The analysis of Quantum Communications systems developed in the previous chapter ignored thermal noise, sometimes called background noise. In this chapter we will consider such noise, which in practical Quantum Communications is always present, although it is neglected by most researchers, at least nowadays.The general scheme of quantum data transmission seen in Sect. 5.2 is reconsidered in Fig. 8.1, with the purpose of evidencing the parameters that apply in the presence of thermal noise.At transmission Alice "prepares" the quantum system H in one of the coherent states |γ i , i ∈ A, as in the previous chapter. These states are pure ("certain"), but the thermal noise, which originates in the receiver and may be conventionally ascribed to the quantum channel, removes the "certainty" of the states |γ i , and therefore they must be described by density operators. Then, if the transmitted state is |γ i , at reception Bob finds the "noisy" density operator ρ(γ i ), with nominal state |γ i , whose expression will be seen in the next sections.Unfortunately, the analysis and especially the optimization in the presence of thermal noise becomes very difficult, and the reason is due to the representation through density operators, whose mathematical structure is intrinsically nonlinear, while in the classical case thermal noise is simply represented as an additive Gaussian noise.The chapter begins with the quantum representation of thermal noise, where the density operators are expressed as a continuum of coherent states. This representation is formulated in a Hilbert space with infinite dimensions, but to get a numerical evaluation the density operators are approximated by matrices of finite dimensions, so that the quantum decision theory seen in Chaps. 5 and 6 can be applied.For the optimization of the measurement operators, explicit results are available only for binary systems and are provided by Helstrom's theory, which holds also in the presence of noise. For multilevel Quantum Communications systems we can apply the numerical optimization, and especially the square root measurement (SRM), which is suboptimal but gives a good approximation of the system performance.