Could somebody disprove the following concept of a Maxwell's demon: Let's say we have a very small container, the size and a form of a nanotube, similar to ellipse, and only two gas molecules randomly flying inside. According to probability we will quite often find molecules in the different ends of a such container. If these molecules will have different speed and energy (is it likely to happen?) then we will get the hotter and the colder molecules in the different ends of the container, what is one of the conditions for a Maxwell's demon to start his work. If hotter molecule will stay at one end of the container for a while it will bounce to its walls and heat it. And contra, the cold molecule at the same time will bounce at the other end of nanotube (container) and cool it. In this way a temperature difference between the different ends of nanotube will be created. Then we could imagine a two thermoelectric wires of an atomic thickness which are connected to both ends of the nanotube and to a small electric generator. If we need to scale this system up, we could take a two largest molecules which could only exist in nature and conduct the experiment in a deep space that gravitation and weight of molecules wouldn't interrupt us. Does nature put some fundamental restriction on size of molecules (which could have a thermal motion)? And also question: is probability of a thermal fluctuations is always the same on all closed systems? Or there could be some systems with increased probability?