Observational evidence of quasar feedback quenching star formation at high redshift
M. Cano-Diaz, R. Maiolino, A. Marconi, H. Netzer, O. Shemmer, G. Cresci
aa r X i v : . [ a s t r o - ph . C O ] D ec Astronomy&Astrophysicsmanuscript no. 2QZ00˙8 c (cid:13)
ESO 2018November 1, 2018 L etter to the E ditor Observational evidence of quasar feedback quenching starformation at high redshift ⋆ M. Cano-D´ıaz , R. Maiolino , , A. Marconi , H. Netzer , O. Shemmer , and G. Cresci INAF-Osservatorio Astronomico di Roma, via di Frascati 33, 00040 Monteporzio Catone, Italy Cavendish Laboratory, University of Cambridge, 19 J. J. Thomson Ave., Cambridge CB3 0HE, UK Dipartimento di Fisica e Astronomia, Universit`a degli Studi di Firenze, Largo E. Fermi 2, 50125 Firenze, Italy School of Physics and Astronomy and the Wise Observatory Tel-Aviv University, Tel-Aviv 69978, Israel Department of Physics, University of North Texas, Denton, TX 76203, USA INAF-Osservatorio Astrofisico di Arcetri, Largo E. Fermi 5, 50125 Firenze, ItalyReceived ; accepted
ABSTRACT
Most galaxy evolutionary models require quasar feedback to regulate star formation in their host galaxies. In particular, at highredshift, models expect that feedback associated with quasar-driven outflows is so e ffi cient that the gas in the host galaxy is largelyswept away or heated up, hence suppressing star formation in massive galaxies. We observationally investigate this phenomenonby using VLT-SINFONI integral field spectroscopy of the luminous quasar 2QZJ002830.4-281706 at z = λ α emission reveals star formation in the quasar host galaxy, with SFR ∼
100 M ⊙ yr − .However, the star formation is not distributed uniformly, but is strongly suppressed in the region with the highest outflow velocity andhighest velocity dispersion. This result indicates that star formation in this region is strongly quenched by the quasar outflow, whichis cleaning the galaxy disk of its molecular gas. This is one of the first direct observational proofs of quasar feedback quenching thestar formation at high redshift. Key words.
Galaxies: formation – Galaxies: high-redshift – Galaxies: evolution – quasars: emission lines
1. Introduction
Most of the recent galaxy formation models invoke energeticoutflows as a way to regulate the evolution of galaxies through-out the cosmic epochs (Silk & Rees, 1998; Bower et al., 2006;Springel et al., 2005). In particular, quasars are expected to drivepowerful outflows that eventually expel most of the gas in theirhost galaxies, thereby quenching both star formation and fur-ther black hole accretion (Granato et al., 2004; Di Matteo et al.,2005; Menci et al., 2005; Bower et al., 2006; Hopkins et al.,2008; King, 2010, e.g. ). According to those models, thesequasar driven outflows are required to prevent massive galaxiesfrom overgrowing, hence explaining the shortage of very mas-sive galaxies in the local universe, and are responsible for thered color and gas poor properties of local elliptical galaxies.Massive, large-scale outflows have been detected in the hostsof local quasars (e.g. Feruglio et al., 2010; Fischer et al., 2010;Sturm et al., 2011; Rupke & Veilleux, 2011). However, modelsexpect that most of the quasar feedback action occurs at high red-shift, when quasars reach their peak activity (z ∼
2) and when starformation in the most massive galaxies is observed to decline.Evidence of outflows in luminous quasars has been detectedup to very high redshift (e.g. Allen et al., 2011; Maiolino et al.,2004; Alexander et al., 2010). Indications that the strength ofthese AGN driven outflows anticorrelates with the starburst con-tribution to the infrared luminosity has been obtained in red-dened quasars by Farrah et al. (2011). However, direct observa- ⋆ Based on data obtained at the VLT through the ESO program077.B-0218(A). tional evidence that high-z quasar driven outflows quench starformation in their host galaxies is still missing.Here we present VLT-SINFONI near-IR integral field spectraof the quasar 2QZJ002830.4-281706 (hereafter 2QZ0028-28).This object, at z = . λ λ α emission is suppressed in the region characterized bythe strongest outflow. We suggest that this is the first observa-tional evidence of feedback associated with a quasar-driven out-flow that quenches star formation at high redshift, or one of thefirst.
