2 Observations and data presentation

In this work, we present spectra of maser emission of the source G43.8-0.1 in the 1.35 cm H₂O line from May 1994 to November 1998. We give these spectra because during this time interval a strong flare of the feature at 38.8 km s^-1 took place, and the overall increase of the maser activity was observed.

The observations were carried out on the RT-22 radio telescope of the Pushchino Radio Astronomy Observatory, Astrospace Centre of the Lebedev Institute of Physics, Russian Academy of Sciences. A cooled FET amplifier of the 22 GHz frequency band was used. The beamwidth of the antenna at 22 GHz is 2.6$^\prime$. The system noise temperature was 200-300 K. The antenna temperature of 1 K for an unpolarised source corresponds to a flux density of 25 Jy.

From the beginning of 1997, time intervals between consecutive observing sessions mainly did not exceed one month. This allowed us to trace the evolution of the three main parameters of spectral components during their flares: flux density, radial velocity and linewidth. In the observations, we used a 96-channel spectrum analyser with a radial-velocity resolution of 0.101 km s^-1. Starting from August 1997, the spectrum analyser includes 128 channels. To investigate the linewidth of the 42.2 km s^-1 feature, we used in some cases a filter bank with a frequency resolution of 2.5 kHz (34 m s^-1 in the H₂O line).

The H₂O spectra of the source G43.8-0.1 for 1994-1998 are presented in Figs. 1a-g. Figure 2 shows variability of the integral flux of the emission. A sharp growth of the flux from the beginning of 1997 is caused by a series of consecutive flares of several features in the velocity interval of 37.5-39.5 km s^-1.

In Fig. 3, flux density variations of the spectral features that dominated in 1994-1998 is shown. The strongest feature of this period appeared at the end of 1996 at 38.2 km s^-1. At first, it was, however, strongly blended with other features; therefore, we could reliably determine its velocity only after the beginning of 1997, and the linewidth -- from March 1997. The peak of the emission (F=3750 Jy) was observed in September 1997, and then the activity rapidly faded. At the end of December 1997, the flux fell to 1200 Jy. At that time other components began to appear in the velocity interval of 37.8-39 km s^-1, i.e. a cascade of less intense flares was observed.

Variations of the radial velocity and linewidth of individual features of the H₂O spectrum are shown in Fig. 4. For the component at $V_{\rm LSR}=42.2$ km s^-1, a smoothed curve is fitted (dashed line); it shows a small regular radial-velocity drift between early 1994 and mid-1997. Then the drift direction changed to the opposite, with a greater velocity gradient. For other features, the velocity drift was more significant.

Figure 2: Time dependence of integral flux of G43.8-0.1 in the H2O line

Figure 3: Time dependence of the flux for the main components of the H2O spectra. Vertical arrow denotes the minimum of the integral flux

Figure 4: Variations of radial velocity and linewidth of main H2O components of G43.8-0.1. The dashed line shows the regular radial-velocity drift of the component at 42.2 km s-1

Figure 5: Linewidth versus flux density for the component at 38.2 km s-1. The cross below shows the maximum errors of the measurements in both coordinates

Important linewidth variations were observed in the feature at $V_{\rm LSR}=38.2$ km s^-1, which flared in 1997. The degree of saturation of the maser can be evaluated from the flux-linewidth dependence, or, more conveniently, $\ln F-\Delta V^{-2}$ dependence, which expresses the logarithm of the Gaussian function. For this purpose, Fig. 5 shows the flux variability versus linewidth for the 42.2 kms^-1 feature. In the case of unsaturated amplification this dependence in the $\ln F, \Delta V^{-2}$ coordinates is presented by a straight line (Elitzur 1992):

$$\Delta V \propto [\ln F(V_0)]^{-1/2}.$$

In the limits of the observational errors (the cross in Fig. 5) all the points fit a straight line. Only one point falls out; it corresponds in time to the onset of considerable changes of the spectrum structure. At this epoch, the linewidth is overestimated, because it was not possible to separate the spectrum into components. The emission of the feature at 38.2 km s^-1 is narrowband: at the flux maximum, the linewidth at half-power was 0.43 km s^-1. Our spectral resolution (0.101 km s^-1) was insufficient to record this line without distortions; therefore, we corrected the line for the filter passband.

Figure 6: Linewidth versus flux density for the component at 42.2 km s-1 in 1994-1997

Figure 7: Time dependence of logarithm of flux density ($\ln F$) for the component at 42.2 km s-1 in 1981-1997

Figure 8: Linewidth ( $\Delta V^{-2}$) versus logarithm of flux density ($\ln F$) for the component at 42.2 km s-1 in 1981-1997. Arrows show the evolution track. Numbers denote the epochs ofobservations

On the other hand, the low-velocity wing of the line was blended by the emission of another line. However, when the main-line flux was above 500 Jy, blending was insignificant and in no way could affect the parameters of the gaussian curve fitted. Nevertheless, we corrected the left wing of the line. Only after that we fitted a gaussian and determined its parameters, including the linewidth, whose value is most sensitive to influence of contaminating features.

Throughout the entire timespan of our observations of G43.8-0.1, for the component at 42.2 km s^-1 there was a functional relation between the flux density and linewidth. At the final stage of evolution of this most "long-living'' component, this functional dependence is shown in Fig. 6. Only one point falls out. All the other points are divided into two groups, which contact near point $\Delta V^{-2}\approx4.25$, $\ln F\approx6.25$. Within each group, the points are mainly arranged along straight lines: $\ln F\propto\Delta V^{-2}$ for the first group and $\ln F\propto-\Delta V^{-2}$ for the second group. Note also that the second group of the points corresponds to the time interval from early 1994 to the end of 1996, while the first one refers to entire 1997.