Difference between revisions of "Transverse traceless gauge"

From Universe in Problems
Jump to: navigation, search
(Created page with "5")
 
Line 1: Line 1:
 
[[Category:Weak field limit and gravitational waves|5]]
 
[[Category:Weak field limit and gravitational waves|5]]
 +
 +
 +
 +
 +
From now on we consider only vacuum solutions. Suppose we use the Lorenz gauge. As shown above, we still have the freedom of coordinate transformations with $\square \xi^\mu =0$, which preserve the gauge.  So, let us choose some arbitrary (timelike) field $u^\mu$ and, in addition to the Lorenz gauge conditions, demand that the perturbation is also transverse $u^{\mu}\bar{h}_{\mu\nu}=0$ with regard to it plus that it is traceless $\bar{h}^{\mu}_{\mu}=0$. Then $h_{\mu\nu}=\bar{h}_{\mu\nu}$ and we can omit the bars. In the frame of observers with 4-velocity $u^\mu$ the full set of conditions that fix the \emph{transverse traceless (TT) gauge} is then
 +
\begin{equation}
 +
\pa_\mu {h^{\mu}}_{\nu}=0,\quad
 +
h_{0\mu}=0,\quad {h^{\mu}}_{\mu}=0.
 +
\end{equation}
 +
 +
Simplest solutions of the vacuum wave equation are plane waves
 +
\[h_{\mu\nu}=h_{\mu\nu}e^{ik_{\lambda}x^\lambda}.\]
 +
Indeed, substitution into $\square h_{\mu\nu}=0$ yields
 +
\[k^{\lambda}k_{\lambda}\cdot h_{\mu\nu}=0.\]
 +
Thus
 +
 +
* either the wave vector is null $k^{\lambda}k_{\lambda}=0$, which roughly translates as that gravitational waves propagate with the speed of light,
 +
* or $h_{\mu\nu}=0$, which means that in any other (non-TT) coordinate frame, in which metric perturbation is non-zero, it is due to the oscillating coordinate system, while the true gravitational field vanishes.
 +
\end{itemize}
 +
 +
 +
 +
 +
 +
<div id="gw25"></div>
 +
<div style="border: 1px solid #AAA; padding:5px;">
 +
=== Problem 1: Transverse traceless (TT) gauge===
 +
Rewrite the gauge conditions for the TT and Lorenz gauge
 +
\begin{enumerate}
 +
\item in terms of scalar, vector and tensor decomposition;
 +
\item in terms of the metric perturbation for the plane wave solution
 +
\[h_{\mu\nu}=h_{\mu\nu}e^{ik_{\lambda}x^\lambda}
 +
=h_{\mu\nu}e^{i\omega t-ikz}\]
 +
with wave vector $k^{\mu}=(\omega,0,0,k)$ directed along the $z$-axis.
 +
\end{enumerate}
 +
<div class="NavFrame collapsed">
 +
  <div class="NavHead">solution</div>
 +
  <div style="width:100%;" class="NavContent">
 +
    <p style="text-align: left;">\begin{enumerate}
 +
\item The two algebraic conditions are
 +
\begin{align}
 +
0&=u^{\mu}h_{\mu\nu}=h_{0\nu}=(\Phi,-\mathbf{w});\\
 +
0&=h^{\mu}_{\mu}=\Phi-6\Psi,
 +
\end{align}
 +
so we get in the new frame
 +
\[\Phi=\Psi=0,\quad \mathbf{w}=0.\]
 +
Then the remaining Lorenz condition is reduced to
 +
\begin{align*}
 +
&0=\pa_0 h^0_\nu +\pa_\alpha h^\alpha_\nu ;\\
 +
\nu=0:\quad& 0=0;\\
 +
\nu=\beta:\quad&0= \pa_{\alpha}{s^{\alpha}}_{\beta},
 +
\end{align*}
 +
so it takes the form
 +
\begin{equation}
 +
\pa_{\alpha}{s^{\alpha}}_{\beta}=0.
 +
\end{equation}
 +
