There is a more serious difficulty with this square-root expression, because it contains an implicit sign ambiguity. In classical physics, such things might not worry us, because the quantities under consideration are ordinary real-valued functions, and we can imagine that we could keep the positive values separate from the negative ones. However, in quantum mechanics, this is not so easy. Part of the reason for this is that quantum wavefunctions are complex, and the two square roots of a complex-number expression do not tend to separate neatly into 'positive' and 'negative' in a globally consistent way (5.4). This should be considered in relation to the fact that quantum mechanics deals with operators acting on complex functions, and things like square roots can lead to essential ambiguities that are not simply resolved by just saying 'take the positive root'.
There is another way of expressing this difficulty. In quantum mechanics, one has to consider that the various possible things that 'might' happen, in a physical situation, can all contribute to the quantum state, and therefore all these alternatives have an influence on whatever it is that does happen. When there is something like a square root involved, each of the two roots has to be considered as a 'possibility', so even an 'unphysical negative energy' has to be considered as a 'physical possibility'. As soon as there is the potential for such a negative-energy state, then there is opened up the likelihood of a spontaneous transition from positive to negative energy, which can lead to a catastrophic instability. In the case of a non-relativistic free particle, we do not have this problem of the possibility of a negative energy, because the positive-definite quantity p2/2m does not have this awkward square root. However, the relativistic expression (m2 + p2)1/2 is more problematic in that we do not normally have a clear-cut procedure for ruling out negative square roots.
It turns out that in the case of a single free particle (or a system of such non-interacting particles), this does not actually cause a real difficulty, because we can restrict attention to superpositions of positive-energy plane-wave solutions of the free Schrödinger equation, which are just those considered in 21.5, and there are no transitions to negative energy states. However, when interactions are present, this is no longer the case. Even for just a single relativistic charged particle in a fixed electromagnetic background field the wavefunction cannot, in general, maintain the condition that it be of positive frequency. In this case, we begin to perceive the tension between the principles of quantum mechanics and those of relativity.
As we shall be seeing in 24.8, the great physicist Paul Dirac found a way to resolve this particular tension. But as a first step, he put forward an ingenious and deeply insightful proposal—his now famous equation for the electron—which got rid of the troublesome square root in a marvellous and unexpected way. This subsequently led to a highly original point of view in which negative energies are eliminated, their effects being taken over by what was then a startling prediction: the existence of antiparticles.