2. Observations and data reduction
We used the near-IR integral field spectrometer SINFONI atthe VLT to observe 2QZ0028-28 in the H and K bands,where H β + [OIII] λλ α + [NII]6548,6584 are red-shifted, respectively, with the high-resolution gratings (deliv-ering R = = ′′ (measured from the broadlines, as discussed in the following). To properly sample this ex-cellent seeing, we used the camera that delivers a pixel scale of0 ′′ . × ′′ .
05, providing a field of view of 3 ′′ × ′′ . Fig. 2: Left panel: Velocity field (first moment map) of the [OIII] λ / s. Right panel: Velocity dispersion (secondmoment map) of the [OIII] λ / s.In both maps the black contours trace the continuum.Fig. 1: Upper panel: 2QZ0028-28 H band spectrum extractedfrom the central 0.5 arcsec, along with the various componentsused for the fit (see Appendix A for details). Vertical dashedlines indicate the rest frame wavelength of each line, by takingthe [OIII] λ ff ected by strong sky line residuals.Data was reduced with the ESO-SINFONI pipeline. Thepipeline subtracts the background, performs the flat-fielding,spectrally calibrates each individual slice, and then reconstructsthe cube. The pipeline delivers cubes where the spatial pixels areresampled to 0 ′′ . × ′′ .
3. Data analysis and results
The final H and K band data cubes were analyzed by fitting eachspatial spectrum separately in the field of view. Details of thefitting procedure for the spectra are given in Appendix A. β analysis: evidencefor a powerfuloutflow The H-band spectrum (Fig. 1) shows a clear broad H β , associ-ated with the broad line region (BLR), and a prominent [OIII]doublet, primarily associated with the quasar narrow line region(NLR). The [OIII] λ / N < / s,as a consequence of the strong blue asymmetry of the line. Thevelocity is strongly negative over the whole region where [OIII]is detected, suggesting that, if the prominent blue wing is dueto a wind, then the outflowing ionized gas is distributed over allof the central ∼ suggesting that the whole central region is characterized by out-flows. However, the velocity dispersion increases significantlyin the SE region, where the strongly blueshifted gas is detected.This correspondence between blueshift and velocity dispersionstrongly suggests that the blueshift observed in this region is dueto an outflow excess and not to galaxy rotation.Fig. 3: Top panel: 2QZ0028-28 K-band spectrum extracted fromthe central arcsec, along with the various components used forthe fit (blue are for H α components, magenta are for [NII] linesof component B). Vertical dotted lines show the rest-frame wave-length for H α and [NII] λ H α emission NW of the nucleus (Fig. 4) andthe spectrum extracted from the region without narrow H α emis-sion, SE of the nucleus, after scaling the two spectra to match theintensity of the broad line. A clear narrow H α component is de-tected, illustrating that the detection of this component and itsdistribution are not artifacts of the spectral fit.While there is a general correspondence between velocityblueshift and dispersion, both with the highest values in the SEquadrant, a detailed analysis reveals interesting di ff erences. Thepeak of the velocity dispersion is located between the two peaksof velocity blueshift. A possible interpretation is that those ar-eas with highest velocity dispersion are associated with regionswhere the outflow is strongly interacting with the gas in the hostgalaxy disk, which also lowers the velocity locally.The outflow may even be intrinsically symmetric, butthe opposite (receding) outflow is likely to be obscured bydust in the host galaxy disk (as observed in local AGNs,M¨uller-S´anchez et al., 2011). It is interesting to note that the ve-locity dispersion map shows a strong excess in a tiny region onthe NW (opposite to the main [OIII] outflow), this may be trac-ing the opposite outflow coming out of the galactic dusty disk.The [OIII] line is primarily tracing gas in the NLR ionizedby the QSO. Therefore, the most likely explanation is that theoutflow is driven by the QSO radiation pressure. A quasar ori-gin of the wind is also supported by the fact that the outflow-ing gas reaches velocities in excess of 1000 km / s (in the [OIII]blue wing), which cannot be explained by models of supernovae- driven outflows (Thacker et al., 2006). The mass of ionized out-flowing gas is simply given by (see Appendix B)M oution = . L ([OIII])n e3 [O / H] M ⊙ (1)where L ([OIII]) is the luminosity of the [OIII] λ erg / s, n e3 is the electron densityin the outflowing gas, in units of 10 cm − , typical of the NLR,and 10 [O / H] is the oxygen abundance is Solar units. If we assumethat the outflow is traced by the broad component of [OIII], thenL ([OIII]) ∼
