, j, k refer to the underlying tangent bundle. The connection 1-forms of fiber would be expressed as: b Out[15]= Conn a In[16]:= BasisExpand[%] b Out[16]= Conn a k8 k8 Basis

Because the default metric on any bundle is assumed to be compatible with its metric, the coecients of the metric satisfy the compatibility equation d(gij ) = Conn ij + Conn ji , where the second index of Conn has been lowered by the metric. This relation is not applied automatically by Ricci, but you can force Ricci to replace derivatives of the metric by connection forms by applying the rule CompatibilityRule. A connection on the tangent bundle of a manifold may or may not have torsion. Ricci assumes that connections are torsion-free, unless you specify that a particular connection may have non-vanishing torsion by adding the option TorsionFree -> False to the DefineBundle command when the bundle is created. The name of the torsion tensor on any bundle is Tor. It is a product tensor

7 SPECIAL FEATURES

46

of rank {1,2}, whose rst index can be associated with any (subbundle of a) tangent bundle, and whose last two are associated with the full underlying tangent bundle (or bundles, if the tangent bundle is a direct sum). It is skew-symmetric in its last two indices. The torsion is dened by j Basis k k Basis j [Basis j , Basis k ] = Tor i jk Basis i . When the rst index is inserted, Tor[U[i]] represents the torsion 2-form associated with the basis 1-form Basis[U[i]]. The rst structure equation relates the exterior derivatives of basis covectors on the tangent bundle to the connection and torsion forms. Ricci does not perform this transformation automatically, but you can cause it to be performed by applying FirstStructureRule: i Out[36]= d[Basis ] In[37]:= % /. FirstStructureRule 1 i Out[37]= - Tor 2 k6 + Basis i ^ Conn k6

The 1/2 appears because, by Riccis default wedge product convention and index conventions, the basis expansion of Tor[U[i]] has to be i Tor j k j k Basis (X) Basis i Tor j k j k Basis ^ Basis

without the factor of 1/2 that appears in the standard notation used in many dierential geometry texts. The 1/2 does not appear in the output from FirstStructureRule when $WedgeConvention = Det, because in that case, i Tor j k j k Basis (X) Basis 1 i - Tor 2 j k j k Basis ^ Basis

The curvature associated with the default connection is called Curv. This is a product tensor of rank {2,2}, which is skew-symmetric in both pairs of indices (because the default connection is always assumed to be compatible with the metric). The rst two indices of Curv can refer to any bundle, and the last two refer to its underlying tangent bundle. When you insert the rst two indices, you get the matrix of curvature 2-forms associated with the standard basis. When all four indices are inserted, you get the components of the curvature tensor,

7 SPECIAL FEATURES dened by i j Basisk j i Basisk [Basis , Basis ] Basis k = Curv k l ij Basis l . i j

47

The second structure equation relates the exterior derivatives of the connection forms to curvature forms. Like the rst structure equation, it is applied only when you specically request it by applying SecondStructureRule: j Out[15]= d[Conn i ] - Conn i k ^ Conn k j

In[16]:= % /. SecondStructureRule //TensorSimplify 1 j Out[16]= - Curv 2 i As described above for Tor, the factor 1/2 appears when $WedgeConvention = Alt, but not when $WedgeConvention = Det. You can cause components of the curvature tensor Curv to be expanded in terms of components of Conn and their directional derivatives by applying the Ricci rule CurvToConnRule. You can apply all three structure equations (CompatibilityRule, FirstStructureRule, and SecondStructureRule) at once by using the rule StructureRules. If an expression involving derivatives of connection forms is converted to a component expression, the derivatives of the connection coecients are maintained in a form such as j Del [Conn k i ]

since covariant derivatives of connection coecients make no invariant sense. This is the only instance in which Ricci does not by default express components of derivatives in terms of covariant derivatives.

7.6

Non-default connections and metrics

You may want to compute covariant derivatives with respect to a connection or a metric other than the default one on a given bundle. Many of Riccis dierentiation commands accept the Connection option to indicate that covariant derivatives are to be taken with respect to a non-default connection. Ordinarily,

7 SPECIAL FEATURES

48

the value of the Connection option should be an expression of the form Conn + diff, where diff is a 3-tensor expression that represents the dierence tensor between the given connection and the default connection. For example, suppose diff is a 3-tensor: In[23]:= Del[a, Connection -> Conn + diff] (Conn + diff) Out[23]= Del [a] In[24]:= %[L[i],L[j],L[k]] k36 Out[24]= a i j ;k - a k36 j diff i - a k i k37 diff j k37 k

Notice that the components are expanded in terms of covariant derivatives of a (with respect to the default connection) and components of diff. You can also call CovD with the Connection option. For example, the following input expression results in the same output as Out[24] above: In[25]:= CovD[ a[L[i],L[j]], {L[k]}, Connection -> Conn + diff] Other operators that accept the Connection option are Div, Grad, and Laplacian. On a Riemannian tangent bundle, you can also specify that operations are to be computed with respect to a metric other than the default one. This is done by adding the Metric option to the appropriate Ricci function. For example, suppose h is a symmetric 2-tensor. To compute the covariant derivative of the tensor a with respect to the metric h, you could use In[60]:= Del[ a , Metric -> h] (h) Out[60]= Del [a] When you insert indices, the components of this expression are computed in terms of covariant derivatives of a and h (with respect to the default metric). Other dierentiation operators that accept the Metric option are Grad, Div, ExtdStar, Laplacian, LaplaceBeltrami, and CovD. Ricci can explicitly compute the Riemannian, Ricci, and scalar curvatures of a non-default metric h. These are indicated by the expressions RiemannTensor[h], RicciTensor[h], and ScalarCurv[h], respectively. When you insert indices into these, you get expressions for the curvatures in terms

7 SPECIAL FEATURES

49

of the curvatures of the default metric and covariant derivatives of h. You can also refer to the Levi-Civita connection of a non-default metric by LeviCivitaConnection[h]; this produces a 3-tensor that is suitable as an input to the Connection option of the dierentiation functions as described above.

8 REFERENCE LIST

50

8
8.1

Reference List
Ricci commands and functions

The list below describes all Ricci functions that are accessible to users. All the essential information about each function is collected here in brief form, including a list of all available options. The numbers at the right-hand sides of the function descriptions refer to section numbers in the main text where these functions are described. (Some less often used functions are documented only here in the Reference List.) For more information and examples on how to use the most common commands, see the main text. For descriptions of Riccis global variables (symbols starting with $), see the end of this chapter. AbsorbMetrics 5.4

AbsorbMetrics[x] simplies x by eliminating any metrics that are contracted with other tensors, and using them instead to raise or lower indices. AbsorbMetrics[x,n] or AbsorbMetrics[x,{n1,n2,...}] applies AbsorbMetrics only to term n or to terms n1,n2,... of x. Option: Mode -> All or NoOneDims. NoOneDims indicates that metrics associated with one-dimensional bundles should not be absorbed, unless they are paired with other metrics. This option exists mainly for internal use by TensorSimplify, so that it can control the order in which dierent metrics are absorbed. Default is All. See also TensorSimplify, LowerAllIndices, CorrectAllVariances. Alt 3.8 Alt[x] is the alternating part of the tensor expression x. If x is alternating, then Alt[x] = x. Alt is also a value for the global variable $WedgeConvention. See also Sym, Wedge, $WedgeConvention. Alternating Alternating is a value for the Symmetries option of DefineTensor. Any 2.1 Any can be used as a value for the Bundle option of the DefineTensor command. Bundle -> Any means that indices from any bundle can be inserted. 2.4

8 REFERENCE LIST

51

Bar

2.2, 2.4 Bar is an internal symbol used by Ricci for representing conjugates of tensors and indices. The internal form for a barred index is L[i[Bar]] or U[i[Bar]]. In input form, these can be abbreviated LB[i] and UB[i]. The internal form for the conjugate of a tensor is Tensor[name[Bar],{...},{...}]. In input form, this is typed Conjugate[name] [...] [...]. See also Tensor, Conjugate, L, U, LB, UB.

Basis

2.4, 5.10

Basis is the name used for generic basis vectors and covectors for any bundle. Basis vectors are generated automatically by BasisExpand. For example, if index name i is associated with bundle b, then Basis[L[i]] and Basis[U[i]] represent contravariant and covariant basis elements for b (i.e., basis elements for b and its dual), respectively; Basis[LB[i]] and Basis[UB[i]] represent basis elements for the conjugate bundle Conjugate[b] and its dual. Basis is dened by Ricci as a product tensor of rank {1,1}, with the special option Bundle -> Same: this means that the rst index can be associated with any bundle, but when a second index is inserted, it must be associated with the same bundle as the rst, or else the result is 0. If you insert an index into Basis[L[i]] or Basis[U[i]], you get Kronecker delta coecients, metric coecients (which result from lowering or raising an index on the Kronecker delta), or 0 as appropriate. See also BasisExpand, BasisName, CoBasisName, Kronecker. BasisExpand 2.4, 5.10

BasisExpand[x] expands all tensors appearing in x into component expressions multiplied by basis vectors and covectors, Basis[L[i]] and Basis[U[j]]. BasisExpand[x,n] or BasisExpand[x,{n1,n2,...}] applies BasisExpand only to term n or to terms n1,n2,... of x. BasisExpand also inserts indices into all unlled index slots, so it can be used to turn a scalar tensor expression into a component expression. See also Basis, BasisGather, CovDExpand, ProductExpand. BasisGather 5.10

BasisGather[x,y] attempts to recognize the basis expression for y in x, and replaces it by y. BasisGather[ x, {y1,y2,...} ] does the same thing for several expressions at once. In eect, BasisGather is a partial inverse for BasisExpand. See also BasisExpand. BasisName 2.4

BasisName -> name is an option for DefineBundle. It species the name to be used in input and output form for the basis vectors for the bundle being dened.

