The momentum flux density is (minus) the Maxwell stress tensor (Section 8.2.3),
T_{i j}=\epsilon_{0}\left(E_{i} E_{j}-\frac{1}{2} \delta_{i j} E^{2}\right)+\frac{1}{\mu_{0}}\left(B_{i} B_{j}-\frac{1}{2} \delta_{i j} B^{2}\right),
(Eq. 8.17) so the momentum radiated per unit (retarded) time is
T_{i j} \equiv \epsilon_{0}\left(E_{i} E_{j}-\frac{1}{2} \delta_{i j} E^{2}\right)+\frac{1}{\mu_{0}}\left(B_{i} B_{j}-\frac{1}{2} \delta_{i j} B^{2}\right) (8.17)
\frac{d p }{d t_{r}}=-\oint \frac{1}{\left(\partial t_{r} / \partial t\right)} \overleftrightarrow{T} \cdot d A,
where (Eq. 10.78)
\frac{\partial t_{r}}{\partial t}=\frac{ᴫ c}{ ᴫ \cdot u } (10.78)
\frac{\partial t_{r}}{\partial t}=\frac{ᴫ c}{ ᴫ \cdot u }=\left(1-\frac{\hat{ᴫ } \cdot v }{c}\right)^{-1}.
(This factor is 1, when v = 0; it can be ignored in deriving the Larmor formula, but it does contribute to the momentum radiated.) The integration is over a large spherical surface centered on the charge:
d A =ᴫ^{2} \sin \theta d \theta d \phi \hat{ᴫ }
As in the case of the power radiated, only the radiation fields contribute (Eq. 11.66):
E _{ rad }=\frac{q}{4 \pi \epsilon_{0}} \frac{ᴫ}{(ᴫ \cdot u )^{3}}[ ᴫ \times( u \times a )] (11.66)
E _{ rad }=\frac{q}{4 \pi \epsilon_{0}} \frac{ ᴫ }{( ᴫ \cdot u )^{3}}[ ᴫ \times( u \times a )], \quad B _{ rad }=\frac{1}{c} \hat{ᴫ } \times E _{ rad }.
Thus
\overleftrightarrow{T} \cdot d A =\epsilon_{0}\left[ E ( E \cdot d A )-\frac{1}{2} E^{2} d A \right]+\frac{1}{\mu_{0}}\left[ B ( B \cdot d A )-\frac{1}{2} B^{2} d A \right]=-\frac{1}{2}\left[\epsilon_{0} E^{2}+\frac{1}{\mu_{0}} B^{2}\right] d A =-\epsilon_{0} E^{2} d A
(note that B · dA = 0, and, for radiation fields, E \cdot d A =0,\left(1 / \mu_{0}\right) B^{2}=\left(1 / \mu_{0} c^{2}\right)(\hat{ ᴫ } \times E ) \cdot(\hat{ ᴫ } \times E )=\left.\epsilon_{0} \hat{ᴫ } \cdot[ E \times(\hat{ ᴫ } \times E )]=\epsilon_{0} \hat{ᴫ } \cdot\left[\hat{ ᴫ } E^{2}- E (\hat{ ᴫ } \cdot E )\right]=\epsilon_{0} E^{2}\right). So
\frac{d p }{d t_{r}}=\epsilon_{0} \oint\left(\frac{ ᴫ \cdot u }{ᴫ c }\right) E^{2} d A =\frac{q^{2}}{16 \pi^{2} \epsilon_{0} c} \oint \frac{ᴫ}{( ᴫ \cdot u )^{5}}[ ᴫ \times( u \times a )]^{2} d A.
We expand the integrand to first order in \beta \equiv v / c: u =c \hat{ ᴫ }- v =c(\hat{ ᴫ}- \beta ), ᴫ \cdot u =c ᴫ (1-\hat{ ᴫ } \cdot \beta ),
\frac{1}{( ᴫ \cdot u )^{5}}=\frac{1}{ᴫ{ }^{5} c^{5}}[1+5(\hat{ ᴫ } \cdot \beta )].