2, implying that the mass of ionized gas involvedin the outflow is ∼ M ⊙ .By assuming a simplified conical (or biconical) outflow dis-tributed out to a radius R kpc (in units of kpc), then the outflowrate of ionized gas is given by (see Appendix B)˙M oution =
164 L ([OIII]) v n e3 [O / H] R kpc M ⊙ yr − (2)where v is the outflow velocity in units of 1000 km s − .The maximum outflow velocity inferred from the [OIII] profileis about 2000 km / s, which is probably representative of the av-erage outflow velocity, while the lower velocities observed inthe [OIII] line profile are very likely projection e ff ects. By us-ing R kpc =
3, we infer an ionized gas outflow rate of about200 M ⊙ yr − . This is a lower limit of the total outflow rate, sincethe ionized component is probably a minor fraction of the globaloutflowing gas mass. If we scale by the same neutral-to-ionizedfraction as in Mrk231 (Rupke, priv. comm.), the total outflowrate can be up to an order of magnitude higher.The inferred kinetic power of the ionized component of theoutflow is about 3 10 erg s − (Eq.B.9), or about ∼ − timesthe bolometric luminosity of the AGNs ( ∼ . erg s − , fromthe continuum luminosity at 5100Å, and by assuming a bolo-metric correction factor of 7). However, if the ionized outflow isaccompaneid by a neutral / molecular outflow an order of magni-tude more massive, as discussed above, then the kinetic poweris also likely to be an order of magnitude higher, i.e. about ∼ α +[NII] analysis: evidencefor quenchedstar formation The H α has a very broad profile (from the BLR), and its over-all profile has been fitted mostly with two broad Gaussians (seeAppendix A for details). Interestingly, the fit also requires thepresence of a weak, but significant narrow component of H α (A*, FWHM ∼ km / s ). The absence of [NII] indicates thatthis narrow H α emission is mostly due to star formation.The map of the narrow H α emission is shown in Fig.4, whichreveals star formation extending over a few kpc from the nu-cleus. However, the star formation is not distributed symmetri-cally around the nucleus, but primarily towards the N and W. Inparticular, the region within a few kpc SE of the nucleus is nearlyfree of any narrow H α tracing star formation. To show that thedetection of narrow H α and that the asymmetric distribution arenot artifacts of the fitting procedure, we extracted a spectrum byintegrating over the NW region putatively containing star for-mation, according to the H α map, and another spectrum in theSE region devoid of star formation. After scaling the two spec-tra so that the intensity of the broad H α wings (associated withthe BLR) are the same, we have obtained the di ff erence of thetwo spectra, which is shown in the bottom panel of Fig. 3. The Fig. 4: Map of the narrow component of H α with contours tracing the [OIII] velocity shift (left panel) and velocity dispersion (rightpanel), as in Fig. 2. Star formation, traced by H α , is heavily suppressed in the SE region where the strongest outflow is traced by[OIII].di ff erential spectrum clearly shows the presence of the narrowH α . The line broadened profile (also making the detection nois-ier than expected ) is because we are integrating over di ff erentregions of the velocity field. The di ff erential spectrum also con-firms that no [NII] is detected in association with the narrowH α , confirming that the latter is tracing star formation and notthe quasar NLR.We mention that the map of the [OIII] narrow component(shown in the appendix) is also characterized by a similar asym-metry towards the NE, suggesting that some of the [OIII] narrowline is associated with star formation. However, the asymmetryis less clean than observed for the H α narrow component, likelybecause of NLR contribution to [OIII] and because of the blend-ing with the “broad” [OIII] component.The integrated emission of the narrow H α yields a total starformation rate in the host galaxy of about 100 M ⊙ yr − (by usingthe conversion factor given in Kennicutt, 1998), which is notunusual in high-z quasars (e.g. Lutz et al., 2008). However, themost interesting result is that the star formation is heavily sup-pressed in the SE region, which is characterized by the excess ofoutflow with high-velocity dispersion. In Fig. 4(left) the whitecontours identify the strongest gas outflow traced by the highlyblueshifted [OIII] line, as in Fig. 2-left, while in Fig. 4-right thewhite contours identify the highest velocity dispersion region,as in Fig. 2-right, which is likely the region where the strongoutflow interacts with the host galaxy disk. We suggest that theheavy suppression of star formation in the region of strongestquasar-driven outflow among the first direct observational proofsof quasar feedback onto the host galaxy quenching star forma-tion at high redshift, as predicted by models.
