A Redshift Survey of the Submillimeter Galaxy Population

a r X i v :a s t r o -p h /0412573v 3 23 D e c 2004

Draft version February 2,2008

Preprint typeset using L A T E X style emulateapj v.04/03/99

A REDSHIFT SURVEY OF THE SUBMILLIMETER GALAXY POPULATION

S.C.Chapman,1A.W.Blain,1Ian Smail,2R.J.Ivison 3,4

Submitted to Astrophysical Journal

ABSTRACT

We have obtained spectroscopic redshifts using the Keck-I telescope for a sample of 73submillimeter (submm)galaxies,with a median 850μm ?ux density of 5.7mJy,for which precise positions are available through their faint radio emission.The galaxies lie at redshifts out to z =3.6,with a median redshift of 2.2and an interquartile range z =1.7–2.8.Modeling a purely submm ?ux-limited sample,based on the expected selection function for our radio-identi?ed sample,suggests a median redshift of 2.3with a redshift distribution remarkably similar to the optically-and radio-selected Quasars.The observed redshift distributions are similar for the AGN and starburst sub-samples.The median R AB =24.6for the sample.However,the dust-corrected ultraviolet (UV)luminosities of the galaxies rarely hint at their huge bolometric luminosities indicated by their radio/submm emission,underestimating the true luminosity by a median factor of ～100for SMGs with pure starburst spectra.Radio and submm observations are thus essential to select the most luminous,high-redshift galaxies.The 850μm,radio,and redshift data is used to estimate the dust temperatures,and characterize photometric https://www.doczj.com/doc/8310426443.html,ing 450μm measurements for a subset of our sample we con?rm that the median dust temperature of T d =36±7K,derived assuming the local FIR–radio correlation applies at high redshift,is reasonable.Individual 450μm detections are consistent with the local radio-FarIR relation holding at z ～2.This median T d is lower than that estimated for similarly luminous IRAS 60μm galaxies locally.We demonstrate that dust temperature variations make it impossible to estimate redshifts for individual submm galaxies using simple long-wavelength photometric methods to better than ?z ?1.We calculate total infrared and bolometric

luminosities (the median infrared luminosity estimated from the radio is 8.5+7.4?4.6×1012

L ⊙),construct a luminosity function,and quantify the strong evolution of the submm population across z =0.5–3.5,relative to local IRAS galaxies.We use the bolometric luminosities and UV-spectral classi?cations to determine a lower limit to the active galactic nucleus (AGN)content of the population,and measure directly the varying contribution of highly-obscured,luminous galaxies to the luminosity density history of the Universe for the ?rst time.We conclude that bright submm galaxies contribute a comparable star formation density to Lyman-break galaxies at z =2?3and including galaxies below our submm ?ux limit this population may be the dominant site of massive star formation at this epoch.The rapid evolution of submm galaxies and QSO populations contrasts with that seen in bolometrically lower luminosity galaxy samples selected in the restframe UV,and suggests a close link between submm galaxies and the formation and evolution of the galactic halos which host QSOs.

Subject headings:cosmology:observations —galaxies:evolution —galaxies:formation —galaxies:

starburst

1.INTRODUCTION

The submillimeter (submm)galaxy population was ?rst detected seven years ago with the Submillimetre Common User Bolometer Array (SCUBA –Holland et al.1999)on the JCMT (Smail,Ivison &Blain 1997;Hughes et al.1998;Barger et al.1998;Eales et al.1999).Their discovery mo-tivated a variety of surveys using both SCUBA and a sim-ilar instrument,MAMBO (Bertoldi et al.2000;Kreysa et al.2002),on the IRAM 30-m telescope.Several surveys were undertaken of blank ?elds,using di?erent strategies to determine their depths and area coverage (Barger et al.1999a,2002;Eales et al.2000;Scott et al.2002;Borys et al.2003;Serjeant et al.2003;Webb et al.2003a;Danner-bauer et al.2004;Greve et al.2004).These surveys detect sources at a rate of approximately one source per night.A second class of survey utilized massive clusters to provide

a mild gravitational lensing boost to aid in the detection and study of SMGs.These surveys uncovered sources at a rate of about two per night,and examples of such sur-veys include Smail et al.(2002),Chapman et al.(2002a),Cowie,Barger &Knei

b (2002),and Knudsen (2004).