Thus the only remaining non-trivial component of the perturbation is the tensor one (the traceless part) $s_{\alpha\beta}$, and it is transverse, in the sense that for a plane wave solution $s_{\alpha\beta}\sim e^{i\,k_{\mu}x^{\mu}}$ the metric perturbation is transverse with respect to the wave vector:
 +
\[k^{\alpha}s_{\alpha\beta}=0.\]
 +
\item Transversality implies $h_{\mu 0}=0$, Lorenz gauge condition that $h_{\mu z}=0$, so the only non-zero components are $h_{xx},h_{yy},h_{xy},h_{yx}$. Of those only two are independent due to symmetry $h_{xy}=h_{yx}$ and tracelessness $h_{xx}+h_{yy}=0$.
 +
\end{enumerate}</p>
 +
  </div>
 +
</div></div>
 +
 +
 +
 +
 +
So any plane-wave solution with $k^{\mu}=(\omega,0,0,k)$ in the $z$-direction in the TT gauge has the form
 +
\begin{equation}
 +
h_{\mu\nu}=\bordermatrix{
 +
~&t&x&y&z\cr
 +
t&0&0&0&0\cr
 +
x&0&h_{+}&h_{\times}&0\cr
 +
y&0&h_{\times}&-h_{+}&0\cr
 +
z&0&0&0&0}e^{ikz-i\omega t},
 +
\end{equation}
 +
or more generally any wave solution propagating in the $z$ direction can be presented as
 +
\begin{align}
 +
h_{\mu\nu}(t,z)&=
 +
\bordermatrix{
 +
~&t&x&y&z\cr
 +
t&0&0&0&0\cr
 +
x&0&1&0&0\cr
 +
y&0&0&-1&0\cr
 +
z&0&0&0&0}h_{+}(t-z)
 +
+\bordermatrix{
 +
~&t&x&y&z\cr
 +
t&0&0&0&0\cr
 +
x&0&0&1&0\cr
 +
y&0&1&0&0\cr
 +
z&0&0&0&0}h_{\times}(t-z)=\\
 +
&=e^{(+)}_{\mu\nu}h_{+}(t-z)
 +
+e^{(\times)}_{\mu\nu}h_{\times}(t-z).
 +
\end{align}
 +
Here $h_{+}$ and $h_\times$ are the amplitudes of the two independent components with linear polarization, and $e^{(\times)}_{\mu\nu},e^{(+)}_{\mu\nu}$ are the corresponding polarization tensors.
 +
 +
 +
 +
 +
 +
<div id="gw26"></div>
 +
<div style="border: 1px solid #AAA; padding:5px;">
 +
=== Problem 2: Two polarizations===
 +
Show that $e^{(\times)}_{\mu\nu}$ and $e^{(+)}_{\mu\nu}$ transform into each other under rotation by $\pi/8$
 +
<div class="NavFrame collapsed">
 +
  <div class="NavHead">solution</div>
 +
  <div style="width:100%;" class="NavContent">
 +
    <p style="text-align: left;">The two-dimensional transformation matrix $\Phi$ for rotation by $\pi/4$ and the polarization tensors are
 +
\[\Phi=(x^{\mu}_{,\nu})=\frac{1}{\sqrt{2}}
 +
\begin{pmatrix}1&1\\-1&1 \end{pmatrix},\quad
 +
(e^{(+)}_{\mu\nu})
 +
=\begin{pmatrix}1&0\\0&-1\end{pmatrix},\quad
 +
(e^{(\times)}_{\mu\nu})
 +
=\begin{pmatrix}0&1\\1&0\end{pmatrix}.\]
 +
In matrix notation the second rank tensor transformation law is $e'=\Phi\Phi e$, and acting twice on each of the polarization vectors we have
 +
\begin{align*}
 +
\Phi\Phi e^{(+)}&=\frac12 \begin{pmatrix}1&1\\-1&1 \end{pmatrix}
 +
\begin{pmatrix}1&1\\-1&1 \end{pmatrix}
 +
\begin{pmatrix}1&0\\0&-1 \end{pmatrix}
 +
=-\begin{pmatrix}0&+1\\1&0 \end{pmatrix}
 +
=-e^{(\times)},\\
 +
\Phi\Phi e^{(\times)}&=\frac12 \begin{pmatrix}1&1\\-1&1 \end{pmatrix}
 +
\begin{pmatrix}1&1\\-1&1 \end{pmatrix}
 +
\begin{pmatrix}0&+1\\1&0 \end{pmatrix}
 +
=+\begin{pmatrix}1&0\\0&-1 \end{pmatrix}
 +
=+e^{(+)}.
 +
\end{align*}</p>
 +
  </div>
 +
</div></div>
 +
 +
 +
 +
 +