4. Conclusions
By using near-IR integral field spectroscopic observations wehave revealed a powerful outflow in the host galaxy of the quasar2QZ0028-28 at z = λ ⊙ yr − , which is, however, a lower limit of the total gas outflow rate. Both the high outflow velocity ( > / s) andthe fact that the wind is mostly traced by the [OIII] line (pro-duced primarily in the NLR) strongly suggest that the outflow ismostly driven by the quasar. The outflow is not symmetric, thehighest velocities and highest velocity dispersion are found inthe region SE of the nucleus.In the K-band, our data clearly reveal the presence of narrowH α emission tracing star formation in the host galaxy, on scalesof several kpc and with a rate of about 100 M ⊙ yr − . However,star formation is not distributed uniformly in the host galaxy,but is mostly found in the regions not directly invested by thestrong outflow. Instead, star formation is heavily suppressed inthe SE region where the strongest outflow is detected. This ob-servational result supports models invoking quasar feedback toquench star formation in massive galaxies at high redshift. Acknowledgements.
We are grateful to the referee for his / her very useful com-ments. MCD is supported by the Marie Curie Initial Training Network ELIXIRunder the contract PITN-GA-2008-214227 from the European Commission. References
Alexander, D. M., Swinbank, A. M., Smail, I., McDermid, R., & Nesvadba,N. P. H. 2010, MNRAS, 402, 2211Allen, J. T., Hewett, P. C., Maddox, N., Richards, G. T., & Belokurov, V. 2011,MNRAS, 410, 860Bower, R. G., Benson, A. J., Malbon, R., Helly, J. C., Frenk, C. S., Baugh, C. M.,Cole, S., & Lacey, C. G. 2006, MNRAS, 370, 645Carilli, C. L., & Wang, R. 2006, AJ, 131, 2763Di Matteo, T., Springel, V., & Hernquist, L. 2005, Nature, 433, 604Farrah, D., Urrutia, T., Lacy, M., et al. 2011, arXiv:1112.1092Feruglio, C., Maiolino, R., Piconcelli, E., Menci, N., Aussel, H., Lamastra, A.,& Fiore, F. 2010, A&A, 518, L155Fischer, J., Sturm, E., Gonz´alez-Alfonso, E., et al. 2010, A&A, 518, L41Granato, G. L., Silva, L., de Zotti, G., Bressan, A., & Danese, L. 2004,Astrophysics and Space Science Library, 301, 129Hopkins, P. F., Cox, T. J., Kereˇs, D., & Hernquist, L. 2008, ApJS, 175, 390Kennicutt, R. C., Jr. 1998, ARA&A, 36, 189Lutz, D., et al. 2008, ApJ, 684, 853King, A. R. 2010, MNRAS, 402, 1516Lapi, A., Cavaliere, A., & Menci, N. 2005, ApJ, 619, 60Maiolino, R., Oliva, E., Ghinassi, F., Pedani, M., Mannucci, F., Mujica, R., &Juarez, Y. 2004, A&A, 420, 889Menci, N., Fontana, A., Giallongo, E., & Salimbeni, S. 2005, ApJ, 632, 49
M¨uller-S´anchez, F., Prieto, M. A., Hicks, E. K. S., et al. 2011, ApJ, 739, 69Nagao, T., Marconi, A., & Maiolino, R. 2006, A&A, 447, 157Netzer, H., Shemmer, O., Netzer, H., Maiolino, R., Oliva, E., Croom, S., Corbett,E., & di Fabrizio, L. 2004, ApJ, 614, 558Rupke, D. S. N., & Veilleux, S. 2011, ApJ, 729, L27Shemmer, O., Netzer, H., Maiolino, R., Oliva, E., Croom, S., Corbett, E., & diFabrizio, L. 2004, ApJ, 614, 547Silk, J., & Rees, M. J. 1998, A&A, 331, L1Sturm, E., et al. 2011, ApJ, 733, L16Springel, V., Di Matteo, T., & Hernquist, L. 2005, MNRAS, 361, 776Thacker, R. J., Scannapieco, E., & Couchman, H. M. P. 2006, ApJ, 653, 86
Appendix A: Details of the spectral fitting