8 REFERENCE LIST

52

BianchiRules

7.2

BianchiRules[L[i],L[j],L[k]] is a rule that converts Riemannian curvature tensors containing the indices L[i], L[j], L[k] in order to two-term sums, using the rst and second Bianchi identities. The three indices may be upper indices, lower indices, or a combination. The rst Bianchi rule is applied to any tensor with with Symmetries -> RiemannSymmetries; the second is applied only to tensors that have been dened as Riemannian curvature tensors by DefineBundle. examples: i Out[16]= Rm j k l In[17]:= % /. BianchiRules[L[j],L[k],L[l]] i Out[17]= Rm k j l m Out[20]= Rm i j k l ; In[21]:= %/.BianchiRules[L[k],L[l],U[m]] m Out[21]= Rm i j k ;l - Rm i j l m ;k SecondBianchiRule, 2.4 i - Rm l j k

See also DefineBundle, ContractedBianchiRules. Bundle

FirstBianchiRule,

Bundle -> name is a DefineTensor option, specifying the name of the bundle or bundles the tensor is to be associated with. The function Bundle[i] returns the name of the bundle associated with the index i. BundleDummy BundleDummy[bundle] returns the symbol that is used for computergenerated dummy indices associated with the bundle. This is the last index name in the list BundleIndices[bundle]. For most bundles, new

8 REFERENCE LIST

53

dummy indices are of the form kn, where k=BundleDummy[bundle] and n is an integer. For one-dimensional bundles, n is omitted. See also BundleIndices, DefineBundle. BundleIndices BundleIndices[bundle] returns a list of the index names currently associated with bundle. See also BundleDummy, DefineBundle. BundleQ BundleQ[x] returns True if x is the name of a bundle, and False otherwise. See also DefineBundle. Bundles The function Bundles[x] returns a list of the bundles associated with each index position in the tensor expression x; each entry in the list is a sublist giving the allowed bundles for that index position. If you insert an index that is not associated with one of the bundles corresponding to that index position, the expression gives 0. See also DefineTensor, Variance. Co Co is an abbreviation for Covariant. CoBasisName 2.4 2.4

CoBasisName -> name is an option for DefineBundle. It species the name to be used in input and output form for the basis covectors for the bundle being dened. CollectConstants 5.7

CollectConstants[x] groups together terms in the tensor expression x having the same tensor factors but dierent constant factors. CollectConstants[x,n] or CollectConstants[x,{n1,n2,...}] applies CollectConstants only to term n or to terms n1,n2,... of x. The constant factors are simplied somewhat: if a constant factor is not too complicated (LeafCount[c] 50), Ricci factors the constant. For larger constant expressions, it applies the Mathematica function Together. See also TensorSimplify, FactorConstants, SimplifyConstants, ConstantFactor, TensorFactor. CommuteCovD 5.17

CommuteCovD[ x, L[i], L[j] ] changes all adjacent occurrences of indices L[i],L[j] after the ; to L[j],L[i] by adding appropriate curvature and torsion terms. The second and third arguments may be upper or lower indices. See also OrderCovD, CovDSimplify.

8 REFERENCE LIST

54

CommutingFrame CommutingFrame is a DefineBundle option. Default is True. CompatibilityRule

2.1

7.5

CompatibilityRule is a rule that transforms derivatives of metric components into connection coecients, using the fact that the default connection is compatible with the metric, which is expressed by d(gij ) = Conn ij + Conn ji . See also FirstStructureRule, SecondStructureRule, StructureRules. Complex 2.1, 2.3, 2.4, 2.6

Ricci recognizes Complex as a value for the Type option of DefineConstant, DefineTensor, DefineBundle, and DefineMathFunction. Con Con is an abbreviation for Contravariant. Conjugate 2.1, 2.2, 2.4, 2.6 2.4

The Ricci package modies the Mathematica function Conjugate to handle tensor expressions and indices. The behavior of a tensor, constant, index, bundle, or mathematical function under conjugation is determined by the Type option specied when the object is dened. In input form, you indicate the conjugate of tensor t by typing Conjugate[t]; you may follow this with one or more sets of indices in brackets, as in Conjugate[t] [L[i],L[j]]. You indicate the conjugate of an index L[i] or U[i] by typing LB[i] or UB[i]. To indicate the conjugate of bundle b, type Conjugate[b]. You can compute the conjugate of any tensor expression x by typing Conjugate[x]. In output form, conjugates of tensors, bundles, and indices appear as barred symbols. The behavior of tensors and indices under conjugation is determined by the Type options specied when the tensor and its associated bundles are dened. For example, assume a, b, and c are 2-tensors of types Real, Complex, and Imaginary, respectively. Assume i is an index associated with a real bundle, and p is an index associated with a complex bundle:

8 REFERENCE LIST

55

Conjugate[a ] = i p Conjugate[b ] = i p

a _ b

_ i p _ i p

Conjugate[c ] = -c _ i p i p See also Re, Im, DefineTensor, DefineConstant, DefineBundle, Declare. Conn 7.5, 7.6 Conn is the name used for the generic connection forms for the default connection on any bundle. It is dened as a product tensor of rank {2,1} and variance {Covariant, Contravariant, Covariant}. When the rst two indices are inserted, it represents an entry of the matrix of connection 1-forms associated with the default basis. When all three indices are inserted, it represents the Christoel symbols of the default connection relative to the default basis. The upper index is the second one, and the index representing the direction in which the derivative is taken is the third one. Thus the Christoel symbol k is represented by ij k Conn [L[j],U[k],L[i]] --> Conn j i

See also Curv, Tor, Curvature, Connection, LeviCivitaConnection, CovDExpand, ConnToMetricRule, CurvToConnRule. Connection 7.6

Connection -> cn is an option for Del, CovD, Div, Grad, and Laplacian. It species that covariant derivatives are to be taken with respect to the connection cn instead of the default connection. ConnToMetricRule 7.2

ConnToMetricRule is a rule that causes indexed components of the default connection Conn on a Riemannian tangent bundle to be converted to expressions involving directional derivatives of the metric coecients. This rule applies only to fully-indexed connection components whose indices all refer to a single Riemannian tangent bundle with the option CommutingFrame -> True. See also CurvToConnRule, Conn, DefineBundle.

8 REFERENCE LIST

56

ConstantFactor ConstantFactor[x] returns the product of all the constant factors in x, which should be a monomial. See also TensorFactor, CollectConstants. ConstantQ 2.3

ConstantQ[x] returns True if there are no explicit tensors in x, and False otherwise. See also DefineConstant. ContractedBianchiRules 7.2

ContractedBianchiRules is a rule that simplies contracted covariant derivatives of the Riemannian and Ricci curvature tensors using contracted versions of the second Bianchi identity. For example (assuming $RiemannConvention = SecondUp, which is the default), l Out[3]= Rm i j k l ; In[4]:= % /. ContractedBianchiRules Out[4]= -Rc i k ;j Similarly, j Out[5]= Rc i j ; In[6]:= % /. ContractedBianchiRules 1 Out[6]= - Sc 2 ;i See also BianchiRules, FirstBianchiRule, SecondBianchiRule. Contravariant 2.4 + Rc j k ;i

Contravariant is a value for the Variance option of DefineTensor, and may be abbreviated Con. If an index slot is Contravariant, indices in that slot are upper by default. See also DefineTensor. CorrectAllVariances 5.15

CorrectAllVariances[x] changes the variances (upper to lower or

8 REFERENCE LIST

57

lower to upper) of indices in x whose variances are not correct for their positions, by inserting appropriate metric coecients. CorrectAllVariances[x,n] or CorrectAllVariances[x,{n1,n2,...}] applies CorrectAllVariances only to term n or to terms n1,n2,... of x. Option: Mode -> All or OneDims. OneDims indicates that only onedimensional indices and indices that appear inside of dierential operators such as Del or Extd should be corrected. This option exists mainly for internal use by TensorSimplify. Default is All. See also TensorSimplify, LowerAllIndices, AbsorbMetrics. Covariant 2.4

Covariant is a value for the Variance option of DefineTensor, and may be abbreviated Co. If an index slot is Covariant, indices in that slot are lower by default. See also DefineTensor. CovD 4.2 If x is a rank-0 tensor expression, then CovD[ x, {L[i],L[j]} ] is the component of the covariant derivative of x in the L[i],L[j] directions. If x has any unlled index slots, then it is rst converted to a component expression by generating dummy indices if necessary before the covariant derivative is computed. The shorthand input form x[ L[i], L[j] ] is automatically converted to CovD[ x, {L[i],L[j]} ] if x has rank 0. In output form, covariant derivatives of a tensor are represented by subscripts following a semicolon. Options (for the explicit form CovD[x,{L[i],L[j]}] only): Connection -> cn species that covariant derivatives are to be taken with respect to the connection cn instead of the default connection. Ordinarily, cn should be an expression of the form Conn + diff, where diff is a 3-tensor expression representing the dierence tensor between the default connection and cn. Metric -> g: a symmetric 2-tensor expression, indicating that the covariant derivatives are to be taken with respect to the Levi-Civita connection of g instead of the default metric for xs bundle (which is assumed to be Riemannian). See also Del, InsertIndices, CovDExpand. CovDExpand 5.12

CovDExpand[x] converts all components of covariant derivatives in x to ordinary directional derivatives and connection coecients. CovDExpand[x,n] or CovDExpand[x,{n1,n2,...}] applies CovDExpand