 +
<div id="gw27></div>
 +
<div style="border: 1px solid #AAA; padding:5px;">
 +
=== Problem 3: Plane wave TT gauge transformation===
 +
Consider the plane wave solution of the wave equation in the Lorenz gauge:
 +
\[\bar{h}_{\mu\nu}
 +
=A_{\mu\nu}e^{i\,k_{\lambda}x^\lambda},
 +
\quad k^\mu k_\mu =0.\]
 +
\begin{enumerate}
 +
\item Show that the TT gauge is fixed by the coordinate transformation $x\to x+\xi$ with
 +
\begin{align}
 +
\xi_{\mu}&=B_{\mu}e^{ik_\lambda x^\lambda};\\
 +
B_{\lambda}
 +
&=-\frac{A_{\mu\nu}l^\mu l^\nu}
 +
{8i \omega^4}k_\lambda
 +
-\frac{A^{\mu}_{\mu}}{4i\omega^2}l_\lambda
 +
+\frac{1}{2i\omega^2}A_{\lambda\mu}l^\mu;\\
 +
&\text{where}\quad
 +
k^\mu=(\omega,\mathbf{k}),\quad
 +
l^{\mu}=(\omega,-\mathbf{k}).
 +
\end{align}
 +
[Padm. p \textsection 9.3, p.403 (wrong signs!), also see  MTW Ex.35.1 for similar formulation; \emph{the solution can probably be derived if we introduce the null frame $k$ and $l$ (the other two spacial basis vectors are not needed) and look for solution in the form
 +
$B=C_1 k+ C_2 l +C_{3} w$.}]
 +
\item What is the transformation to the Lorenz gauge for arbitrary gravitational wave in vacuum?
 +
\end{enumerate}
 +
<div class="NavFrame collapsed">
 +
  <div class="NavHead">solution</div>
 +
  <div style="width:100%;" class="NavContent">
 +
    <p style="text-align: left;">\begin{enumerate}
 +
\item The gauge transformation for $\bar{h}_{\mu\nu}$ and therefore $A_{\mu\nu}$ is
 +
\[A_{\mu\nu}\to A_{\mu\nu}'
 +
=A_{\mu\nu}-2\pa_{(\mu}\xi_{\nu)}
 +
+\eta_{\mu\nu}\pa_{\lambda}\xi^\lambda,\]
 +
so
 +
\begin{equation}
 +
A_{\mu\nu}\to A_{\mu\nu}'
 +
=A_{\mu\nu}-2ik_{(\mu}B_{\nu)}
 +
+i\eta_{\mu\nu}k_{\lambda}B^\lambda,
 +
\end{equation}
 +
and the TT gauge conditions take the form
 +
\begin{align}
 +
0&={{A'}^{\mu}}_{\mu}={A^{\mu}}_{\mu}+2ik^\mu B_\mu;\\
 +
0&={A'}_{0\;\mu}=A_{0\mu}-ik_0 B_{\mu}
 +
-ik_\mu B_0  +i\delta^0_\mu \; k_\lambda B^\lambda.
 +
\end{align}
 +
Also remember that Lorenz gauge implies $k_{\mu}A^{\mu\nu}=0$, thus $k^{\alpha}A_{\alpha\mu}=-\omega A_{0\mu}$, and therefore
 +
\begin{equation}
 +
l^{\mu}k_{\mu}=2\omega^2,\quad
 +
A_{\mu\nu}l^{\mu}=2\omega A_{0\mu},\quad
 +
k^{\lambda}B_{\lambda}
 +
=-\frac{1}{2i}{A^{\lambda}}_{\lambda}.
 +
\end{equation}
 +
Plugging this into the gauge conditions, one can see that the first one (tracelessness) is obeyed immediately, and the second one (transversality) after a little bit of more algebra.
 +
\item The general first-order vacuum solution of Einstein's equations in the Lorenz gauge can be presented through its spatial Fourier transform (temporal part is integrated over due to fixed dispersion relation)
 +
\[h_{\mu\nu}(x)=\int d^{3}k h_{\mu\nu}(\mathbf{k})
 +
e^{ik_{\lambda}x^{\lambda}},\quad\text{with}\quad
 +
k^\mu k_\mu =0,\quad\Leftrightarrow\quad
 +
k_{0}^{2}\equiv \omega^2 =\mathbf{k}^2.\]
 +
Then the TT gauge will be fixed by the coordinate fransformation
 +
\[\xi^\mu =\int d^{3}k B^{\mu}[h_{\mu\nu}(\mathbf{k})]
 +