These surveys have gradually built up a large enough sample of Sub-Millimeter Galaxies (SMGs)to produce a statistically useful count (e.g.,Blain et al.2002).How-ever,until very recently most of our detailed knowledge of the properties of SMGs came from a handful of SMGs identi?ed in lensing cluster ?elds (e.g.,Ivison et al.1998,2000,2001;Frayer et al.1998,1999,2004),as follow-up of blank ?eld sources without the lensing boost required signi?cantly more resources (e.g.,Gear et al.2000;Lutz et al.2001;Dannerbauer et al.2002;Dunlop et al.2004).A breakthrough in our understanding of the proper-

1California Institute of Technology,Pasadena,CA,91125,U.S.A.

2Institute

for Computational Cosmology,University of Durham,South Rd,Durham DH13LE,UK

3Astronomy Technology Centre,Royal Observatory,Blackford Hill,Edinburgh EH93HJ,UK 4Institute for Astronomy,University of Edinburgh,Blackford Hill,Edinburgh EH93HJ,UK

1

2

ties of SMGs came from exploiting ultradeep20-cm radio maps.A strong correlation exists between the far-infrared (FIR)and radio?ux densities of galaxies,both locally and at high redshifts(e.g.,Helou et al.1985;Condon1992; Garrett2002),and so deep radio imagery with the Very Large Array(VLA)5can be used to help pinpoint and study SMGs(Ivison et al.1998;Smail et al.2000;Barger, Cowie&Richards2000;Chapman et al.2001,2002b). These radio maps provide a～1.5′′beam and～0.5′′as-trometric precision relative to the optical frame,su?cient to accurately locate the counterpart of a submm source. The use of radio imaging has culminated in the success-ful identi?cation of at least65%of SMGs brighter than S850μm>5mJy and their photometric characterization in the optical/near-infrared waveband(Ivison et al.2002; Chapman et al.2003a;Wang et al.2004;Borys et al.2004; Greve et al.2004).

The ability to precisely locate the position of a submm emitting source is essential if we wish to study their prop-erties in any detail.In particular,this is a necessary?rst step when trying to derive redshifts and luminosities for these systems.It had been hoped that long-wavelength ob-servations of the dust emission spectrum of these galaxies might be prove a reliable route to derive their redshifts and luminosities.The submm/radio?ux ratio was?rst used by Carilli&Yun(1999)in this manner to estimate the typical redshift of SMGs,however the technique was rec-ognized immediately to have limited accuracy(～50%red-shift errors)if there was a range of dust temperatures(T d) present.This uncertainty is particularly important when deriving luminosities and related properties from submm

observations as the submm?ux density is S850μm∝T?3.5

d for a?xed FIR luminosity at z?2.In addition,ther

e is a strong degeneracy between T d and(1+z)(Blain1999), which limits the usefulness o

f simple photometric redshifts for estimatin

g luminosities for the SMG population.Re-?nement of the modeling and?tting techniques appears not to have overcome this basic source of uncertainty(e.g., Aretxaga et al.2003;Wiklind2003).Indeed,even surveys at several submm wavelengths(e.g.,Hughes et al.2002) cannot completely overcome the degeneracy between dust temperature and redshift(Blain1999;Blain,Barnard& Chapman2003–see also Aretxaga et al.2004for a con-trary view).

As a consequence,precise redshifts are crucial for in-terpreting almost every aspect of SMGs.Prior to2002, only a handful of spectroscopic redshifts were available for unambiguously-identi?ed SMGs(Ivison et al.1998,2000; Barger et al.1999b;Lilly et al.1999).Recent attempts to measure redshifts for SMGs have met with more suc-cess(Chapman et al.2002c,2003a;Barger et al.2002; Ledlow et al.2002;Smail et al.2003a,2003b;Kneib et al.2004).However,the resulting sample is still restricted in size and unrepresentative of the general properties of the SMG population(with a bias towards optically-bright counterparts and a preponderance of strong-lined AGN).A redshift survey of a large and representative sample of submm galaxies is therefore urgently required. Chapman et al.(2003b–hereafter C03)demonstrated that spectroscopic redshifts can be obtained for even the optically faintest SMGs,spanning a factor of100×in I-band?ux,allowing a much more representative sample of the population to be studied.Their approach involved constructing densely-packed distributions of submm galax-ies across～10′?elds(matched to the area coverage of multi-object spectrographs on10-m telescopes)with pre-cise positions from radio counterparts.These samples could then be e?ciently and e?ectively targeted using deep spectroscopy in the UV/blue spectral region.With a large, unbiased sample of SMGs constructed in this manner it is possible to address questions about the SMG population with more certainty:including their dust temperatures (T d)and SED properties,their luminosities at various wavelengths and luminosity evolution,their contribution to the FIR background,and their relation to other popu-lations of galaxies and AGN at high redshift.