8 REFERENCE LIST

58

only to term n or to terms n1,n2,... of x. Directional derivatives with respect to basis elements are represented by Del[ Basis[L[i]],...]. For example: k1 v // CovDExpand --> Del [v ] - Conn i ;j j i j v i k1

Ordinary directional derivatives can be converted back to covariant derivatives by calling BasisExpand. See also BasisExpand, Del, Conn. CovDSimplify 5.19

CovDSimplify[x] attempts to simplify x as much as possible by ordering all dummy indices, including those that occur after the ; in component expressions. CovDSimplify[x,n] or CovDSimplify[x,{n1,n2,...}] applies CovDSimplify only to term n or to terms n1,n2,... of x. CovDSimplify rst calls TensorSimplify, then OrderCovD, then TensorSimplify again. See also OrderCovD, CommuteCovD, TensorSimplify, SuperSimplify. Curv 7.5, 7.6 Curv is the generic curvature tensor associated with the default connection on any bundle. It is generated when covariant derivatives are commuted, and when SecondStructureRule is applied to derivatives of connection forms. Curv is a product tensor with rank {2,2}, which is Alternating in its rst two and last two indices. Inserting the rst two indices yields the matrix of curvature 2-forms associated with the default basis. Inserting all four indices yields the coecients of the curvature tensor. See also Conn, Curvature, CurvToConnRule. Curvature 7.5

Curvature[cn] is the curvature tensor associated to the connection cn. Ordinarily, cn should be an expression of the form Conn + diff, where diff is a 3-tensor expression representing the dierence tensor between the default connection and cn. See also Conn, Curv, CurvToConnRule. CurvToConnRule 7.2, 7.5

CurvToConnRule is a rule that converts components of curvature tensors (including Riemann, Ricci, and scalar curvature tensors for Riemannian tangent bundles) to expressions involving connection coecients and their covariant derivatives. It is a partial inverse for SecondStructureRule. See also Curv, Conn, ConnToMetricRule, SecondStructureRule. Declare 2.1, 2.3, 2.4

Declare[ name, options ] can be used to change certain options

8 REFERENCE LIST

59

for previously-dened Ricci objects. Declare[ {name1,name2,...}, options ] changes options for several names at once; all names in the list are given the same options. The rst argument, name, must be the name of a bundle, constant, index, tensor, or mathematical function that was previously dened by a call to DefineBundle, DefineConstant, DefineIndex, DefineTensor, or DefineMathFunction. The allowable options depend on what type of object is given as the rst argument; the meanings of the options are the same as in the corresponding Define command. Any options not listed explicitly in the call to Declare are left unchanged. Constant: If name is a symbolic constant, only the Type option is allowed. It overrides any Type specication given when DefineConstant was called. Index: If name is an index name, the only allowable option is TeXFormat. Bundle: If name is a bundle name, the allowable options are FlatConnection, ParallelFrame, OrthonormalFrame, PositiveDefinite, CommutingFrame, and TorsionFree. Tensor: If name is a tensor name, the allowable options are Type, TeXFormat, and Bundle. Math function: If name is the name of a mathematical function, the only allowable option is Type. See also DefineBundle, DefineConstant, DefineIndex, DefineTensor, DefineMathFunction. DefineBundle 2.1, 7.1

DefineBundle[ name, dim, metric, {indices} ] denes a bundle. name: A symbol that uniquely identies the bundle. dim: The dimension of the bundle. Can be a positive integer, a symbolic constant, or any constant expression. metric: A name for the bundles metric. The metric is automatically dened by DefineBundle as a Symmetric or Hermitian 2-tensor. It is assumed to be compatible with the bundles default connection. {indices}: A list of index names to be associated with the bundle. If only one index name is given, the braces may be omitted. For onedimensional bundles, only one index name is allowed. For higherdimensional bundles, additional indices can be added to the list at any time by calling DefineIndex. Options: Type -> Complex or Real. Default is Real. If Complex, the conjugate bundle is referred to as Conjugate[name], and is associated with barred indices.

8 REFERENCE LIST

60

TangentBundle -> bundle. Species the name of the underlying tangent bundle for the bundle being dened. This can be a list of bundles, representing the direct sum of the bundles in the list. The default is the value of $DefaultTangentBundle if dened, otherwise this bundle itself (or the direct sum of this bundle and its conjugate if the bundle is complex). If $DefaultTangentBundle has not been given a value the rst time DefineBundle is called, it is set to the rst bundle dened (and its conjugate if it is a complex bundle). FlatConnection -> True or False. True means the default connection on this bundle has curvature equal to 0. The default is False. ParallelFrame -> True or False. True means the default basis elements have vanishing covariant derivatives, so the connection and curvature forms are set equal to 0. This option forces FlatConnection -> True. The default is False. OrthonormalFrame -> True or False. True means the metric coecients with respect to the default basis are always assumed to be constants. For one-dimensional bundles, the metric coecient is explicitly set to 1. The default is False. CommutingFrame -> True or False. This option is meaningful only for tangent bundles or their subbundles. True means the default basis for this bundle consists of commuting vector elds (as is the case, for example, when coordinate vector elds are used). This option causes the connection coecients for this bundle to be symmetric in their two lower indices, and Lie brackets of basis vectors to be replaced by zero. PositiveDefinite -> True or False. True means the bundles metric is positive denite. False means no assumption is made about the sign of the metric. This aects the simplication of complex powers and logarithms of inner products with respect to the metric. The default is True. TorsionFree -> True or False. False means that the default connection on this bundle may have nonvanishing torsion; this makes sense only for tangent bundles or their subbundles. The default is True. BasisName -> name. Species a name to be used in input, output, and TEX forms for the default basis vectors for this bundle. The default name is Basis. Basis vectors are always represented internally by the name Basis. Covariant basis vectors can be named independently by the CoBasisName option. CoBasisName -> name. Species a name to be used in input, output, and TEX forms for the default covariant basis vectors for this bundle. The default is the name given in the BasisName option, or Basis if BasisName is also omitted. Basis covectors are always represented internally by the name Basis.

8 REFERENCE LIST

61

MetricType -> Riemannian or Normal. The default is Normal, which means that the bundles metric has no special properties. If Riemannian is specied, this bundle is assumed to be a Riemannian tangent bundle, and DefineBundle automatically denes the Riemannian, Ricci, and scalar curvature tensors. Contractions of the Riemannian and Ricci curvatures are automatically converted to Ricci and scalar curvatures, respectively. To apply other identities, use FirstBianchiRule, SecondBianchiRule, ContractedBianchiRules, or BianchiRules. When you specify MetricType -> Riemannian, you may also give the RiemannTensor, RicciTensor, ScalarCurv, and RiemannConvention options below. RiemannTensor -> name: a name to be given to the Riemannian curvature tensor for this bundle. This tensor is automatically dened as a real, covariant 4-tensor with Symmetries -> RiemannSymmetries. This option takes eect only if MetricType -> Riemannian is specied. The default is Rm. RicciTensor -> name: a name to be given to the Ricci tensor for this bundle. This tensor is automatically dened as a real, covariant, symmetric 2-tensor. This option takes eect only if MetricType -> Riemannian is specied. The default is Rc. ScalarCurv -> name: a name to be given to the scalar curvature function for this bundle. This tensor is automatically dened as a real 0-tensor. This option takes eect only if MetricType -> Riemannian is specied. The default is Sc. RiemannConvention -> SecondUp or FirstUp: determines the sign convention used for the Riemannian curvature tensor for this bundle. This option takes eect only if MetricType -> Riemannian is specied. The default is SecondUp, or the value of the global variable $RiemannConvention if it has been set. The meaning of this option is indicated by the relations below. (Note that the Riemann curvature tensor is always dened as a covariant 4-tensor, so all four of its indices are normally down. It is done this way so that all index positions are equivalent for the purpose of applying symmetries of the curvature tensor. The RiemannConvention option controls only the sign of the curvature components.) RiemannConvention -> FirstUp: (X Y Y X [X,Y ] )Z = X i Y j Z k g Rm
k ikj m

Rm mkij Basis

= Rcij
m

RiemannConvention -> SecondUp: (X Y Y X [X,Y ] )Z = X i Y j Z k g Rm i k kj = Rcij Rm kmij Basis

8 REFERENCE LIST

62

Quiet -> True or False. Default is False, or the value of $Quiet if set. See also DefineTensor, DefineIndex, UndefineBundle, Declare. DefineConstant 2.3

DefineConstant[c] denes the symbol c to be a constant with respect to dierentiation in all space variables. The notation c[ L[i] ] is always interpreted as a covariant derivative of c, and thus 0; and Conjugate[c] is replaced by c if the constant is Real, and by -c if it is Imaginary. Options: Type -> type. This can be a single keyword or a list of keywords, chosen from among Complex, Real, Imaginary, Positive, Negative, NonPositive, NonNegative, Integer, Odd, or Even. Determines how the constant behaves under conjugation, exponentiation, and logarithms. Default is Real. Quiet -> True or False. Default is False, or the value of $Quiet if set. See also Conjugate, UndefineConstant, Declare. DefineIndex 2.1, 2.2

DefineIndex[ {i,j,k}, bundle ] causes the index names i, j, k to be associated with bundle. The rst argument must be a symbol or list of symbols. One or more of these may have already been dened as indices, as long as they are associated with the same bundle as specied in this command. The second argument must be the name of a bundle. Options: TeXFormat -> "texformat". Determines how the index appears in TeXForm output. Default is the index name itself. This should be specied only when one single index is being dened. Note that, as in all Mathematica strings, backslashes must be typed as double backslashes; thus to get an index name to appear as in TEX output, you would type TeXFormat -> "\\beta". Quiet -> True or False. Default is False, or the value of $Quiet if set. See also DefineBundle, UndefineIndex, UndefineBundle, Declare.