e^{ik_{\lambda}x^{\lambda}}.\]
 +
Working with individual plane-wave solutions is equivalent to working in the full Fourier space.
 +
\end{enumerate}</p>
 +
  </div>
 +
</div></div>
 +
 +
 +
 +
 +
<div id="gw28></div>
 +
<div style="border: 1px solid #AAA; padding:5px;">
 +
=== Problem 3: Curvature of a plane wave===
 +
Consider the plane-wave solution, in which
 +
\[R_{\mu\nu\rho\sigma}
 +
=C_{\mu\nu\rho\sigma}e^{ik_{\lambda}x^{\lambda}}.\]
 +
\begin{enumerate}
 +
\item Using the Bianchi identity, show that all components of the curvature tensor can be expressed through $R_{0\alpha0\beta}$;
 +
\item Show that in the coordinate frame such that $k^\mu =(k,0,0,k)$ is directed along the $z$-axis the only possible nonzero components are $R_{0x0x}$, $R_{0y0y}$ and $R_{0x0y}$, obeying $R_{0x0x}=-R_{0y0y}$, leaving only two independent non-zero components;
 +
\item in the TT gauge (denoted by the superscript $TT$)
 +
\[R_{0\alpha 0\beta}
 +
=-\tfrac{1}{2}\pa_0^2 g_{\alpha\beta}^{TT};\]
 +
\end{enumerate}
 +
[based on Padm p403].
 +
<div class="NavFrame collapsed">
 +
  <div class="NavHead">solution</div>
 +
  <div style="width:100%;" class="NavContent">
 +
    <p style="text-align: left;">\begin{enumerate}
 +
\item The covariant derivatives in the  Bianchi identity
 +
\[R_{abcd;e}+R_{abec;d}+R_{abde;c}=0\]
 +
in the first order can be replaced by partial derivatives. Then for a plane-wave solution
 +
\[k_{e}R_{abcd}+k_{d}R_{abec}+k_{c}R_{abde}=0.\]
 +
By setting $\{eabcd\}=\{0\alpha 0\beta \gamma\}$ we get ($k_0\equiv \omega \neq 0$ for a null vector)
 +
\[R_{\alpha 0\beta \gamma}
 +
=-\frac{k_\beta}{\omega}R_{\alpha 0\gamma 0}
 +
+\frac{k_\gamma}{\omega}R_{\alpha 0\beta 0},\]
 +
and by setting $\{eabcd\}=\{0\alpha \beta \gamma\delta \}$ and using the previous result, that
 +
\[R_{\alpha\beta\gamma\delta}=
 +
\frac{k_\alpha k_\gamma}{\omega^2}
 +
R_{0\beta 0\gamma}
 +
+\big(\text{three terms restored by symmetry}\big).\]
 +
\item From vacuum Einstein equations
 +
\[0=R^{\lambda \mu\lambda \nu}
 +
=R_{0\mu 0\nu}-R_{\alpha\mu\alpha\nu},\]
 +
expressing everything through $R_{0\alpha 0\beta}$, we get
 +
\begin{align}
 +
00:&\quad R_{0\alpha 0\alpha}=0;\\
 +
0\beta:&\quad k_{\alpha} R_{0\alpha 0\beta}=0;\\
 +
\beta\gamma:&\quad 0=0.
 +
\end{align}
 +
The $00$ equation translates to the wave being traceless, and the $0\beta$ equation to it being transverse: in the chosen frame
 +
\[R_{0x0x}+R_{0y0y}=0,\qquad R_{0z0\alpha}=0.\]
 +
\item In the TT gauge $g_{00}$ and $g_{0\alpha}$ are zero, so in the first order
 +
\[R_{0\alpha 0\beta}=\frac{1}{2}\big(
 +
\pa_0 \pa_\beta g_{0\alpha}
 +
+\pa_\alpha \pa_0 g_{0\beta}
 +
-\pa_0 \pa_0 g_{\alpha\beta}
 +
-\pa_\alpha \pa_\beta g_{00}\big)
 +
=-\tfrac12 \pa_0^2 g_{\alpha\beta}.\]
 +
As the curvature tensor is gauge invariant, the last relation allows one to calculate the metric components in the TT gauge $g_{\alpha\beta}^{TT}$ without the prior gauge fixing.
 +
\end{enumerate}</p>
 +
  </div>
 +
</div></div>