In this paper,we present an expanded sample from the ten SMGs with robust spectroscopic redshifts described by C03,to provide a total sample of73redshifts for unambiguously-identi?ed SMG’s.We discuss the prop-erties and observations of this sample,along with selec-tion e?ects in§2.We present the basic observational re-sults obtained for this sample,including the redshift distri-butions,variation in SEDs with redshift as characterized by the submm/radio?ux ratio,and optical properties in §3.In§4we then use basic assumptions to derive dust temperatures and bolometric luminosities for our sample, compare the UV properties of the galaxies with their ra-dio/submm emission,assess their contribution to the lu-minosity and star formation histories of the Universe and the FIR background(FIRB)and discuss their evolution-ary connections with other high-redshift populations.Fi-nally,in§5we give our main conclusions.All calcula-tions assume a?at,ΛCDM cosmology with?Λ=0.7and H0=71km s?1Mpc?1.

2.SAMPLE DEFINITION,OBSERVATIONS AND ANALYSIS The parent sample of SMG’s used for our survey con-sists of150sources detected at850μm(>3σ)with SCUBA/JCMT,lying in seven separate?elds(?eld centers listed in Table1):CFRS03,Lockman Hole,HDF,SSA13, Westphal-14,ELAIS-N2,and SSA22.104of these SMGs have radio identi?cations from deep VLA radio maps at 1.4GHz.This radio-identi?ed subset are the focus of this paper.

In all?elds the SCUBA submm data was retrieved from the JCMT archive6and reduced in a consistent man-ner using the SURF reduction tools(Jenness et al.1998) and our own software to extract beam-weighted submm ?uxes.7In some cases,additional radio sources were tar-geted in SCUBA’s photometry mode(Holland et al.1999) to e?ciently construct large samples of SMGs to target

5The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities,Inc.

6The JCMT is operated by the Joint Astronomy Centre on behalf of the United Kingdom Particle Physics and Astronomy Research Council, the Netherlands Organisation for Scienti?c Research,and the National Research Council of Canada.The JCMT archive is provided through the Canadian Data Archive Center.

7Map?uxes are obtained by extracting the e?ective beam imprint,(?0.5,1,?0.5)×δSCUBA chopping/nodding pro?le(e.g.,Scott et al. 2002;Borys et al.2003).

Chapman et al.3

in contiguous regions around?elds mapped by SCUBA (e.g.,Chapman et al.2001a,2002b,2003a).In addition, follow-up SCUBA photometry was used to verify the real-ity and submm?ux densities of11of the sources detected in SCUBA maps.These new SCUBA observations were obtained during JCMT observing runs in2002and2003, with sky opacity at225GHz,τ225=0.04–0.09.The ob-serving strategy was to integrate for a?xed length of time (1.0hrs)on all targeted galaxies,with additional time allo-cated to targets which did not achieve our nominal RMS sensitivity goal of dS850=1.5mJy due to weather condi-tions.We note the observational mode used to identify sources in Table3,based on whether their submm de-tection was obtained entirely in photometry mode(′P′), entirely in mapping mode(′M′)or through a combination of the two modes(′MP′).