8 REFERENCE LIST

63

DefineMathFunction

2.5, 2.6

DefineMathFunction[f] declares f to be a scalar-valued function of one real or complex variable that can be used in tensor expressions. If x is any rank-0 tensor expression, then f[x] is also interpreted as a rank-0 tensor expression. Ricci has built-in denitions for the Mathematica functions Sin, Cos, Tan, Sec, Csc, Cot, and Log. If you intend to use any other functions in tensor expressions, you must call DefineMathFunction. Limitation: Currently, Ricci has no provision for using mathematical functions depending on more than one variable in tensor expressions. Options: Type -> type. This can be a single keyword or a list of keywords, chosen from among Real, Complex, Imaginary, Automatic, Positive, Negative, NonPositive, or NonNegative. Real means f[x] is always real; Imaginary means f[x] is always imaginary; Complex means f[x] can assume arbitrary complex values; and Automatic means Conjugate[f[x]] = f[Conjugate[x]]. Positive, Negative, NonPositive, and NonNegative all imply Real, and mean that the function values always have the corresponding attribute. Default is Real. Quiet -> True or False. Default is False, or the value of $Quiet if set. See also Declare. DefineRelation 6.1

DefineRelation[ tensor, x, condition ] denes a relation of the form tensor := x /; condition associated with the tensors name. The rst argument tensor must be a single tensor name with or without indices, and can be written in Riccis input form. If index names used in tensor are associated with bundles, then the relation is applied only when indices from those bundles appear in those positions. The condition argument is optional; if present, it should be a true/false condition that must be satised before the relation will be applied. It may involve the tensor name or indices that appear in tensor. For example, if i is associated with bundle a and p with bundle b, DefineRelation[ t[ L[i],L[p] ], 0 ]

8 REFERENCE LIST

64

species that t[ , ] is to be replaced by 0 whenever its rst index is associated with a and its second with b. In general, the relation substitutes x (suitably modied) in place of any expression that matches tensor after insertion of indices, renaming of indices, covariant dierentiation, conjugation, or raising or lowering of indices. DefineRelation has the HoldAll attribute, so its arguments will not be evaluated when the command is executed. This means you must write out tensor, x, and condition in full, rather than referring to a variable name or a preceding output line such as Out[10], so that DefineRelation can properly process the indices that appear in the arguments. Option: Quiet -> True or False. Default is False, or the value of $Quiet if set. See also DefineRule, UndefineRelation. DefineRule 6.2

DefineRule[ rulename, lhs, rhs, condition ] denes a rule named rulename of the form lhs :> rhs /; condition The rst two arguments, lhs and rhs, are arbitrary tensor expressions with or without indices; tensors in them can be written in Riccis input form (tensor name, followed optionally by one or more sets of indices in brackets). Other expressions used in lhs must match Mathematicas internal form for such expressions. The condition argument is optional; if present, it should be a true/false condition that must be satised before the relation will be applied. It may involve the tensor name or indices that appear in lhs. The rule is appended to a list containing previous rules dened with the same name, unless the option NewRule -> True is specied. To delete a rule entirely, make the assignment rulename = . If index names used in lhs are associated with bundles, then the rule is applied only when indices from those bundles appear in those positions. For example, if i is associated with bundle a and p with bundle b, DefineRule[ zerorule, t[ L[i], L[p] ], 0 ]

8 REFERENCE LIST

65

causes rule zerorule to replace t[ , ] by 0 whenever its rst index is associated with a and its second with b. In general, the rule substitutes rhs (suitably modied) in place of any expression that matches lhs after renaming of indices or raising or lowering of indices. If lhs is a single tensor with or without indices, then the rule is also applied to expressions that match lhs after insertion of indices, covariant dierentiation, or conjugation. DefineRule has the HoldAll attribute, so its arguments will not be evaluated when the command is executed. This means you must write out lhs, rhs, and condition in full, rather than referring to a variable name or a preceding output line such as Out[10], so that DefineRule can properly process the indices that appear in the arguments. Options: Quiet -> True or False. Default is False, or the value of $Quiet if set. NewRule -> True or False. Default is False. See also DefineRelation. DefineTensor DefineTensor[ name, rank ] denes a tensor. name: A symbol that uniquely identies the tensor. rank: Rank of the tensor. For ordinary tensors, this must be a non-negative integer. For product tensors, this is a list of positive integers. Options: Symmetries -> sym: Symmetries of the tensor. For ordinary tensors, the allowed values are Symmetric, Alternating or Skew, Hermitian, SkewHermitian, RiemannSymmetries, NoSymmetries, or the name of an arbitrary symmetry dened by calling DefineTensorSymmetries. For a product tensor, the value of the Symmetries option must be either NoSymmetries or a list of symmetries whose length is the same as that of the list of ranks; each symmetry on the list can be Symmetric, Alternating or Skew, Hermitian, SkewHermitian, or NoSymmetries, and applies to the corresponding group of indices in the list of ranks. Default is NoSymmetries. The values of the Symmetries option have the following meanings: Symmetric: The tensor is fully symmetric in all indices. Alternating: The tensor changes sign if any two indices are interchanged. Skew is a synonym for Alternating. 2.4, 7.4

8 REFERENCE LIST

66

Hermitian: The tensor must be a real 2-tensor associated with a bundle and its conjugate. It is symmetric in its two indices, with the additional property that t[L[i],L[j]] and t[LB[i],LB[j]] are both zero whenever i,j are associated with the same bundle. For product tensors, this can be used only for a group of two indices. SkewHermitian: The tensor must be a real 2-tensor associated with a bundle and its conjugate. It is skew-symmetric in its two indices, with the additional property that t[L[i],L[j]] and t[LB[i],LB[j]] are both zero whenever i,j are associated with the same bundle. For product tensors, this can be used only for a group of two indices. RiemannSymmetries: The tensor must have rank 4. The symmetries are those of the Riemann curvature tensor: skew-symmetric in the rst two indices, skew-symmetric in the last two indices, symmetric when the rst two and last two are interchanged. The (rst) Bianchi identity is not automatically applied, but can be explicitly applied to any tensor with these symmetries by means of FirstBianchiRule or BianchiRules. Type -> type. For rank-0 tensors, this can be a single keyword or a list of keywords, chosen from among Complex, Real, Imaginary, Positive, Negative, NonPositive, or NonNegative. For higherrank tensors, only Real, Complex, or Imaginary is allowed. Determines how the tensor behaves under conjugation, exponentiation, and logarithms. The default is always Real. TeXFormat -> "texformat": Determines how the tensor appears in TeXForm output. Default is the tensors name. Note that, as in all Mathematica strings, backslashes must be typed as double backslashes; thus to get a tensors name to appear as in TEX output, you would type TeXFormat -> "\\beta". Bundle -> bundle: The bundle with which the tensor is associated. This can be a list, meaning the tensor is associated with the direct sum of all the bundles in the list. If a complex bundle is named and the Type option is not Complex, the tensor is automatically associated with the direct sum of the bundle and its conjugate. Bundle -> Any indicates that the tensor can accept indices from any bundle. The default value for the Bundle option is $DefaultTangentBundle. If the tensor is given an index associated with a bundle not in this list, the result is 0. For a product tensor, the Bundle option can be either a single bundle, meaning that all index positions are associated with the same bundle, or a list whose length is the same length as the list of ranks. Each entry in the list species the bundle for the corresponding index slots. Each entry can be either a single bundle name or a list of bundle names indicating a direct sum of bundles for that group of slots.

8 REFERENCE LIST

67

Bundle -> Same is a special option for product tensors. It means that indices from any bundle can be inserted in the rst set of index slots, but indices in the remaining slots must be from the same bundle or its conjugate. This option is used internally by Ricci in dening the Basis tensor. Bundle -> Automatic is another special form that can be used for product tensors. It means that indices from any bundle can be inserted in the rst set of index slots, but indices in the remaining slots must come from the underlying tangent bundle of the rst indices. This option is used internally by Ricci in dening the Conn, Tor, and Curv tensors. Variance -> Covariant or Contravariant: Species whether the tensor is covariant (so that components have lower indices) or contravariant (upper indices). This can be a list whose length is equal to the total rank of the tensor, in which case each entry in the list species the variance of the corresponding index slot. The default is Covariant. Values may be abbreviated Co and Con. Quiet -> True or False: Default is False, or the value of $Quiet if set. See also UndefineTensor, Declare, Conjugate, DefineBundle, FirstBianchiRule, BianchiRules, Conn, Tor, Curv, Basis, DefineTensorSymmetries. DefineTensorSymmetries DefineTensorSymmetries[ name, {perm1,sgn1,...,permN,sgnN} ] denes a TensorSymmetry object that can be used in DefineTensor. Here perm1,...,permN must be lists that are nontrivial permutations of {1,2,...,k}, and sgn1,...,sgnN must be constants (ordinarily 1). A tensor with this symmetry has the property that, if any of the permutations perm1,...,permN is applied to its indices, the value of the tensor is multiplied by the corresponding constant sgn1,...,sgnN. Ricci automatically tries all non-trivial iterates of each specied permutation, until the indices are as close as possible to lexical order. (However, Ricci does not try composing dierent permutations in the list.) See also TensorSymmetry, DefineTensor, UndefineTensorSymmetries. Del 4.1 Del[x] is the total covariant derivative of the tensor expression x. If x is a k-tensor, then Del[x] is a (k + 1)-tensor. The last index slot of Del[x] is the one generated by dierentiation; it is always covariant, and is associated with the underlying tangent bundle of the bundle(s) of x. If x is a scalar function (0-tensor), then Del[x] is automatically replaced by Extd[x].