Revision as of 14:25, 26 December 2012



From now on we consider only vacuum solutions. Suppose we use the Lorenz gauge. As shown above, we still have the freedom of coordinate transformations with $\square \xi^\mu =0$, which preserve the gauge. So, let us choose some arbitrary (timelike) field $u^\mu$ and, in addition to the Lorenz gauge conditions, demand that the perturbation is also transverse $u^{\mu}\bar{h}_{\mu\nu}=0$ with regard to it plus that it is traceless $\bar{h}^{\mu}_{\mu}=0$. Then $h_{\mu\nu}=\bar{h}_{\mu\nu}$ and we can omit the bars. In the frame of observers with 4-velocity $u^\mu$ the full set of conditions that fix the \emph{transverse traceless (TT) gauge} is then \begin{equation} \pa_\mu {h^{\mu}}_{\nu}=0,\quad h_{0\mu}=0,\quad {h^{\mu}}_{\mu}=0. \end{equation}

Simplest solutions of the vacuum wave equation are plane waves \[h_{\mu\nu}=h_{\mu\nu}e^{ik_{\lambda}x^\lambda}.\] Indeed, substitution into $\square h_{\mu\nu}=0$ yields \[k^{\lambda}k_{\lambda}\cdot h_{\mu\nu}=0.\] Thus

  • either the wave vector is null $k^{\lambda}k_{\lambda}=0$, which roughly translates as that gravitational waves propagate with the speed of light,
  • or $h_{\mu\nu}=0$, which means that in any other (non-TT) coordinate frame, in which metric perturbation is non-zero, it is due to the oscillating coordinate system, while the true gravitational field vanishes.

