Trouble understanding Berry's phase derivation

32 Views Asked by At

Following the the wikipedia article about the adiabatic theorem, and Sakurai's Modern QM, we start with the definition of the geometric phase that we get when doing a loop with a parameter R which depends on time :

$$ \gamma_n(T)=i \int_0^T\left\langle n(R(t)) \mid \frac{\partial}{\partial t}(n(R(t)))\right\rangle d t $$

Let's calculate this time derivative. I use the chain rule with a differential :

$$ \frac{\partial}{\partial t} n(R(t))=dn_{R(t)}\left(R^{\prime}(t)\right) $$

In order to get the notations with a gradient (which are in direction of $R$), we can put the components of the differential, and then those differentials are those of real-valued functions so we can write them using gradients :

$$ =\left(\nabla_R n_1(R(t)) \cdot R^{\prime}(t), \ldots, \nabla_{R}n_m(R(t)) \cdot R^{\prime}(t)\right) $$

And now, if I understand correctly, the authors use the following ambiguous notation :

$$ =\nabla_R n(R(t)) \cdot R^{\prime}(t) $$

Finally, our geometric phase is :

$$ \gamma_n(T)=i \int_0^T\left\langle m(R(t)) \mid \nabla_R n(R(t)) \cdot R^{\prime}(t)\right\rangle d t $$

Here, we're making a change of variables, but honestly I'm not sure how to formalize it. Using $R^{\prime}(t)=d R / d t$, and cancelling $dt$s, we get the usual definition of Berry's phase :

$$ \gamma_m(T)=i \int_{R(0)}^{R(T)}\left\langle n(R) \mid \nabla_R n(R)\right\rangle \cdot dR $$

I think something doesn't add up because what we're integrating now is a sort of "inner product which takes as the second argument a vector of vectors" and that definitely doesn't sound right. The author also mentions that the gradient is an operator, a bit like in QM, so I might miss something obvious. I'm not at ease with the notion of a gradient of a vector field. Is there something wrong with my formulas ? Thanks in advance !