8 REFERENCE LIST

68

Del[v,x] is the covariant derivative of x in the direction v; v must be a 1-tensor expression. In output form, Del[v,x] appears as: Del [x] v Covariant derivatives with respect to (contravariant) basis vectors are displayed without the name of the basis vector. For example, Del[ Basis[L[i]], x ] appears as Del [x] i Options: Connection -> cn species that the covariant derivative is to be taken with respect to the connection cn instead of the default connection. Ordinarily, cn should be an expression of the form Conn + diff, where diff is a 3-tensor expression representing the dierence tensor between the default connection and cn. Metric -> g: a symmetric 2-tensor expression, indicating that the covariant derivative is to be taken with respect to the Levi-Civita connection of g instead of the default metric for xs bundle (which is assumed to be Riemannian). See also Grad, CovD, Extd, Div, Laplacian, CovDExpand. Det 7.3 If x is a 2-tensor expression, Ricci interprets Det[x] as the determinant of x, using the metric to convert x to a {Contravariant,Covariant} tensor if necessary. Det is also a value for the global variable $WedgeConvention. See also Dot, $WedgeConvention. Dimension Dimension[bundle] returns the dimension of bundle. DefineBundle. Div See also 4.3 Div[x] is the divergence of the tensor expression x, which is the covariant derivative of x contracted on its last two indices. If x is a k-tensor, Div[x] is a (k 1)-tensor. Ricci assumes (without checking) that the last index of x is associated with the underlying tangent bundle of xs bundle. Div is the formal adjoint of -Del. Options:

8 REFERENCE LIST

69

Connection -> cn species that the divergence is to be taken with respect to the connection cn instead of the default connection. Ordinarily, cn should be an expression of the form Conn + diff, where diff is a 3-tensor expression representing the dierence tensor between the default connection and cn. Metric -> g: a symmetric 2-tensor expression, indicating that the divergence is to be taken with respect to the Levi-Civita connection of g instead of the default metric for xs bundle (which is assumed to be Riemannian). See also Del, Laplacian. DivGrad DivGrad is a value for the global variable $LaplacianConvention. Dot 3.4, 7.3 When x and y are tensor expressions, x.y or Dot[x,y] represents the contraction of the last index of x with the rst index of y. When x and y are 2-tensors, this can be interpreted as matrix multiplication. If Dot is applied to three or more tensor expressions, all arguments except possibly the rst and last must have rank greater than 1. Ricci automatically expands dot products of sums and scalar multiples, and replaces dot products of 1-tensors with inner products. See also Inner, Inverse, Kronecker. ERROR If you attempt to insert the wrong number of indices into a tensor expression, Ricci returns ERROR[expression]. Even Even is a value for the Type option of DefineConstant. Expand 5.3 2.3 4.5

If you apply the Mathematica function Expand to an expression that includes tensors with indices, Ricci automatically converts it to TensorExpand. See also TensorExpand. Extd 4.6 Extd[x] is the exterior derivative of x, which must be an alternating covariant tensor expression. In output form, Extd[x] appears as d[x]. Ricci automatically expands exterior derivatives of sums, scalar multiples, powers, and wedge products. See also Del, ExtdStar, LaplaceBeltrami. ExtdStar 4.7

ExtdStar[x] is the adjoint of the operator Extd applied to x, which must be an alternating covariant tensor expression. If x is a k-form, then

8 REFERENCE LIST

70

ExtdStar[x] is a (k 1)-form. In output form, ExtdStar[x] appears as shown here: In[26]:= ExtdStar[x] * Out[26]= d [x] Option: Metric -> g: a symmetric 2-tensor expression, representing a metric to be used in place of the default metric for xs bundle (which is assumed to be Riemannian). See also Extd, LaplaceBeltrami, Div. FactorConstants 5.8

FactorConstants[x] applies the Mathematica function Factor to the constant factor in each term of the tensor expression x. FactorConstants[x,n] or FactorConstants[x,{n1,n2,...}] applies FactorConstants only to term n or to terms n1,n2,... of x. See also CollectConstants, SimplifyConstants, ConstantFactor, TensorFactor. FirstBianchiRule 7.2

FirstBianchiRule is a rule that attempts to turn sums containing two 4-tensors with Symmetries -> RiemannSymmetries into a single term, using the rst Bianchi identity. For example: i Out[7]= 2 Rm j k l i + 2 Rm l j k

In[8]:= % /. FirstBianchiRule i Out[8]= 2 Rm k j l See also SecondBianchiRule, BianchiRules, ContractedBianchiRules. FirstStructureRule 7.5

FirstStructureRule is a rule that implements the rst structure equation for exterior derivatives of basis covectors. For example (assuming $WedgeConvention = Alt):

8 REFERENCE LIST

71

i Out[9]= d[Basis ] In[10]:= % /. FirstStructureRule k6 Out[10]= Basis i ^ Conn k6 1 i + - Tor 2

See also SecondStructureRule, CompatibilityRule, StructureRules, $WedgeConvention. FirstUp FirstUp is a value for the RiemannConvention option of DefineBundle. FlatConnection FlatConnection is a DefineBundle option. The function FlatConnection[bundle] returns True or False. See also DefineBundle. FormQ FormQ[x] returns True if x is a covariant alternating tensor expression, and False otherwise. See also DefineTensor. Grad 4.4 Grad[x] is the gradient of the tensor expression x. It is the same as Del[x], except that the last index is contravariant instead of covariant. If x is a scalar function (0-tensor), then Grad[x] is a vector eld. See also Del, Laplacian. Hermitian Hermitian is a value for the Symmetries option of DefineTensor. HodgeInner 3.6 2.1 2.1

HodgeInner[x,y] represents the Hodge inner product of the alternating tensors x and y, taken with respect to the default metric(s) on the bundle(s) with which x and y are associated. It is equal to the usual inner product multiplied by a numerical scale factor (depending on $WedgeConvention) chosen so that, if {ei } are orthonormal 1-forms, then the following k-forms are orthonormal: ei1 . . . eik , i1 < . . . < ik .

8 REFERENCE LIST

72

The arguments x and y must be alternating tensor expressions of the same rank. In output form, HodgeInner[x,y] appears as <<x, y>>. See also Inner, HodgeNorm, Int, ExtdStar. HodgeNorm HodgeNorm[x] is the norm of the alternating tensor expression x, with respect to the Hodge inner product. It is automatically converted by Ricci to Sqrt[HodgeInner[x,Conjugate[x]]]. See also HodgeInner, Norm. Im 2.6 Ricci converts Im[x] to (x - Conjugate[x])/(2I). See also Conjugate. Imaginary 2.3, 2.4

Imaginary is a value for the Type option of DefineConstant, DefineTensor, and DefineMathFunction. IndexOrderedQ 6.1

IndexOrderedQ[{indices}] returns True if {indices} are ordered correctly according to Riccis index ordering rulesrst by name, then by altitude (lower before upper)and False otherwise. See also OrderDummy, OrderCovD, DefineRelation, DefineRule. IndexQ IndexQ[i] returns True if i is an index name, and False otherwise. See also DefineBundle, DefineIndex. Inner 3.5

If x and y are tensor expressions, Ricci interprets Inner[x,y] as the inner product of x and y, taken with respect to the default metric(s) on the bundle(s) with which x and y are associated. If x and y have opposite variance (for example, if x is a 1-form and y is a vector eld), then Inner[x,y] represents the natural pairing between x and y. The arguments x and y must have the same rank. Ricci automatically expands inner products of sums and scalar multiples. In output form, Inner[x,y] appears as <x, y>. See also Norm, HodgeInner. InsertIndices 2.6

If x is a tensor expression of rank k, then InsertIndices[ x, {L[i1],...,L[ik]} ] causes the indices L[i1], . . . , L[ik] to be inserted into the k index slots in order, yielding a component expression. In input form, this can be abbreviated x[ L[i1], ..., L[ik] ]. Indices in each slot may be lower (L[i]) or upper (U[i]). If x has rank 0, then InsertIndices[ x, {} ] or x[] converts it to a component expression by generating dummy indices as necessary.

8 REFERENCE LIST

73

If one or more indices are inserted into a rank-0 expression, then InsertIndices is automatically converted to CovD. See also CovD. Int 3.7 If x and y are alternating tensor expressions with Rank[x] Rank[y], Int[x,y] represents the generalized interior product of x into y. When x is a vector eld, Int[x,y] is interior multiplication of x into y (with a numerical factor depending on $WedgeConvention). In general, Int[x, ] is the adjoint (with respect to the Hodge inner product) of wedging with x on the left: if alpha, beta, and gamma are alternating tensors of ranks a, b, a + b, respectively, then <<alpha ^ beta, gamma>> = <<beta, Int [gamma]>>. alpha See also HodgeInner, Wedge, Dot, $WedgeConvention. Integer 2.3

Ricci recognizes Integer as a value for the Type option of DefineConstant. Inverse 7.3

If x is a 2-tensor expression, Ricci interprets Inverse[x] as the matrix inverse of x. See also Dot, Kronecker. Kronecker 7.3

Kronecker[L[i],U[j]] is the Kronecker delta symbol, which is equal to 1 if i = j and 0 otherwise. Kronecker is a generic tensor representing the identity endomorphism of any bundle. See also Dot, Inverse. L 2.2, 2.4 L[i] represents a lower index i. L[i[Bar]] is the internal representation for a lower barred index i; it can be abbreviated in input form by LB[i]. See also LB, U, UB. LaplaceBeltrami 4.8 LaplaceBeltrami[x] is the Laplace-Beltrami operator = dd + d d applied to the dierential form x. It is automatically replaced by Extd[ExtdStar[x]] + ExtdStar[Extd[x]]. Option: Metric -> g: a symmetric 2-tensor expression, representing a metric to be used in place of the default metric for xs bundle (which is assumed to be Riemannian). See also Extd, ExtdStar, Laplacian.