\end{itemize}



Problem 1: Transverse traceless (TT) gauge

Rewrite the gauge conditions for the TT and Lorenz gauge \begin{enumerate} \item in terms of scalar, vector and tensor decomposition; \item in terms of the metric perturbation for the plane wave solution UNIQ-MathJax58-QINU with wave vector UNIQ-MathJax10-QINU directed along the UNIQ-MathJax11-QINU-axis. \end{enumerate}



So any plane-wave solution with $k^{\mu}=(\omega,0,0,k)$ in the $z$-direction in the TT gauge has the form \begin{equation} h_{\mu\nu}=\bordermatrix{ ~&t&x&y&z\cr t&0&0&0&0\cr x&0&h_{+}&h_{\times}&0\cr y&0&h_{\times}&-h_{+}&0\cr z&0&0&0&0}e^{ikz-i\omega t}, \end{equation} or more generally any wave solution propagating in the $z$ direction can be presented as \begin{align} h_{\mu\nu}(t,z)&= \bordermatrix{ ~&t&x&y&z\cr t&0&0&0&0\cr x&0&1&0&0\cr y&0&0&-1&0\cr z&0&0&0&0}h_{+}(t-z) +\bordermatrix{ ~&t&x&y&z\cr t&0&0&0&0\cr x&0&0&1&0\cr y&0&1&0&0\cr z&0&0&0&0}h_{\times}(t-z)=\\ &=e^{(+)}_{\mu\nu}h_{+}(t-z) +e^{(\times)}_{\mu\nu}h_{\times}(t-z). \end{align} Here $h_{+}$ and $h_\times$ are the amplitudes of the two independent components with linear polarization, and $e^{(\times)}_{\mu\nu},e^{(+)}_{\mu\nu}$ are the corresponding polarization tensors.



Problem 2: Two polarizations

Show that $e^{(\times)}_{\mu\nu}$ and $e^{(+)}_{\mu\nu}$ transform into each other under rotation by $\pi/8$



Problem 3: Plane wave TT gauge transformation

Consider the plane wave solution of the wave equation in the Lorenz gauge: \[\bar{h}_{\mu\nu} =A_{\mu\nu}e^{i\,k_{\lambda}x^\lambda}, \quad k^\mu k_\mu =0.\] \begin{enumerate} \item Show that the TT gauge is fixed by the coordinate transformation UNIQ-MathJax31-QINU with \begin{align} \xi_{\mu}&=B_{\mu}e^{ik_\lambda x^\lambda};\\ B_{\lambda} &=-\frac{A_{\mu\nu}l^\mu l^\nu} {8i \omega^4}k_\lambda -\frac{A^{\mu}_{\mu}}{4i\omega^2}l_\lambda +\frac{1}{2i\omega^2}A_{\lambda\mu}l^\mu;\\ &\text{where}\quad k^\mu=(\omega,\mathbf{k}),\quad l^{\mu}=(\omega,-\mathbf{k}). \end{align} [Padm. p \textsection 9.3, p.403 (wrong signs!), also see MTW Ex.35.1 for similar formulation; \emph{the solution can probably be derived if we introduce the null frame $k$ and $l$ (the other two spacial basis vectors are not needed) and look for solution in the form $B=C_1 k+ C_2 l +C_{3} w$.}] \item What is the transformation to the Lorenz gauge for arbitrary gravitational wave in vacuum? \end{enumerate}



Problem 3: Curvature of a plane wave

Consider the plane-wave solution, in which \[R_{\mu\nu\rho\sigma} =C_{\mu\nu\rho\sigma}e^{ik_{\lambda}x^{\lambda}}.\] \begin{enumerate} \item Using the Bianchi identity, show that all components of the curvature tensor can be expressed through UNIQ-MathJax39-QINU; \item Show that in the coordinate frame such that UNIQ-MathJax40-QINU is directed along the UNIQ-MathJax41-QINU-axis the only possible nonzero components are UNIQ-MathJax42-QINU, UNIQ-MathJax43-QINU and UNIQ-MathJax44-QINU, obeying UNIQ-MathJax45-QINU, leaving only two independent non-zero components; \item in the TT gauge (denoted by the superscript UNIQ-MathJax46-QINU) UNIQ-MathJax67-QINU \end{enumerate} [based on Padm p403].