Radio data for these?elds either existed from previous work by members of our group(Lockman Hole,ELAIS-N2),was rereduced for the purpose of this study(CFRS-03,SSA22,Westphal-14),or obtained from the public release(HDF,Richards2000).The SSA13radio data were obtained from E.A.Richards(private communica-tion),and is so far unpublished(a subsequent reduction of the SSA13data is described in Fomalont et al.2004). The radio data for the Lockman Hole and ELAIS-N2?elds come from Ivison et al.(2002)who identi?ed counterparts to the submm galaxies in these regions from Scott et al. (2002).For those?elds which we rereduced the radio data was retrieved from the VLA archive when available and combined with new data obtained by our group in SSA22(36hrs,A-array),Westphal-14(24hrs,B-array), and CFRS-03(16hrs,B-array).The radio maps were re-duced in an identical manner to those described in Ivison et al.(2002).The resulting radio maps have depths range from4μJy to15μJy rms;see Table1.

Deep optical imaging in the B,R,and I passbands is available for all of our?elds.This consists of several hour integrations with mosaic CCD cameras on4-m and 8-m class telescopes,taken either from public archives,or obtained by our group during observing runs throughout 2000–2002.HDF(BRI),SSA13(I),and Lockman(RI) imaging was obtained with the SUPRIME camera on the Subaru telescope.The HDF data was retrieved from the public release presented in Capak et al.(2004).The SSA13 and Lockman data was retrieved from the Subaru archive and reduce by our group.The BR imaging in SSA13was obtained with the Kitt Peak4m telescope and the MO-SAIC camera,and reduced with the MSCRED tasks in IRAF.The B imaging in Lockman and the ELAIS-N2 imaging(BRI)was obtained with the wide-?eld camera on the William Herschel Telescope(WHT)and reduced in a standard manner using IRAF.g,Rs,I imaging in SSA22and Westphal-14was obtained from the public re-lease in Steidel et al.(2003),and the details are described therein.Additional SSA22imaging was obtained with the LFC instrument(g)on the Palomar5m telescope and the CFHT-12k mosaic camera(R,I),and was reduced with MSCRED in IRAF.References to all these instruments are listed in Table1.Near-infrared imaging is also avail-able for the majority of submm sources in our?elds from a number of di?erent instruments and telescopes,typically reaching at least K=20and J=22.Details of the opti-cal,radio,and submm data in each?eld is given in Table1. The near-IR properties of our SMGs are discussed fully in Smail et al.(2004a).

SMGs with radio identi?cations allow the position of the rest-frame FIR emission to be unambiguously identi-?ed with a position in the optical imaging to within the relative astrometric alignment of the radio/optical frames. Optical images were distortion corrected,and tied to the same astrometric grid as the radio data using large num-bers of optically-bright radio sources across the?eld,re-sulting in an rms positional uncertainty of typically～0.5′′(see the detailed discussions in Richards2000;Ivison et al. 2002).The R AB-magnitudes(subsequently called R)in2′′apertures of the targets range from R=18.3to R>27.5 (Fig.1).

Targets were selected for spectroscopic followup from the seven?elds,chosen at random and prioritized equally, without preference for optical brightness.Observations of two sources were obtained with ESI echelle spectrograph on the Keck-II telescope on the night of2001July16 and have been previously discussed by Chapman et al. (2002c).Over the course of seven observing runs between 2002March and2004February we observed98of the104 sources in our radio-SMG sample with the Low Resolu-tion Imaging Spectrograph(LRIS,Oke et al.1995)spec-trograph on the Keck-I telescope8,obtaining reliable red-shifts for a total of73galaxies.The?rst ten spectroscopic identi?cations from our program were presented in C03. The details of the spectroscopic con?gurations for our observing runs and their success rates are presented in Table2.Observations taken with LRIS using several dif-ferent settings of gratings and cameras.Data taken before 2002March was obtained before the commissioning of the large mosaic CCD blue camera,and used a smaller format blue device.All subsequent data was taken with the larger format(4k×4k)blue camera(Steidel et al.2004).Our ob-servations use either the5600?A[D560]or6800?A[D680] dichroic to divide the light between the red and blue cam-eras.The400l/mm[B400]grism was always used in the blue arm to provide wavelength coverage from the atmo-spheric limit out to the dichroic wavelength for most of the slitlets on the masks.This grism provides reasonable resolution(～5–6?A)with our adopted1.2–1.4′′slitwidths. Either the400l/mm[R400]or600l/mm[R600]gratings were used in the red arm,dependent on the dichroic se-lected.Spectral resolutions of～6–8?A are achieved in the red.