The initial spectral fit is performed on a spectrum extracted froma central aperture of ten pixels (i.e. 0.5 arcsec) in the H band and20 pixels (i.e. 1.0 arcsec) in the K band, which guarantee highS / N.The emission lines are fitted with multiple Gaussians and,in the case of the broad lines, by also using power-law profiles,as in Nagao et al. (2006). The continuum is fitted with a singlepower law, which represents the emission of the QSO accretiondisk (plus possibly some minor contribution from the host galaxystellar continuum). Starting from this initial fit in the central re-gion, we fitted the spectra individually at all spatial pixels byleaving most of the parameters free, except for the velocity, dis-persion, and relative intensity of the components describing thebroad lines (H α and H β ). Indeed, since the BLR is unresolved,the shape and shift of the broad line profile (components C + D + Ein the case of H β ) must be constant over the field of view, andthere the global intensity variation only reflects the seeing PSF.We tried to also include an Fe II template in the spectral fitting(as in Netzer et al., 2004); however, this is always set to zeroby the fitting procedure, confirming the lack of significant FeIIemission inferred by the visual inspection of the spectrum. Thisresult contrasts with the measurement of Netzer et al. (2004),who obtain a flux of the Fe II emission that is about 0.37 timesthe H β emission, and we ascribe the discrepancy to the muchlower signal-to-noise in the latter spectrum.Figure 1 shows the various components used to fit the H-band spectrum of the central region and, in the bottom panel,the fit residuals. The region included within the two green linesis a ff ected by strong sky residuals and was not considered inthe fit. In this figure we can see that the [OIII] profile canbe nicely fitted with two Gaussians, one relatively narrow (A,FWHM ∼
600 km / s) and a second one (B) blueshifted by about700 km / s and with FWHM ∼ / s (much broader thantypically found in the NLR of lower luminosity AGNs). The[OIII]4959 line is fitted with the same components, linked tohave an intensity equal to one third of the [OIII] λ α narrowcomponent, suggesting that the narrow component of [OIII] re-ceives a significant contribution from star formation. However,the imperfect correspondence between the two maps suggeststhat a fraction of the narrow [OIII] is also contributed by theNLR.The [OIII]4959 line is heavily blended with another line at λ rest = β + [OIII] group is particularly weak, therefore suggesting adi ff erent origin of this line. In the residuals we observe a narrowcomponent that is likely associated to the same unidentified line.The fit of the broad H β requires three Gaussians and a powerlawprofile (as in Nagao et al. 2006). We also included the H β contri-bution associated with the two [OIII] components. However, oneshould bear in mind that the intensity of these weaker H β com-ponents is di ffi cult to evaluate, since these are blended withinthe broad, complex H β profile. In Table A.1 we list the parame-ters inferred for all of the components used in the fitting of theSINFONI spectra.The H α profile (dominated by the broad component, trac-ing the BLR) is clearly di ff erent from the H β profile, which is aproperty that is common to many other quasars and AGNs andthat is ascribed to complex radiative transfer within the densegas of the BLR. The bulk of the H α profile was fitted by usingtwo very broad Gaussians (H and G in Fig. 3), which give theseeing in the K-band, which is roughly consistent with the see-ing measured in the H-band observation. As for the broad H β ,the relative intensity, shift, and width of these two lines are keptfixed over the field of view, and their overall intensity variationreflects the seeing PSF.We forced the inclusion of component B in the H α profile,by using the same velocity shift and width as the correspond-ing [OIII] component, while the intensity was left free to vary.This component of H α is certainly