8 REFERENCE LIST

74

Laplacian

4.5

Laplacian[x] is the covariant Laplacian of the tensor expression x. The sign convention of Laplacian is controlled by the global variable $LaplacianConvention. When $LaplacianConvention = DivGrad (the default), Laplacian[x] is automatically replaced by Div[Grad[x]]. When $LaplacianConvention = PositiveSpectrum, Laplacian[x] is replaced by -Div[Grad[x]]. Options: Metric -> g: a symmetric 2-tensor expression, representing a metric to be used in place of the default metric for the underlying tangent bundle of x. Connection -> cn species that covariant derivatives are to be taken with respect to the connection cn instead of the default connection. Ordinarily, cn should be an expression of the form Conn + diff, where diff is a 3-tensor expression representing the dierence tensor between the default connection and cn. See also Div, Grad, LaplaceBeltrami. LB 2.2, 2.4 LB[i] is the input form for a lower barred index i. It is converted automatically to the internal form L[i[Bar]]. See also L, U, UB, Bar. LeviCivitaConnection 7.6

LeviCivitaConnection[g] represents the Levi-Civita connection of the arbitrary Riemannian metric g. It behaves as a product tensor of rank {2,1}. When indices are inserted, the connection coecients (Christoffel symbols) are computed in terms of the background connection of gs bundle (assumed to be Riemannian) and covariant derivatives of g. The upper index is the second one, just as for the default connection Conn. See also Conn, Curvature, RiemannTensor, RicciTensor, ScalarCurv. Lie 4.9 Lie[v,x] is the Lie derivative of the tensor expression x in the direction v. The rst argument v must be a vector eld (a contravariant 1-tensor expression). If v and x are both vector elds, Lie[v,x] is their Lie bracket, and Ricci automatically puts the factors in lexical order by inserting a minus sign if necessary. Applying LieRule causes Lie derivatives of dierential forms to be expanded in terms of Int and exterior derivatives. See also Del, LieRule. LieRule 4.9

LieRule is a rule that transforms Lie derivatives of dierential forms to expressions involving exterior derivatives and Int according to the relation

8 REFERENCE LIST

75

Lie [w] --> Int [d[w]] + d[Int[ [w]]] v v v See also Lie, Int, Extd. LowerAllIndices 5.20

LowerAllIndices[x] lowers all of the indices in x (except those appearing on metrics or Basis covectors) by inserting appropriate metrics with raised indices. LowerAllIndices[x,n] or LowerAllIndices[x,{n1,n2,...}] applies LowerAllIndices to term n or to terms n1,n2,... of x. See also CorrectAllVariances. Method Method is an option for the Ricci simplication command OrderDummy, which species how hard the command should work to simplify the expression. Metric Metric -> metricname is an option for Del, Div, Grad, ExtdStar, CovD, Laplacian, and LaplaceBeltrami. The function Metric[bundle] returns the bundles metric. See also DefineBundle. MetricQ MetricQ[x] returns True if x is a metric with or without indices, and False otherwise. See also DefineBundle. MetricType 7.2

MetricType is a DefineBundle option, which species whether the bundles metric has special properties. Currently the only allowable values are MetricType -> Normal (no special properties), and MetricType -> Riemannian (for a Riemannian tangent bundle). The default is Normal. Negative 2.3, 2.4, 2.6

Ricci recognizes Negative as a value for the Type option of DefineConstant, DefineTensor, and DefineMathFunction. NewDummy 5.16

NewDummy[x] converts all dummy indices occurring in the tensor expression x to computer-generated dummy indices. NewDummy[x,n] or NewDummy[x,{n1,n2,...}] applies NewDummy to term n or to terms n1,n2,... of x. For most bundles, the generated dummy names are

8 REFERENCE LIST

76

of the form kn, where k is the last index name associated with bundle and n is an integer. For one-dimensional bundles, only the dummy name k itself is generated. See also RenameDummy, OrderDummy. NewRule NewRule is a DefineRule option. NonNegative 2.3, 2.4, 2.6 6.2

Ricci recognizes NonNegative as a value for the Type option of DefineConstant, DefineTensor, and DefineMathFunction. NonPositive 2.3, 2.4, 2.6

NonPositive is a value for the Type option of DefineConstant, DefineTensor, and DefineMathFunction. NoOneDims NoOneDims is a value for the Mode option of AbsorbMetrics, used internally by TensorSimplify. Norm Norm[x] is the norm of the tensor expression x. It is automatically converted by Ricci to Sqrt[Inner[x,Conjugate[x]]]. See also Inner, HodgeNorm. Normal Normal is a value for the MetricType option of DefineBundle. NoSymmetries 2.4 7.2

NoSymmetries is a value for the Symmetries option of DefineTensor. Odd [47 Odd is a value for the Type option of DefineConstant. OneDims OneDims is a value for the Mode option of CorrectAllVariances, used internally by TensorSimplify. OrderCovD 5.18 2.3

OrderCovD[x] puts all of the indices appearing after ; in the tensor expression x in order according to Riccis index-ordering rules (rst alphabetically by name, then by altitude), by adding appropriate curvature and torsion terms. OrderCovD[x,n] or OrderCovD[x,{n1,n2,...}] applies OrderCovD only to term n or to terms n1,n2,... of x. See also CommuteCovD, CovDSimplify, IndexOrderedQ.

8 REFERENCE LIST

77

OrderDummy

5.6

OrderDummy[x] attempts to put the dummy indices occurring in the tensor expression x in a canonical form. OrderDummy[x,n] or OrderDummy[x,{n1,n2,...}] applies OrderDummy to term n or to terms n1,n2,... of x. All pairs of dummy indices are ordered so that the lower member appears rst whenever possible. Then OrderDummy tries various permutations of the dummy index names in each term of x, searching for the lexically smallest version of the expression among all equivalent versions. Option: Method -> n species how hard OrderDummy should work to nd the best possible version of the expression. The allowable values are 0, 1, and 2. The default is Method -> 1, which means that dummy index names are interchanged in pairs until the lexically smallest version of the expression is found; this is used by TensorSimplify. Method -> 2 causes OrderDummy to try all possible permutations of the dummy index names; this is used by SuperSimplify. Method -> 0 means dont try interchanging names at all. If n is the number of dummy index pairs in the expression, the time taken by Method -> 1 is proportional to n2 , while the time taken by Method -> 2 is proportional to n!. This can be very slow, especially if there are more than 4 dummy pairs per term. See also TensorSimplify, SuperSimplify, RenameDummy, NewDummy. ParallelFrame ParallelFrame is a DefineBundle option. Plus 2.6 Ricci transforms expressions of the form (a + b)[L[i],...] into InsertIndices[ a+b, {L[i],...} ]. Any number of upper and/or lower indices can be inserted in this way, provided that the number of indices is consistent with the rank of a+b. Positive 2.3, 2.4, 2.6 2.1

Ricci recognizes Positive as a value for the Type option of DefineConstant, DefineTensor, and DefineMathFunction. PositiveDefinite PositiveDefinite is a DefineBundle option. PositiveSpectrum PositiveSpectrum is $LaplacianConvention. a value for the global 4.5 variable 2.1

8 REFERENCE LIST

78

Power

2.6

If x is a tensor expression and p is a positive integer, Ricci interprets x^p as the p-th symmetric power of x with itself. Ricci transforms expressions of the form (x^p)[i] into InsertIndices[x^p,{i}] whenever i is an index (L[j] or U[j]). Ricci causes a power of a product such as (a b)^p to be expanded into a product of powers a^p * b^p, provided a and b do not contain any indices; if they do contain indices, then (a b)^p is transformed to Summation[a b]^p, to prevent the expression from being expanded to a^p b^p. Summation is not printed in output form. See also SymmetricProduct, Summation, PowerSimplify, ProductExpand. PowerSimplify 5.14

PowerSimplify[x] attempts to simplify powers that appear in the tensor expression x. PowerSimplify[x,n] or PowerSimplify[x,{n1,n2,...}] applies PowerSimplify only to term n or to terms n1,n2,... of x. PowerSimplify transforms x by Converting products like a^p b^p to Summation[a b]^p when a and b contain indices. Summation is an internal Ricci function that prevents such products from being automatically expanded by Mathematica. It does not appear in output form; instead, Summation[a b]^p prints as (a b)^p. Expanding powers like (a^b)^c to a^(b c), provided it knows enough about a, b, and c to know this is legitimate (for example, if c is an integer, or if a is positive and b is real). Expanding and collecting constants in any expression that appears as base or exponent of a power. See also TensorSimplify, Power, Summation. ProductExpand 5.13

ProductExpand[x] expands symmetric products and wedge products of 1-tensors that occur in x, and rewrites them in terms of tensor products. ProductExpand[x,n] or ProductExpand[x,{n1,n2,...}] applies ProductExpand only to term n or to terms n1,n2,... of x. See also BasisExpand, Sym, Alt, SymmetricProduct, Wedge. Quiet Quiet is an option for some of the dening and undening commands in the Ricci package. The option Quiet -> True silences the usual messages printed by these commands. The default is the value of the global variable $Quiet, which is initially False.

