Superparamagnetism is essentially, ferromagnetism with single domain. It is considered paramagnetism, because if you apply external magnetic field all particles can be magnetised, except their susceptibility is much larger than in paramagnets. 

Superparamagnetism is due to reduction of size of the systems ferromagnetic and ferrimagnetic. The reduction of size  causes that the systems became en monodomain and they lost its stability of the magnetization. In this point, the competence between the anisotropy and thermal energy are comparable.

The mains characteristics of a superparamagnet are:

1. Magnetic moment than highest that the paramagnet

2. it not have hysteresis curves.

3. Additionally, these nanoparticles are single-domain particles

As described correctly by Stefan and Yeni, superparamagnetic materials are actually ferromagnets. If a ferromagnetic material is synthesized as a sufficiently small particle, it will only have one domain. Macroscopically, however, the domains of the individual particles will be randomly distributed and there will be no net magnetization of the powder/suspension. Under the influence of an external magnetic field, the magnetization directions of all particles will align and a net magnetization results. This behavior is analogous to the description of a paramagnet. However, the individual entities are not atomic dipoles, but those of the nanoparticles comprising several thousand atoms. Consequently, the individual magnetic moments are much larger than in (atomic) paramagnets. Hence the term "super"-paramagnet.

If I'm not mistaken, this behavior was first characterized by Bean and Livingston in the Journal of Applied Physics in 1959.

One important point not mentioned above is that superparamagnetism is a time/temperature dependent phenomenon.

Essentially, a ferromagnetic or ferrimagnetic particle behaves like a superparamagnetic when thermal fluctuations are large enough to over overcome the particle's coercivity and randomize its moment. For example, at room temperature a (10 nm)^3 magnetite grain will be SP, but if you cool it to, say, 10 K it will behave like an SD particle. This is because at low temperatures the magnitude of thermal fluctuations is much lower and cannot overcome the grain's coercivity. 

Similarly, SP behavior is time dependent. This is because thermal fluctuations are a statistical phenomenon. So for very short timescales the likelihood of a fluctuation large enough to randomize the moment is low, but the probability increases if you wait longer (the longer you wait at a bus stop the more likely a bus will come).

For magnetite, the SP/SD threshold is about (30 nm)^3 at room temperature and for a measurement time of ~100 s. Let's say we have a slightly smaller particle, such that is behaves SP when measured over 100 s (maybe ~(27-28 nm)^3, very rough guess). If we were to measure very quickly, say 1 s, then it would behave SD and not SP. (There are very rough numbers to serve as an illustration). It is this time dependence that we take advantage of when measuring frequency dependence of magnetic susceptibility. 

Paramagnetic substances are weak magnets. They do not have their own magnetic order. Strong magnetic materials (including the ferromagnetic or ferrimagnetic) below the Curie point have a spontaneous magnetic order. Superparamagnetic particle (SPP) is single-domain particle of ferro- or ferrimagnetic substance, which consists in itself of many atoms (ions) coupling by strong exchange interaction, which is absent in the paramagnetic material. Therefore, the magnetic moments of the atoms of a ferromagnet (or elementary cells of ferrimagnetic) directed collinear. The thermal fluctuations randomly change the orientation of the magnetic moment of individual atoms of a paramagnetic material independently from each other. In SPP, the magnetic moments of all atoms oscillate synchronously, and the overall (total) moment of a particle as whole behaves under the influence of temperature as a single magnetic moment. But in the ensemble of such particles, magnetic moments of individual SPP behave independently from each other, as the atoms of a paramagnetic material. SPP is some analogue of the atom of paramagnetic substance, or a certain “superatom”.

The role of the Curie point TC in the superparamagnetism plays the blocking temperature TB < TC. The analogy between paramagnetism and superparamagnetism is not complete. Some differences were mentioned by other authors earlier. There's more. There are other. The magnetic moments of paramagnetic atoms can have almost any orientation in the space, while the superparamagnetic particles have the most stable states (directions) associated with the axes of easy magnetization due to magnetic anisotropy of crystalline particles or to their shape anisotropy.

Please, see also the following discussion:

https://www.researchgate.net/post/How_can_differentiate_ferromagnetism_Paramagnetism_and_superparamagnetism_in_RT-VSM_magnetism_study

You're getting good answers here but let me try to clarify one point.