Integration times were between1.5–6.0hrs in dark or gray conditions,split into30-min exposures.Conditions varied from photometric to light cirrus,and seeing ranged between0.7′′and1.1′′.Data reduction followed standard multi-slit techniques using custom iraf scripts.The spec-tra typically probe an observed wavelength range of3100–8000?A.

2.1.Spectroscopic Identi?cations

8The data presented herein were obtained at the W.M.Keck Observatory,which is operated as a scienti?c partnership among the California Institute of Technology,the University of California and the National Aeronautics and Space Administration.The Observatory was made possible by the generous?nancial support of the W.M.Keck Foundation.

4

Fig.1.—This?gure compares the properties of our spectroscopically-identi?ed sample to the parent catalog of all submm galaxies identi?ed in the radio waveband and observed spectroscopically.We show the relative distributions of R-magnitude,radio?ux, and submm?ux(the shaded histograms are the spectroscopic sample).As expected the spectroscopically-unidenti?ed galaxies are typically fainter in the optical,but have similar1.4GHz/850μm ratios,consistent with the suggestion that they are likely to lie at similar redshifts to our spectroscopically-identi?ed sample,but are on average more obscured in their restframe UV.

To obtain redshifts from our spectroscopic observa-

tions,one-dimensional spectra were compared with tem-

plate spectra and emission line catalogs.Of the98radio-

SMGs observed,redshifts were obtained with con?dence

for73galaxies,for a total spectroscopic completeness of

74%.The distribution of the optical,radio,and submm

?uxes of the parent and spectroscopically identi?ed sam-

ple are shown in Fig.1.Representative spectra are shown

in Fig.3.Table1lists the number of radio-SMGs ob-

served with LRIS in each?eld,and the number of success-

ful redshift measurements.Field to?eld variations re?ect

weather quality,as well as intrinsic source properties(e.g.,

Lyαline strength).

Twelve SMGs from our sample have pre-

viously published redshifts from other groups:

SMM J141741.90+522823.6,SMM J141742.20+523026.0

by Eales et al.(2000);SMM J030244.56+000632.3

by Webb et al.(2003);SMM J123629.13+621045.8,

SMM J123632.61+620800.1,SMM J123634.51+621241.0,

SMM J123635.59+621424.1,SMM J123607.53+621550.4,

SMM J123721.87+621035.3,SMM J131201.17+424208.1,

SMM J131215.27+423900.9,SMM J131225.20+424344.5

by Barger et al.(2001a,2001b,2003).9The R AB mags

of these sources are amongst the brightest in our sample,

with an average R=22.3±2.2.All these redshifts from

the literature are in agreement with our measurements

within errors.

Three further sources were tentatively identi-

?ed by Simpson et al.(2004)using the Subaru-

OHS spectrograph.SMM J163658.19+410523.8agrees

with our redshift as noted in Simpson et al.

(2004).However,SMM J105158.02+571800.3and

SMM J141809.00+522803.8disagree with our measured

redshifts by d z=0.20and d z=0.16respectively.Our red-

shift for SMM J105158.02+571800.3(z=2.239)is de-

rived from two UV-absorption features and the detec-

tion of Hα(Swinbank et al.2004),and we regard our

redshift as a more robust identi?cation.Our redshift

for SMM J141809.00+522803.8is derived primarily from

strong Lyαin emission,but lies at a redshift(z=2.71)

that makes it di?cult to followup in nebular lines using

near-IR spectrographs,casting some doubt on the reality

of the Simpson et al.redshift(Simpson et al.2004in fact

suggest that their redshift is likely to be spurious based

on the weakness of the features and the residuals present

from sky-line subtraction).

The primary criteria for considering a redshift as robust

is the identi?cation of multiple emission/absorption lines.

Our redshift identi?cations are con?rmed by the detec-

tion of other lines and continuum features in75%(55)of

the identi?ed spectra:AGN lines(C ivλ1549,S ivλ1397,

N vλ1240),as well as weaker stellar and interstellar fea-

tures,and continuum breaks.We also consider redshifts

to be robust if we detected a large equivalent width line

(>20?A),and there is supporting evidence that this line

is Lyα.Only one quarter(18)of the73spectroscopic

redshifts are single-line identi?cations.There are three

items of supporting evidence which are used to support

the single-line Lyαidenti?cations:the identi?cation of a 9Note that not all these sources were measured as SMGs,but were listed with spectroscopic redshifts in catalogs of radio or X-ray sources.