8 REFERENCE LIST

79

Rank Rank[x] returns the rank of the tensor expression x. If t is a tensor without indices, Rank[t] is the total rank as specied in the call to DefineTensor. If t is a tensor with all index slots lled, Rank[t] = 0. If t is a product tensor with some, but not all, index slots lled, then Rank[t] is the total rank of t minus the number of lled slots. For any other tensor expression x, Rank[x] depends on the meaning of the expression. See also DefineTensor. Re Ricci converts Re[x] to (x + Conjugate[x])/2. See also Conjugate. Real 2.1, 2.3, 2.4, 2.6 Ricci recognizes Real as a value for the Type option of DefineBundle, DefineConstant, DefineTensor, and DefineMathFunction. RenameDummy 5.5 2.6

RenameDummy[x] changes the names of dummy indices in x to standard names. RenameDummy[x,n] or RenameDummy[x, {n1,n2,...} ] applies RenameDummy only to term n or to terms n1,n2,... of x. RenameDummy chooses dummy names in alphabetical order from the list of index names associated with the appropriate bundle, skipping those names that already appear in x as free indices. When the list of index names is exhausted, computer-generated names of the form kn are used, where k is the last index name in the list and n is an integer. For one-dimensional bundles, n is omitted. See also TensorSimplify, OrderDummy, NewDummy. RicciSave 2.7

RicciSave["filename"] causes the denitions of all symbols dened in Mathematicas current context, along with Riccis predened tensors and global variables, to be saved in Mathematica input form into the le filename. (The current context is usually Global, and usually includes all symbols you have used in the current session.) The previous contents of filename are erased. You can read the denitions back in by typing <<filename. RicciTensor 7.6

RicciTensor -> name is a DefineBundle option, which species a name to be used for the Ricci tensor for this bundle. The default is Rc. This option takes eect only when MetricType -> Riemannian has been specied. RicciTensor[g] represents the Ricci curvature tensor of the arbitrary metric g. When indices are inserted, the components of the Ricci tensor are computed in terms of covariant derivatives of g and the curvature of

8 REFERENCE LIST

80

the default metric on gs bundle, which is assumed to be Riemannian. Note: If you insert any upper indices, they are considered to have been raised by the default metric on gs bundle, not by g. To use g to raise indices, you must explicitly multiply by components of Inverse[g]. See also DefineBundle, LeviCivitaConnection. RiemannConvention RiemannTensor, ScalarCurv, 7.2

RiemannConvention is a DefineBundle option, which species the index convention of the Riemannian curvature tensor. Riemannian 7.2

MetricType -> Riemannian is a DefineBundle option for dening a Riemannian tangent bundle. RiemannSymmetries RiemannSymmetries is DefineTensor. RiemannTensor a value for the Symmetries option of 7.6

RiemannTensor -> name is a DefineBundle option, which species a name to be used for the Riemannian curvature tensor for this bundle. The default is Rm. This option takes eect only when MetricType -> Riemannian has been specied. RiemannTensor[g] represents the Riemannian curvature tensor of the arbitrary metric g. Like the default Riemannian curvature tensor, RiemannTensor[g] is a covariant 4-tensor. When indices are inserted, the components of the curvature tensor are computed in terms of covariant derivatives of g and the curvature of the default metric on gs bundle, which is assumed to be Riemannian. Note: If you insert any upper indices, they are considered to have been raised by the default metric on gs bundle, not by g. To use g to raise indices, you must explicitly multiply by components of Inverse[g]. See also DefineBundle, ScalarCurv. Same Bundle -> Same is a special DefineTensor option for product tensors. ScalarCurv 7.6 LeviCivitaConnection, RicciTensor,

ScalarCurv -> name is a DefineBundle option, which species a name to be used for the scalar curvature for this bundle. The default is Sc. This option takes eect only when MetricType -> Riemannian has been specied.

8 REFERENCE LIST

81

ScalarCurv[g] represents the scalar curvature of the arbitrary metric g. When indices are inserted, the scalar curvature is computed in terms of covariant derivatives of g and the curvature of the default metric on gs bundle, which is assumed to be Riemannian. See also RiemannTensor, DefineBundle. ScalarQ ScalarQ[x] is True if x is a rank 0 tensor expression with no free indices, and False otherwise. SecondBianchiRule 7.2 RicciTensor, LeviCivitaConnection,

SecondBianchiRule is a rule that attempts to turn sums containing two dierentiated Riemannian curvature tensors into a single term using the second Bianchi identity. For example: 2 Rm + 2 Rm /. SecondBianchiRule i j k l; m i j l m; k --> -2 Rm i j m k; l See also FirstBianchiRule, ContractedBianchiRules, BianchiRules, DefineBundle. SecondStructureRule 7.5

SecondStructureRule is a rule that implements the second structure equation for exterior derivatives of the generic connection forms. For example: Extd[Conn[L[i],U[j]]] /. SecondStructureRule --> 1 j - Curv 2 i k + Conn i ^ Conn k j

See also FirstStructureRule, CompatibilityRule, StructureRules, Curv, Conn, CurvToConnRule. SecondUp 7.2

SecondUp is a value for the RiemannConvention option of DefineBundle. SimplifyConstants 5.9

SimplifyConstants[x] applies the Mathematica function Simplify to the constant factor in each term of the tensor expression x.

8 REFERENCE LIST

82

SimplifyConstants[x,n] or SimplifyConstants[x,{n1,n2,...}] applies SimplifyConstants only to term n or terms n1,n2,... of x. See also CollectConstants, FactorConstants, ConstantFactor. Skew Skew is a synonym for Alternating. SkewHermitian SkewHermitian is a value for the Symmetries option of DefineTensor. SkewQ SkewQ[x] is True if x is an alternating or skew-Hermitian tensor expression, and False otherwise. StructureRules 7.5 2.4

StructureRules is the union of CompatibilityRule, FirstStructureRule, and SecondStructureRule. Summation If a and b contain indices, then Ricci transforms (a b)^p internally to Summation[a b]^p, to prevent the expression from being expanded to a^p * b^p. Summation is not printed in output form; instead, Summation[a b]^p appears as if it were (a b)^p. See also Power, PowerSimplify. SuperSimplify 5.1, 5.2

SuperSimplify[x] attempts to put the tensor expression x into a canonical form, so that two expressions that are equal are usually identical after applying SuperSimplify. SuperSimplify[x,n] or SuperSimplify[x,{n1,n2,...}] applies SuperSimplify only to term n or to terms n1, n2, . . . of x. SuperSimplify works exactly the same way as TensorSimplify, except that it calls OrderDummy with the option Method -> 2, so that all possible permutations of dummy index names are tried. This is generally much slower for expressions having more than 4 dummy index pairs: for an expression with k dummy index pairs per term, the time taken by TensorSimplify is proportional to k 2 , while the time taken by SuperSimplify is proportional to k!. See also TensorSimplify, OrderDummy, TensorCancel. Sym 3.9 Sym[x] represents the symmetrization of the tensor expression x. If x is symmetric, then Sym[x] = x. See also Alt, SymmetricProduct. Symmetric Symmetric is a value for the Symmetries option of DefineTensor. 2.4

8 REFERENCE LIST

83

SymmetricProduct

3.3

If x, y, and z are tensor expressions, SymmetricProduct[x,y,z] or x * y * z or x y z represents their symmetric product. Symmetric products are represented internally by ordinary multiplication. In output form, Ricci inserts explicit asterisks whenever non-scalar tensor expression are multiplied together, to remind the user that multiplication is being interpreted as symmetric product. Mathematically, the symmetric product is dened by ab = Sym(ab). See also Times, Power, Sym, ProductExpand. SymmetricQ SymmetricQ[x] is True if x is a symmetric or Hermitian tensor expression, and False otherwise. Symmetries Symmetries is a DefineTensor option. TangentBundle TangentBundle is a DefineBundle option. The function TangentBundle[x] returns the tangent bundle list for the bundle x. See also UnderlyingTangentBundle. Tensor Riccis internal form for tensors is Tensor[ name, {i,j,...}, {k,l,...} ] where name is the tensors name, i,j,... are the tensor indices (each of the form L[i] or U[i]), and k,l,... are the indices resulting from covariant dierentiation. In input form, this can be typed name [i,j,...] [k,l,...]. Either set of indices in brackets can be omitted if it is empty. See also DefineTensor, UndefineTensor. TensorCancel 5.21 2.4 2.1 2.4

TensorCancel[x] attempts to simplify each term of x by combining and cancelling common factors, even when the factors have dierent names for their dummy indices. TensorCancel[x,n] or TensorCancel[x,{n1,n2,...}] applies TensorCancel only to term n or to terms n1,n2,... of x. See also TensorSimplify, SuperSimplify. TensorData TensorData[name] is a list containing data for the tensor name, used internally by Ricci. See also DefineTensor.

