For an Array A, say, and some positive integer n, say, Maple interpretes A^n as raising each entry separately to the same power n. Without the Physics package loaded, A^n can also be written as A . A . ... . A (n times). But with the Physics package loaded, this equality is broken (at least in Maple 2017): If A is a 2D square Array, A . A all of a sudden is no longer equal to A^2, but rather to convert(A,Matrix)^2, i.e., to the square of the Array considered as a Matrix. The presence of the dot operator seems to make the Physics enviroment convert A to a Matrix. This seems to me to be a bug.

Consider the following expression (omega being the socalled minimal spin connection in the vierbein formalism of general relativity):

with(Physics):
with(Tetrads):
omegaDef := omega[mu,c,d] = 1/2*(
   +e_[~rho,c]*(d_[mu](e_[rho,d]) - d_[rho](e_[mu,d]))
   -e_[~rho,d]*(d_[mu](e_[rho,c]) - d_[rho](e_[mu,c]))
)
-1/2*e_[~a,mu]*e_[~rho,c]*e_[~sigma,d]*(d_[rho](e_[sigma,a]) - d_[sigma](e_[rho,a]));
Define(omegaDef):

The Define command raises "Error, (in Physics:-Define) numeric exception: division by zero" in Schwarschild spacetime (loaded with g_[sc]), but not, say, in Tolman spacetime (loaded with g_[tolman]). Furthermore, if either the first two terms, or the last term, are/is removed in omega, then no error is raised in Schwarzschild spacetime. What is going on?

I have two tensors, E_ and F_ below, that I believe should be equal. But they are not, and I cannot understand why. The problem does not appear in, say, Schwarschild spacetime, but it appears in Boyer-Lindquist spacetime, metric [5,29,1]; perhaps it appears only if the vierbein is nondiagonal?

The code: Loading the packages and the metric:

restart:
with(Physics):
with(Tetrads):
g_[[5,29,1]];   # The Boyer-Lindquist metric

Set up galilean and nongalilean Levi-Civita's, respectively, following the recipe given elsewhere:

Define(varepsilon[a,b,c,d] = Array((1..4)$4,rhs(LeviCivita[nonzero])),quiet):
Setup(levicivita = nongalilean):
# Checking that the Levi-Civita's are indeed different
varepsilon[1,2,3,4];   # The galilean case
LeviCivita[1,2,3,4];   # The nongalilean case

Define the two tensors E_ and F_, using mixed type Levi-Civita's for the latter:

Define(
   E_[~a,mu] = varepsilon[~a,~b,~c,~d]*LeviCivita[mu,nu,rho,sigma]*e_[b,~nu]*e_[c,~rho]*e_[d,~sigma],
   F_[~a,mu] = varepsilon[~a,b,c,d]*LeviCivita[mu,~nu,~rho,~sigma]*e_[~b,nu]*e_[~c,rho]*e_[~d,sigma]
,quiet):
E_[definition];
F_[definition];

Compare the two expressions, which should be equal, I believe.

expr := simplify(TensorArray(E_[~a,mu] - F_[~a,mu])) assuming a::real,theta > 0,theta < Pi;
eval(expr,{a = 1,m = 1,r = 2,theta = Pi/4});   # Just to make the difference completely obvious

[I have trouble copy-pasting the output from these two lines, so you will have to execute the worksheet provided below to see it.]

However, they are not equal. Why not?

Download worksheet: MixedTypeLeviCivitas.mw.

In the realm of tetrads where both world indices and Lorentz indices are present, contractions, say, using simultaneously the Minkowskian (galilean) Levi-Civita symbol,

and the curvilinear Levi-Civita (pseudo-)tensor,

can be considered. Although each of the two types of Levi-Civitas can easily be obtained separately by specifying Setup(levicivita = galilean) or Setup(levicivita = nongalilean), I cannot figure out how to have them both available at the same time. Any suggestions?

PS: I am, of course, aware of the fact that the two Levi-Civitas are related by some appropiate square-root of the determinant of the metric, but I have no desire to fiddle around with explicit such determinants if they can be avoided.

This is probably a question to Edgardo: In another thread, the following quantity is considered [Eqs. (5) and (6) combined]:

expand(gamma_[definition]);

This is all very well, but it seems to depend on the metric loaded: if the Schwarzschild metric g_[sc] is loaded, then the above output results, but if the Minkowski metric g_[minkowski] is loaded, then the output of the above expansion is identically zero. Does that make sense? Is that intentional? The explicitly evaluated Ricci rotation coefficients vanish identically for the Minkowski metric, of course, but if evaluation is performed for that case, then why not also for the Schwarschild metric [not meaning to say that I want evaluation]?