The ON2 water maser spectra are complex. During the active phase the spectrum
extends from -10 to 
. However, in the spectra the
features appear, as a rule, to form groups clustered in certain spectral
ranges. Isolated features do not appear often. For this reason a method was
employed to fit gaussian curves to the spectra, on the basis of the
velocity of each feature.
First of all, each spectrum observed was considered as the superposition of gaussian curves. The aim of the analysis was to determine the flux and the radial velocity for the peak of each gaussian and the gaussian width at half power. In order to gain a clearer insight into the variation of these parameters (specially the flux and the radial velocity because the width was not always determined with enough precision) we searched in the spectra for an existing emission feature, i.e. a feature which belongs to the same maser condensation. To determine whether a spectral feature corresponded to a maser condensation we assumed that:
1) there is only a monotonous trend of velocity for each spectral feature;
2) there are no fast velocity jumps;
3) the similarity in shape in a given spectrum (and in spectral fragments) between adjacent components suggested that we were dealing with the same feature;
4) a feature should have more or less continuous flux variations.
On the basis of the above criteria we distinguished more than 60 features and 35 of them were in an active phase for more than two years. The emission of more than 10 features was observed for a considerable time, between 4-8 years. However, these condensations were not in the active phase for all of this time. For some periods, of up to 2 years, there was no emission from them. We denoted with numbers the most interesting features.
All the features that we identified in the interval from -6 to 
are shown in Figs. 1 (click here)a-c. As may be seen from
Fig. 1 (click here) the dots are not chaotically distributed and it is possible
to fit most of them with more or less smooth curves. So, the curves of
Figs. 1 (click here)a-c, according to the above criteria, belong to separate
maser condensations. In some cases (particularly, when the maser emission
of the condensations was weak), the corresponding features appeared
episodically in the spectra for brief intervals. In such situations there
are only isolated dots or short curve segments. Some of these segments may
belong to the same spectral feature. As a result, for these features we
will have a disrupted curve.
Figure 1: Time variation in the velocity of 
maser
features (dots). The solid lines delineate those features which belong to
the same maser condensation. Dashed lines show the segment of the curves
where no emission was detected in a given time
Most of the dots are well fitted by smooth curves. This supports the idea
that there are preferred velocities at which the features
appear. We indicate with arrows and the letters "M" and "m" in
Figs. 1 (click here)a-c the times of the flux maxima and minima respectively.
Only for three features (12, 16 and 17) do the flux maxima coincide with
radial velocity maxima.
Time variations of the fluxes and differences between the radial velocities
of close spectral features are presented in Figs. 2 (click here)-6 (click here).
In each figure the flux of two-three features is shown. This allows
comparison of the evolution of those maser condensation which, most
probably, are located in the vicinity of the shell and whose
flux variations probably have a common cause.
As may be seen from Fig. 2 (click here) the fluxes of two spectral components, whose radial velocities are close, clearly anticorrelate. The anticorrelation was continuously observed during 1981-1986, having more or less periodic character with a period of about 1.5 years.
Figure 2: Flux variations of spectral features 8 and 9 which are
anticorrelated
Flux variation of the strong features at negative velocities during 1981-1986 is shown in Fig. 3 (click here). In this period a slow change of the maser to a minimum activity took place.
Figure 3: Flux variations of the two main spectral features at negative
velocities
Time evolution of the most long lived features is given in Fig. 4 (click here).
For convenience in the analysis of flux and velocity variations we
considered (on the basis of the arguments below) that features 16 and 17
are identical. They have similar radial velocities and one feature merges
into the other. After plotting smoothed curves (Fig. 4 (click here)b, dashed
line) the anticorrelation of the fluxes of the two main spectral features
(12 and 16/17) becomes clear. Moreover, there is correlation between the
flux variation of component 12 and the difference of radial velocities. The
maximum velocity separation between features in the spectra are
observed at times when one of the features (either of them) has a flux
maximum.
Figure 4: a) Flux variations of the most long lived spectral
features and b) difference of their radial velocities. Dashed lines
shown the fitted smoothed curves. Solid and dotted arrows indicate the
position of flux maxima
The flux variations of the most intense features in 1988-1996 are shown in
Fig. 5 (click here). In Fig. 6 (click here) time variations of the radial velocities
of features 10c and 11a relative to feature 17 are given. At first the
variation has an almost sinusoidal character with amplitude about 
. Later, the two components approach each other, the velocity
difference reducing linearly from 1.8 to 
.
Figure 5: Flux variations of the strongest components during
the transition of the maser to the active phase
Figure 6: Variations in the radial velocity of features 10c and 11
relative to feature 17
According to Fig. 1 (click here) many features show radial velocity drifts. To
analyze the character of the drift for the many features in ON2
we introduced the parameters 
and 
. They are the
sum of the drifts at a given time for all distinct features. The sum of the
drifts squares was also computed. Parameters and
characterize of the state of the system (maser condensations). The values
of the drifts in a given time were determined from the smoothed curves of
Fig. 1 (click here), using a sampling interval of 1/3 year. The results are
shown in Figs. 7 (click here)a-c, where also the averaged values are given. In
plotting Fig. 7 (click here) we did not take into account the features at
velocities 
, since they were rather weak and belong to
the S component.
Figure 7: a) Variation of the quadratic sum of the drifts of the
spectral features in radial velocity, b) the sum of the
drifts values and c) the mean value of the drifts. The solid arrows
indicate the positions of the main maxima of total flux and (dotted) of the
local maxima. Dashed lines are fitted smoothed curves. The position of the
local maximum in 1992-1993 was not determined due to the growth of the
main maximum
The curve giving the sum of the velocity drifts of the components
(Fig. 7 (click here)b) has a sharp jump, which divides it into two equal parts.
In each of them there is a clear tendency to drift changing from 
to 
. The drift jump took place during
1989, i.e., two years before the beginning of the maser activity
enhancement in ON2. The analysis of the component widths did not give
interesting results, so we do not show these data here.