8 REFERENCE LIST

84

TensorExpand

5.3

TensorExpand[x] expands products and positive integral powers in x, just as Expand does, but maintains correct dummy index conventions and does not expand constant factors. TensorExpand[x,n] or TensorExpand[x,{n1,n2,...}] applies TensorExpand to term n or to terms n1,n2,... of x. If you apply the Mathematica function Expand to an expression containing tensors with indices, it is automatically converted to TensorExpand. See also TensorSimplify, Expand, BasisExpand, CovDExpand, ProductExpand. TensorFactor TensorFactor[x] returns the product of all the non-constant factors in x, which should be a monomial. See also ConstantFactor. TensorMetricQ TensorMetricQ[tensorname] returns True if tensorname is the metric of some bundle, and False otherwise. See also DefineBundle. TensorProduct 3.1

TensorProduct[x,y,z] or TProd[x,y,z] represents the tensor product of x, y, and z. Ricci automatically expands tensor products of sums and scalar multiples. In output form, tensor products appear as x (X) y (X) z See also SymmetricProduct, Wedge. TensorQ TensorQ[name] returns True if name is the name of a tensor, and False otherwise. See also DefineTensor. TensorRankList TensorRankList stores the list of ranks for a product tensor. Used internally by Ricci. See also DefineTensor. TensorSimplify 5.1

TensorSimplify[x] attempts to put the tensor expression x into a canonical form, so that two expressions that are equal will usually be identical after applying TensorSimplify. TensorSimplify[x,n] or TensorSimplify[x,{n1,n2,...}] applies TensorSimplify to term n or to terms n1,n2,... of x. TensorSimplify expands products and positive integer powers, simplies powers, uses metrics to raise and lower indices, tries to rename all dummy indices in a canonical order, and collects all terms

8 REFERENCE LIST

85

containing the same tensor factors but dierent constant factors. When there are two or more dummy index pairs associated with the same bundle, TensorSimplify tries exchanging dummy index names in pairwise, choosing the lexically smallest result. It does not reorder indices after the ; (use OrderCovD or CovDSimplify to do that). TensorSimplify calls CorrectAllVariances, TensorExpand, AbsorbMetrics, PowerSimplify, RenameDummy, OrderDummy, and CollectConstants. See also SuperSimplify, OrderCovD, CovDSimplify, TensorCancel, FactorConstants, SimplifyConstants. TensorSymmetry A tensor symmetry is an object of the form TensorSymmetry[name,d,{perm1,sgn1,...,permN,sgnN}], where d is a positive integer, perm1,...,permN are non-trivial permutations of {1, 2, . . . , d}, and sgn1,...,sgnN are constants, usually 1. TensorSymmetry objects can be dened with DefineTensorSymmetries and used in the DefineTensor command. See also DefineTensorSymmetries, DefineTensor. TeXFormat TeXFormat is an option for DefineTensor and DefineIndex. Times 2.6, 3.3

Ricci uses ordinary multiplication to represent symmetric products of tensors. Ricci transforms expressions of the form (a * b)[L[i],...] into InsertIndices[ a*b, {L[i],...} ]. Any number of upper and/or lower indices can be inserted in this way, provided that the number of indices is consistent with the rank of a*b. In output form, Ricci modies Mathematicas usual ordering of factors: constants are printed rst, then scalars, then higher-rank tensor expressions, and Ricci inserts explicit asterisks between tensor factors of rank greater than 0 to remind the user that multiplication is being interpreted as symmetric product. See also SymmetricProduct, Power, ProductExpand. Tor 7.5 Tor is the name for the generic torsion tensor for the default connection in any bundle. It is a product tensor of rank {1,2}; inserting the rst index yields the torsion 2-forms associated with the default basis. Inserting all three indices yields the components of the torsion tensor. TorsionFree TorsionFree is a DefineBundle option. The function TorsionFree[bundle] returns True or False. 2.1, 7.5

8 REFERENCE LIST See also DefineBundle. TotalRank

86

TotalRank[tensorname] returns the total rank of the tensor tensorname. This function is used internally by Ricci. To compute the rank of an arbitrary tensor expression, use Rank instead. See also Rank, DefineTensor. TProd TProd is an abbreviation for TensorProduct. Tr 7.3 Tr[x] represents the trace of the 2-tensor expression x, computed with respect to the default metric on xs bundle. See also Dot. Transpose 7.3 3.1

If x is a tensor expression, Ricci interprets Transpose[x] as x with its index positions reversed. See also Dot. Type 2.1, 2.3, 2.4, 2.6 Type is an option for DefineBundle, DefineConstant, DefineTensor, and DefineMathFunction. U 2.2, 2.4 U[i] represents an upper index i. U[i[Bar]] is the internal representation for an upper barred index i; it can be abbreviated in input form by UB[i]. See also UB, L, LB. UB 2.2, 2.4 UB[i] is the input form for an upper barred index i. It is converted automatically to the internal form U[i[Bar]]. See also U, L, LB, Bar. UndefineBundle 2.1

UndefineBundle[bundle] clears the denition of bundle, its metric, and all its indices. If bundle is a Riemannian tangent bundle, it also clears the denitions of its Riemann, Ricci, and scalar curvature tensors. If you try to perform computations with previous expressions involving bundles indices, you may get unpredictable results. Option: Quiet -> True or False. Default is False, or the value of $Quiet if set. See also DefineBundle, UndefineIndex, Declare.

8 REFERENCE LIST

87

UndefineConstant UndefineConstant[c] removes cs denition as a constant. Option: Quiet -> True or False. Default is False, or the value of $Quiet if set. See also DefineConstant, Declare. UndefineIndex UndefineIndex[i] or UndefineIndex[{i,j,k}] removes the association of indices with their bundles. If you try to perform computations with previous expressions involving these indices, you may get unpredictable results. Option: Quiet -> True or False. Default is False, or the value of $Quiet if set. See also DefineIndex, DefineBundle, UndefineBundle, Declare. UndefineRelation 6.1

UndefineRelation[tensor] deletes the relation previously dened for tensor. The argument must exactly match the rst argument of the corresponding call to DefineRelation. Note: There is no UndefineRule function; to remove the denition of a rule dened by DefineRule, simply execute an assignment such as rulename =., or call DefineRule with the option NewRule -> True. Option: Quiet -> True or False. Default is False, or the value of $Quiet if set. See also DefineRelation. UndefineTensor UndefineTensor[tensorname] clears the denition of tensorname. If you try to perform computations with previous expressions involving tensorname, you may get unpredictable results. Option: Quiet -> True or False. Default is False, or the value of $Quiet if set. See also DefineTensor, Declare.

8 REFERENCE LIST

88

UndefineTensorSymmetries UndefineTensorSymmetries[name] deletes the TensorSymmetry object created by DefineTensorSymmetries[name,...]. UnderlyingTangentBundle UnderlyingTangentBundle[x] returns a list of bundles representing the underlying tangent bundle of the expression. The tangent bundle is the direct sum of the bundles in the list. It is assumed that all tensors used in a given expression have the same underlying tangent bundle. Variance Variance is a DefineTensor option. The function Variance[x] returns a list of variances of the tensor expression x, one for each index slot. See also DefineTensor, Bundles. VectorFieldQ VectorFieldQ[x] returns True if x is a contravariant 1-tensor expression, and False otherwise. Wedge 3.2 2.4

Wedge[x,y,z] represents the wedge or exterior product of x, y, and z. The arguments of Wedge must be alternating tensor expressions. They need not be covariant tensors, however; Ricci handles wedge products of covariant, contravariant, and mixed tensors all together, using the metric to raise and lower indices as needed. Ricci automatically expands wedge products of sums and scalar multiples, and arranges factors in lexical order by inserting appropriate signs. The interpretation of Wedge in terms of tensor products is determined by the global variable $WedgeConvention. In output form, wedge products are indicated with a caret: x ^ y ^ z See also $WedgeConvention, Alt, TensorProduct, ProductExpand.

8.2

Global variables

This section describes Riccis global variables: these are symbols that can be set by the user to control the way Ricci evaluates certain expressions. For example, to change the value of $WedgeConvention to Det, just execute the assignment $WedgeConvention = Det

8 REFERENCE LIST

89

The variables described below all begin with $, and are all saved automatically if you issue the RicciSave command. $DefaultTangentBundle 2.1

The global variable $DefaultTangentBundle can be set by the user to a bundle or list of bundles. It is used by DefineBundle as the default value for the TangentBundle option, by DefineTensor as the default value for the Bundle option, and by BasisExpand and similar commands as the default bundle for tensors such as Curv, Conn, and Tor that are not intrinsically associated with any particular bundle. By default, $DefaultTangentBundle is set to the rst bundle the user denes (or the direct sum of this bundle and its conjugate if the bundle is complex). If you want to override this default behavior, you can give a value to $DefaultTangentBundle before calling DefineBundle. For example: $DefaultTangentBundle = {horizontal, vertical} species that the tangent bundle of all subsequently-dened bundles is to be the direct sum of the bundles named horizontal and vertical. See also DefineTensor, DefineBundle, BasisExpand. $LaplacianConvention 4.5

$LaplacianConvention determines which sign convention is used for the covariant Laplacian on functions and tensors. $LaplacianConvention = DivGrad means that Laplacian[x] = Div[Grad[x]], while $LaplacianConvention = PositiveSpectrum means that Laplacian[x] = -Div[Grad[x]]. Default is DivGrad. See also Laplacian. $MathFunctions $MathFunctions is a list of names that have been dened as scalar mathematical functions for use by Ricci. See also DefineMathFunction. $Quiet The global variable $Quiet is used by Ricci to determine whether the dening and undening commands report on what they are doing. Setting $Quiet=True will silence these chatty commands. The default is False. It can be overridden for a particular command by specifying the Quiet option as part of the command call. $RiemannConvention 7.2

The global variable $RiemannConvention can be set by the user to specify a default value for the RiemannConvention option of DefineBundle. The allowed values are FirstUp and SecondUp (the default). See also DefineBundle.

8 REFERENCE LIST

90

$TensorFormatting

2.2

The global variable $TensorFormatting can be set to True or False by the user to turn on or o Riccis special output formatting of tensors and indices. Default is True. $TensorTeXFormatting The global variable $TensorTeXFormatting can be set to True or False by the user to turn on or o Riccis special formatting of tensors in TeXForm. Default is True. $WedgeConvention 3.2

The global variable $WedgeConvention can be set by the user to determine the interpretation of wedge products. The allowed values are Alt and Det. The default is Alt. Suppose is a p-form and a q-form: $WedgeConvention = Alt: 1 det n!

= Alt( )

and Basis 1 . . . Basis n =

$WedgeConvention = Det: (p + q)! Alt( ) and Basis 1 . . . Basis n = det p!q!

See also Wedge, Alt.