For single atoms or objects with relatively small total angular momentum J, you probably know that there is only a finite number of states (2J+1) available. In a magnetic field these are energetically split (Zeeman splitting) and their occupation probability is governed by Boltzmann factors. Put everything together, and you obtain the Brillouin function BJ(H,T) which describes how M(H,T) will behave. For different J, these BJ(T) essentially differ in slope at H=0.

When Langevin invented Langevin paramagnetism, the angular momentum quantisation was not yet known. Nevertheless, he had the idea to apply the same kind of reasoning for a classical magnetic moment (which is free to have ANY orientation). the resulting magnetization function is L(H,T), the Langevin function. It is at the same time the Brillouin function in the limit J->infinity (i.e. the quantization is quasi-continuous in the limit of large angular momenta).

In a nanoparticle with hundreds or thousands of atoms, the total spin is a large number and therefore it behaves like a classical spin [L(H,T]] with large moment. Since the susceptibility scales as the square of the moment, it is larger in superparamagnets than with the same average number density of atoms, each responding individually to the applied field.

In the computation of the Langevin function, an underlying assumption is that thermodynamic equilibrium is given in the (ensemble of) particles. Then M(H,T) is automatically a single valued function and there is no hysteresis.

It takes terms of magnetic anisotropy to generate hyseresis. If large enough, the associated anisotropy energy barriers will (massively) slow down the time it takes to reach thermodynamic equilibrium (the time constant is called relaxation time). Roughly said, if your measurement is "faster than the relaxation" time, then you will observe hysteresis. (see previous comments). Usually, when there is no hysteresis (or other signatures of fast fluctuations, as e.g. in Mößbauer spectroscopy) we call the system superparamagnetic (= fluctuating particle moments, near equilibrium on the time scale of the measurement process), and when hysteresis turns up (by lowering the temperature), then we say that the particles are (partially) blocked.

However, the presence of anisotropy will also modify the shape of the equilibrium  magnetization curves!  This can be an appreciable effect which I see rarely being taken into account or even discussed (modeling gets pretty much more laborious, then).

Finally, different (anisotropy) mechanisms exist which may lead to hysteresis. Among these would be magnetocrystalline (volume and surface) anisotropies, shape anisotropy (nonspherical particles) and dipolar interaction between particles. In case of magnetic particles embedded in some host material (matrix), host mediated interactions will also play their role, if present.

Thank you very much to all of you, It was very nice and informative discussion , helped me to remove all  my confusion.My humble thanks to all of you.

Dear Aiman Shamim, since I do not know which aspects you already understood and which ones remained unclear, I propose that you ask specific questions. It will probably be more easy then to give dedicated answers.

(note added: Aiman seems to have modified his last post, this is why the above answer of mine appears to be quite out of context....)

KF writes: “Nevertheless, he had the idea to apply the same kind of reasoning for a classical magnetic moment (which is free to have ANY orientation). The resulting magnetization function is L(H,T), the Langevin function. It is at the same time the Brillouin function in the limit J->infinity (i.e. the quantization is quasi-continuous in the limit of large angular momenta).

In a nanoparticle with hundreds or thousands of atoms, the total spin is a large number and therefore it behaves like a classical spin [L(H,T]] with large moment”.

It is the reason why I wrote: The magnetic moments of paramagnetic atoms can have ALMOST ANY orientation in the space.

How could define, localized surface plasmon resonance of metal nanoparticles ? how it is different from molecules having conjugation in them?

Aiman, please make this a separate thread.

Mainly superparamagnetism is occurred for grains with size less than 5 nm and its magnetic loop at room temperature is not saturated with zero coersivity (Hc=0). Also, the superparamagnetism can be defined by the magnetization of a grain of volume V = (5 nm)3 and anisotropy of 4 X 106 erg/cm3 is thermally unstable at room temperature. On the other hand, for particles whose radius is smaller than about 1 nm the external magnetic field is unable to produce saturation, because it cannot compete against thermal excitations. Note that true ferromagnetism is limited to infinite magnets, because thermal excitations in finite magnets cause

the net moment to fluctuate between opposite directions and yield, ultimately, a zero spontaneous magnetization