Chapman et al.

5

continuum break (if the continuum is detected)across the proposed Ly αline from the red to the blue (3/18),the ab-sence of emission lines which do not match the proposed Ly α-derived redshift (all 18),and an asymmetrical line pro?le,with the blue-wing truncated,typical of Ly αemis-sion from high-redshift galaxies (8/18).In very few cases were identi?cations ambiguous using these

criteria.

Fig. 2.—We show possible SEDs describing the emission

from a typical SMG,lying near the median redshift for the model-corrected sample (z =2.4),with ?ux densities of 6mJy at 850μm and 50μJy at 1.4-GHz (near our radio detection limit).Overlaid are SED templates at three dust tempera-tures (25,32&45K)spanning the typical range observed in the SMGs.In the upper panel,the SEDs are normalized to the radio point to emphasize how sources with hotter charac-teristic dust temperatures,and lower implied dust masses,are missed in the submm at z >～2.In the lower panel,the SEDs are normalized to the 850-μm point,highlighting how sources with cooler characteristic temperatures are undetectable in the radio at the redshifts higher than the sample median.

While single line Ly αidenti?cations may not be convinc-ing to some readers,several arguments support our inter-pretation.For many of the single emission line detections,the observed wavelength lies below 4000?A (sometimes be-low 3700?A ),precluding a reasonable identi?cation as [O ii ]λ3727at z <0.07given the optical faintness and submm/radio detection.

We also have two independent tests of the reliability of our redshifts.Firstly,we have obtained Keck/NIRSPEC and VLT/ISAAC near-IR spectroscopic observations for a signi?cant fraction of our sample to probe the nebular line emission to measure the star formation rates,estimate metallicities and study kinematics.These observations have successfully detected restframe H α(and frequently [N ii ])emission in 26cases (Swinbank et al.2004),con-?rming the UV-based redshifts in the present paper.Ten of the 18single-line identi?cations have been con?rmed in H α.The near-IR H αspectra were also able to break de-generacies in 5cases where spectral identi?cations based

only on UV-absorption lines were consistent with two sim-ilar redshifts.These H α/[N ii ]results were also used to aid in the spectroscopic classi?cation of our sample,as indi-cated in Table 3.

Secondly,15of our SMG redshifts have been con?rmed with CO line emission using the IRAM Plateau de Bure Interferometer (Neri et al.2003;Greve et al.2005),includ-ing 2single-line identi?cations.These detections not only con?rm the precision of the UV-based redshifts for the counterparts we targeted,but equally importantly they also con?rm that these galaxies are gas-rich systems suit-able to be the source of the luminous far-infrared emission detected in the submm waveband.

The strength of the Ly αlines for the SMGs vary tremen-dously in both line ?ux (L νfrom 1–60μJy)and restframe equivalent width,which ranges from ?3?A (absorption)to >150?A (we note that we see no obvious variation in the radio/submm properties of SMGs as a function of Ly αline strength).With the generally faint rest-frame UV continua exhibited by our SMG (65%are fainter than R AB >24.4),there is a bias in our sample against ob-taining redshifts for the weaker emission line sources at the faintest continuum ?uxes.This is re?ected in Fig.1,where the increased failure rate for obtaining spectroscopic redshifts is apparent for R AB >～24.This bias is highlighted by the fact that sources with identi?able Ly αabsorption only appear in our sample for SMGs with R AB <～25.We discuss the spectral properties and incompleteness of our sample in more detail in §3.4.

3.RESULTS

3.1.Sample properties and submm-radio selection e?ects To understand the characteristics of the submm popu-lation we ?rst have to quantify how the selection criteria for our sample (e.g.,radio,submm and optical ?ux lim-its,spectroscopic incompleteness)may have in?uenced the observed properties.