associated with the NLR.However, since this component is relatively broad, its intensityis poorly constrained (see uncertainty in the flux of this com-ponent in Table A.1), and degenerate with the other two broadH α Gaussian components (as is the case for the correspondingcomponent in H β ). The [NII] doublet associated with compo-nent B is also forced to have the same shift and width as thecorresponding [OIII] component. The relative intensity of thetwo [NII]6548,6584 lines are forced to be in the ratio of 1:3.The intensity of these [NII] lines is even less constrained thanthe corresponding H α B component, since the wavelength of[NII]6548 nearly overlaps the intense component G of H α . Theresulting [NII] / H α ratio of component B is low, but highly uncer-tain (log ( F [ NII ] / F H α ) = − . ± . α profile. However, in principle, we should include two nar-row (FWHM ∼ / s) H α components, one associated withthe quasar NLR and another one associated with any putativestar formation in the host galaxy. However, the quality of ourdata does not really allow us to fit two separate narrow compo-nents, because these would be totally degenerate. We thereforefit a single narrow component not tied to have the same velocityand width of component “A” of [OIII]. We label this narrow H α component with “A*” (meaning that it may be partly associatedwith component A of [OIII], but not necessarily). We investigatethe relation of the H α component “A*” with the [OIII] compo-nent “A” a posteriori . We note that it is not possible to followa similar approach on H β (i.e. introduce a component “A*”, notlinked to the [OIII] components, since, besides the problem ofthe blending with other components, the signal-to-noise on H β is much lower).The parameters resulting from the best fit in the K-band aregiven in Table A.1. We note that there is no room for narrow A*[NII] emission, at a level of F [NII]6584 (A ∗ ) < .
21 F H α (A ∗ ). Thelack of an [NII] narrow component is confirmed by the “di ff er- Fig. A.1: Flux maps of the broad (B, left) and narrow (A, right) components of [OIII].ential” spectrum presented in Sect. 3.2 and in Fig. 3, which istotally independent of any fitting procedure. As discussed in thebody of the paper, the lack of [NII] at a level below one fourthof H α indicates that this narrow H α emission is mostly tracingstar formation, and not the NLR.Figure A.2 shows the velocity field and the velocity disper-sion of the narrow component of H α . Both maps are very noisy,owing to the weakness of the line. The velocity field does notclearly indicate a rotation pattern, which would be expected bygas in a galactic disk, except possibly for an NW-SE gradient,but the latter may be associated with some contribution to H α narrow from the outflow in the SE region. However, only a frac-tion of the disk is actually traced by the H α narrow, and this,together with the noisy velocity map, may prevent identificationof a clear rotation pattern. Moreover, it is well known that hostgalaxy disks of optically selected quasars tend to be face on, asa consequence of selection e ff ects (Carilli & Wang, 2006), so itis not expected that quasar host galaxies have prominent rota-tional patterns. The velocity dispersion map is very noisy, but itis consistent with being uniform over the area where H α narrowis detected.We finally note that the best-fit velocity and FWHM of com-ponent A* of H α are similar to component A of [OIII]. Thisfurther suggests that the latter component of [OIII] is partly con-tributed by the ionized gas in the star-forming regions traced bythe narrow H α . The inferred F A ∗ H α / F A[OIII] ∼ .