ᴫ \times( u \times a )= u ( ᴫ \cdot a )- a ( ᴫ \cdot u )=ᴫ c(\hat{ ᴫ }- \beta )(\hat{ᴫ } \cdot a )- a ᴫ c(1-\hat{ ᴫ } \cdot \beta )=ᴫ c[\hat{ ᴫ}(\hat{ ᴫ } \cdot a )- a + a (\hat{ ᴫ } \cdot \beta )- \beta (\hat{ᴫ} \cdot a )].
[ ᴫ \times( u \times a )]^{2}=ᴫ^{2} c^{2}\left[(\hat{ᴫ } \cdot a )^{2}-2(\hat{ᴫ } \cdot a )^{2}+a^{2}+2(\hat{ ᴫ } \cdot a )^{2}(\hat{ ᴫ } \cdot \beta )-2(\hat{ ᴫ } \cdot \beta )(\hat{ᴫ } \cdot a )^{2}-2 a^{2}(\hat{ᴫ } \cdot \beta )\right.
+2( a \cdot \beta )(\hat{ᴫ } \cdot a )]=ᴫ^{2} c^{2}\left[a^{2}-(\hat{ᴫ } \cdot a )^{2}-2 a^{2}(\hat{ᴫ } \cdot \beta )+2(\hat{ ᴫ } \cdot a )( a \cdot \beta )\right].
\frac{ᴫ}{( ᴫ \cdot u )^{5}} \quad[ r \times( u \times a )]^{2}=\frac{1}{ᴫ^{2} c^{3}}[1+5(\hat{ ᴫ } \cdot \beta )]\left[a^{2}-(\hat{ ᴫ} \cdot a )^{2}-2 a^{2}(\hat{ᴫ } \cdot \beta )+2(\hat{ ᴫ } \cdot a )( a \cdot \beta )\right]
=\frac{1}{ᴫ^{2} c^{3}}\left[a^{2}-(\hat{ ᴫ } \cdot a )^{2}-2 a^{2}(\hat{ ᴫ } \cdot \beta )+2(\hat{ᴫ } \cdot a )( a \cdot \beta )+5(\hat{ ᴫ } \cdot \beta ) a^{2}-5(\hat{ᴫ} \cdot \beta )(\hat{ ᴫ } \cdot a )^{2}\right]
=\frac{1}{ᴫ^{2} c^{3}}\left[a^{2}-(\hat{ ᴫ } \cdot a )^{2}+3 a^{2}(\hat{ ᴫ } \cdot \beta )+2(\hat{ᴫ } \cdot a )( a \cdot \beta )-5(\hat{ ᴫ } \cdot \beta )(\hat{ ᴫ } \cdot a )^{2}\right]
=\frac{1}{ᴫ^{2} c^{3}}\left[(\hat{ᴫ} \times a )^{2}+\hat{ ᴫ } \cdot\left[3 a^{2} \beta +2( a \cdot \beta ) a \right]-5(\hat{ ᴫ } \cdot \beta )(\hat{ ᴫ } \cdot a )^{2}\right].
To integrate the first term we set the polar axis along a, so (\hat{ ᴫ } \times a )^{2}=a^{2} \sin ^{2} \theta \text {, while } \hat{ ᴫ }=\sin \theta \cos \phi \hat{ x }+\sin \theta \sin \phi \hat{ y }+\cos \theta \hat{ z }. The \phi integral kills the \hat{ x } \text { and } \hat{ y } components, leaving
\oint \frac{1}{ᴫ^{2}}(\hat{ ᴫ } \times a )^{2} d A =2 \pi a^{2} \hat{ z } \int_{0}^{\pi} \sin ^{3} \theta \cos \theta d \theta=0.
[Note that if v = 0 then d p / d t_{r}= 0—a particle instantaneously at rest radiates no momentum. That’s why we had to carry the expansion to first order in \beta , whereas in deriving the Larmor formula we could afford to set v=0.] The second term is of the form (\hat{ᴫ} \cdot g ) , for a constant vector g. Setting the polar axis along g, so \hat{ ᴫ} \cdot g =g \cos \theta , the x and y components again vanish, leaving
\oint \frac{1}{ᴫ^{2}}(\hat{ᴫ} \cdot g ) d A =2 \pi g \hat{ z } \int_{0}^{\pi} \cos ^{2} \theta \sin \theta d \theta=\frac{4 \pi}{3} g.
The last term involves (\hat{ ᴫ} \cdot \beta )(\hat{ ᴫ} \cdot a )^{2} ; this time we orient the polar axis along a and let v lie in the xz plane:
\beta =\beta_{x} \hat{ x }+\beta_{z} \hat{ z }, \text { so }(\hat{ ᴫ } \cdot \beta )=\beta_{x} \sin \theta \cos \phi+\beta_{z} \cos \theta . Then
\oint \frac{1}{ᴫ^{2}}(\hat{ ᴫ} \cdot \beta )(\hat{ ᴫ } \cdot a )^{2} d A =a^{2} \int\left(\beta_{x} \sin \theta \cos \phi+\beta_{z} \cos \theta\right) \cos ^{2} \theta[\sin \theta \cos \phi \hat{ x }+\sin \theta \sin \phi \hat{ y }+\cos \theta \hat{ z }] \sin \theta d \theta d \phi
=a^{2}\left(\beta_{x} \hat{ x } \int \sin ^{3} \theta \cos ^{2} \theta \cos ^{2} \phi d \theta d \phi+\beta_{z} \hat{ z } \int \cos ^{4} \theta \sin \theta d \theta d \phi\right)=\frac{4 \pi}{15} a^{2}\left(\beta_{x} \hat{ x }+3 \beta_{z} \hat{ z }\right)=\frac{4 \pi}{15}\left(a^{2} \beta +2( a \cdot \beta ) a \right).
Putting all this together,
\frac{d p }{d t_{r}}=\frac{q^{2}}{16 \pi^{2} \epsilon_{0}} \frac{1}{c^{4}}\left\{\frac{4 \pi}{3}\left[2 a ( a \cdot \beta )+3 a^{2} \beta \right]-5 \frac{4 \pi}{15}\left[a^{2} \beta +2( a \cdot \beta ) a \right]\right\}=\frac{\mu_{0} q^{2}}{6 \pi c^{3}} a^{2} v.
The angular momentum radiated is
\frac{d L }{d t_{r}}=-\oint \frac{1}{\left(\partial t_{r} / \partial t\right)}( ᴫ \times \overleftrightarrow{T}) \cdot d A.
Because of the “extra” ᴫ in the integrand, it seems at first glance that the radiation fields alone will produce a result that grows without limit (as ᴫ → ∞); however, the coe!cient of this term is precisely zero: (ᴫ \times \overleftrightarrow{T}) . d A =-\epsilon_{0} E^{2}( ᴫ \times d A )=0 (for radiation fields). The finite contribution comes from the cross terms, in which one field (\operatorname{in} \overleftrightarrow{T}) is a radiation field and the other a Coulomb field \left( E _{\text {coul }} \equiv \frac{q}{4 \pi \epsilon_{0}} \frac{ᴫ}{( ᴫ \cdot u )^{3}}\left(c^{2}-v^{2}\right) u \right) :
T_{i j}=\epsilon_{0}\left(E_{i}^{(r)} E_{j}^{(c)}+E_{i}^{(c)} E_{j}^{(r)}-\delta_{i j} E ^{(r)} \cdot E ^{(c)}\right)+\frac{1}{\mu_{0}}\left(B_{i}^{(r)} B_{j}^{(c)}+B_{i}^{(c)} B_{j}^{(r)}-\delta_{i j} B ^{(r)} \cdot B ^{(c)}\right).
This time
\overleftrightarrow{T} \cdot d A =\epsilon_{0}\left[ E ^{(r)}\left( E ^{(c)} \cdot d A \right)-\left( E ^{(r)} \cdot E ^{(c)}\right) d A \right]-\frac{1}{\mu_{0}}\left( B ^{(r)} \cdot B ^{(c)}\right) d A,
so
( ᴫ \times \overleftrightarrow{T}) \cdot d A =\epsilon_{0}\left( ᴫ \times E ^{(r)}\right)\left( E ^{(c)} \cdot d A \right).
Thus
\frac{d L }{d t_{r}}=-\frac{q^{2} c}{16 \pi^{2} \epsilon_{0} \gamma^{2}} \oint \frac{ᴫ}{( ᴫ \cdot u )^{5}}\{ᴫ \times[ ᴫ \times( u \times a )]\} u \cdot d A , \text { where } \gamma \equiv \frac{1}{\sqrt{1-(v / c)^{2}}}.