17, at the verge ofthe range typically observed in star-forming galaxies, indicatesthat the flux of component A of [OIII] is not incompatible withbeing partly originated by star formation, but probably a contri-bution by the AGN NLR is required. As mentioned above, thesimilarity of the F [OIII] (A) map and the F H α (A ∗ ) map also sup-ports the scenario where part of component A of [OIII] is asso-ciated with star formation. Appendix B: A simple model of the ionized outflow
In this section we discuss how the physical properties of the ion-ized outflow can be constrained through the observational pa-rameters of the [OIII] line, by adopting a simple model for the ionized wind. The [OIII]5007 line luminosity associated with theoutflow is simply given byL([OIII]) = Z V ǫ [OIII] f dV (B.1)where V is the volume occupied by the outflowing ionized gas, fthe filling factor of the [OIII] emitting clouds in the outflow, and ǫ [OIII] the [OIII]5007 emissivity that, at the temperature typical ofthe NLR ( ∼ K), has a weak dependence on the temperature( ∝ T . ). It can be expressed by ǫ [OIII] = .
11 10 − h ν [OIII] n O + n e erg s − cm − , (B.2)where h ν [ OIII ] is the energy of the [OIII]5007 photons (in units oferg), n O + and n e are the volume densities of the O + ions and ofelectrons, respectively (in units of cm − ). Under the reasonableassumption that most of the oxygen in the ionized outflow is inthe O + form, then ǫ [OIII] ≈ − h ν [OIII] n [O / H] erg s − cm − (B.3)where 10 [O / H] gives the oxygen abundance in solar units.The mass of outflowing ionized gas is given byM ionout = Z V .
27 m H n e f dV (B.4)where m H is the mass of the hydrogen atom, and where we haveneglected the mass contributed by species heavier than helium.By combining Eqs. B.1 and B.4 we obtainM ionout = .
33 10 C L ([OIII]) h n e3 i [O / H] M ⊙ (B.5)where L ([OIII]) is the luminosity of the [OIII]5007 lineemitted by the outflow, in units of 10 erg s − , h n e3 i ( = R V n e f dV / R V f dV ) is the average electron density in the ion-ized gas clouds, in units of 10 cm − , and C = h n e3 i / h n i is a“condensation factor”, where h n i = R n f dV / R f dV. We canassume C = Fig. A.2: Velocity field (left) and velocity dispersion (right) of the narrow component (A*) of H α . Line Component Fitting Function λ obs FWHM Flux Velocity ( µ m) (km / s) (10 − erg cm − s − ) (km / s)[OIII]5007 A Gaussian 1.7061 652 10.9 ± ± β A Gaussian 1.6564 652 0.79 ± ± ± ± ± ± α A* Gaussian 2.2361 616 1.9 ± ± ± ± < B Gaussian 2.2382 1797 1.7 ± Table A.1: Results of the spectral fit for the individual components.The names of the components correspond to those shown in Figs. 1 and 3. Notes: The velocities of the components are obtainedby assuming the peak of [OIII] λ For this [NII] line, the width and velocity were forced, to estimatethe upper limit to the values found for the corresponding A* component in H α .If we assume a simplified model of the outflow (justified bythe limited information currently available to us) where the windoccurs in a conical region, with opening angle Ω , composed ofionized clouds uniformly distributed and outflowing with veloc-ity v, out to a radius R, then the mass outflow rate of ionized gasis given by˙M ionout = h ρ i V v Ω R (B.6)where h ρ i V is the average mass density in the whole volume oc-cupied by the outflow, which is given by h ρ i V = M ionout V (B.7)where the volume occupied by the conical outflow is given byV = π R Ω π . Unless f =
1, generally h ρ i V , .
27 m H h n e i , sincethe latter numerical density (defined above) is averaged amongthe emitting clouds, not over the whole volume. By replacing Eqs. B.5 and B.7 into Eq. B.6 we obtain thatthe ionized outflow rate is given by˙M ionout = C L ([OIII]) v h n e3 i [O / H] R kpc M ⊙ yr − (B.8)where L ([OIII]), n e3 , and C ( ≈
1) were defined above, v isthe outflow velocity in units of 1000 km s − , and R kpc is theradius of the outflowing region, in units of kpc. The outflow rateis independent of both the opening angle Ω of the outflow and ofthe filling factor f of the emitting clouds (under the assumptionof clouds with the same density).The kinetic power (associated with the ionized component)is then given byP ionK = .
17 10 C L ([OIII]) v h n e3 i [O / H] R kpc erg s − . (B.9)(B.9)