Now, \{ᴫ \times[ ᴫ \times( u \times a )]\}= ᴫ [ ᴫ \cdot( u \times a )]-( u \times a ) ᴫ^{2} , and ( u \times a )=c(\hat{ ᴫ}- \beta ) \times a =c[(\hat{ᴫ } \times a )-( \beta \times a )],
\hat{ ᴫ } \cdot( u \times a )=-c \hat{ ᴫ } \cdot( \beta \times a ), \text { so }\{ ᴫ \times[ ᴫ \times( u \times a )]\}=- ᴫ ^{2} c\{(\hat{ ᴫ } \times a )-( \beta \times a )+\hat{ ᴫ }[\hat{ ᴫ } \cdot( \beta \times a )]\}.
Meanwhile u \cdot d A =( u \cdot \hat{ᴫ }) ᴫ^{2} d \Omega \text {, where } d \Omega \equiv \sin \theta d \theta d \phi \text {. Expanding to first order in } \beta,
\frac{d L }{d t_{r}}=-\frac{\mu_{0} q^{2} c^{3}}{16 \pi^{2}} \oint \frac{ᴫ}{( ᴫ \cdot u )^{5}}\left(-ᴫ^{2} c\right)\{(\hat{ ᴫ } \times a )-( \beta \times a )+\hat{ᴫ}[\hat{ ᴫ } \cdot( \beta \times a )]\} ᴫ( ᴫ \cdot u ) d \Omega
=\frac{\mu_{0} q^{2}}{16 \pi^{2}} \int[1+4(\hat{ᴫ} \cdot \beta )]\{(\hat{ ᴫ } \times a )-( \beta \times a )+\hat{ ᴫ }[\hat{ᴫ } \cdot( \beta \times a )]\} d \Omega
\left.=\frac{\mu_{0} q^{2}}{16 \pi^{2}} \int\{(\hat{ ᴫ } \times a )-( \beta \times a )+\hat{ᴫ }[\hat{ ᴫ } \cdot( \beta \times a )]+4(\hat{ ᴫ } \cdot \beta )(\hat{ ᴫ } \times a )]\right\} d \Omega
The first integral is
– a \times \int \hat{ ᴫ } \sin \theta d \theta d \phi=- a \times \int(\sin \theta \cos \phi \hat{ x }+\sin \theta \sin \phi \hat{ y }+\cos \theta \hat{ z }) \sin \theta d \theta d \phi= 0,
the second is
-( \beta \times a ) \int \sin \theta d \theta d \phi=-4 \pi( \beta \times a ),
for the third we set the polar axis along g \equiv \beta \times a and get
g \int \hat{ᴫ} \cos \theta \sin \theta d \theta d \phi=2 \pi g \hat{ z } \int \cos ^{2} \theta \sin \theta d \theta=\frac{4 \pi}{3} g =\frac{4 \pi}{3}( \beta \times a ),
and for the last we put the polar axis along a and let \beta =\beta_{x} \hat{ x }+\beta_{z} \hat{ z } lie in the xz plane:
4 a \int\left(\beta_{x} \sin \theta \cos \phi+\beta_{z} \cos \theta\right) \sin \theta(-\hat{ \phi }) \sin \theta d \theta d \phi
=4 a \int\left(\beta_{x} \sin \theta \cos \phi+\beta_{z} \cos \theta\right)(\sin \phi \hat{ x }-\cos \phi \hat{ y }) \sin ^{2} \theta d \theta d \phi
=-4 a \beta_{x} \hat{ y } \int \cos ^{2} \phi \sin ^{3} \theta d \theta d \phi=-\frac{16 \pi}{3} a \beta_{x} \hat{ y }=\frac{16 \pi}{3}( \beta \times a ).
Putting this all together, we conclude
\frac{d L }{d t_{r}}=\frac{\mu_{0} q^{2}}{16 \pi^{2}}\left[-4 \pi( \beta \times a )+\frac{4 \pi}{3}( \beta \times a )+\frac{16 \pi}{3}( \beta \times a )\right]=\frac{\mu_{0} q^{2}}{6 \pi c}( v \times a )