Fig.1shows our spectroscopic completeness as a func-tion of R -band magnitude,1.4-GHz radio ?ux and 850μm submm ?ux.The median properties of the parent sample are R AB =25.4±1.8,S 1.4GHz =75±127μJy and S 850μm =6.0±2.9mJy,while the spectroscopically-identi?ed popu-lation has R AB =24.6±1.7,S 1.4GHz =78±106μJy and S 850μm =5.7±3.0mJy,As expected,our spectroscopic sample is biased towards the optically brighter galaxies (the median R -band magnitude for the unidenti?ed spec-troscopic targets is R AB =26.1±1.2),but there is no discernable di?erence in the submm or radio distributions (which are e?ectively decoupled from the restframe UV emission).This suggests that the long-wavelength proper-ties of our spectroscopic sample are likely to be represen-tative of the more general submm population.

A crucial feature of the present study is that by analyz-ing only the radio-identi?ed SMG we are considering only part of the total SMG population.We must therefore de-termine the in?uence of the resulting selection function be-fore drawing wider conclusions about ?ux-limited submm samples.

About 65%of the bright (>5mJy)SMG population are detectable in the deepest radio maps obtainable with the VLA (Ivison et al.2002;Chapman et al.2003a;Wang et al.2004;Borys et al.2004).Greve et al.(2004)and

6

Fig. 3.—Representative spectra for twelve SMGs from our complete sample.The strongest UV lines used in the redshift identi?cations are marked by dashed lines.All spectra have been shifted to a common rest-frame wavelength scale.

Chapman et al.7

Ivison et al.(2005),have recently suggested that the frac-tion of bright SMGs,robustly con?rmed at both850μm and1200μm,which are detected in the radio may be even higher–～80%.The remaining～35%of SMGs not de-tected in our radio maps(and therefore not included in the distributions of Fig.1)could in principle have a wide range of properties and redshifts.

We can elucidate the e?ects of our radio pre-selection further by considering the spectral energy distributions (SEDs)of SMGs in more detail.A range of possible SEDs for a canonical6mJy radio-SMG at z=2.4with a50-μJy radio counterpart is shown in Fig.2.The upper panel shows SEDs(Dale&Helou2002)spanning a range of dust temperatures,all with the same radio?ux,and therefore comparable FIR luminosities.The range in dust tempera-tures depicted is less than a factor two,but results in close to a factor ten range in submm?ux(see Blain et al.2002, 2004a for additional details of this selection e?ect).As a consequence the hottest SED(with the lowest implied dust mass)shown in Fig.2(with a characteristic temper-ature of45K)falls below the SCUBA detection threshold in our sample of3mJy.

This trend becomes more dramatic at lower redshifts: by shifting the SED templates to shorter wavelengths,it becomes apparent that only the coolest sources can be de-tected above our～3mJy submm?ux limit.By contrast, the selection bias vanishes at higher redshifts;increasingly hotter sources become detectable with SCUBA above our radio limit(～30μJy in the typical?eld).

However,our requirement of a radio detection to pin-point the spectroscopic counterpart of the submm source means that the opposite selection bias comes into play in our sample(Fig.2lower panel);the coldest SMG’s at z=2.4lie below our radio?ux limit.This bias become an increasing concern at higher redshifts where warmer SMG’s fall beneath our radio limit.We will quantify the in?uence of our radio selection using knowledge of the range of observed SMG SED properties in§4.1.

The observed redshift distribution in Fig.4illustrates succinctly how the radio-undetected SMGs may relate to our radio-SMG sample in this context:we expect to miss sources lying between the submm and radio model curves due to our requirement of a radio detection to pinpoint the host galaxies.The radio-undetected minority of the submm population is likely to overlap signi?cantly in red-shift with our present radio-SMG sample.These galaxies would have characteristic dust temperatures that are typ-ically cooler than the radio-detected galaxies at their red-shifts(via the(1+z)–T d degeneracy).Fig.4shows that the two populations begin to overlap signi?cantly at z～2.5.

A well-studied example of a radio-undetected SMG in this redshift range is the extremely red SMM J04431+0210 identi?ed by Smail et al.(1999),the redshift of this galaxy was measured as z=2.51by Frayer et al.(2003)and con-?rmed beyond doubt as the submm source through the detection of molecular gas in the CO(3–2)line by Neri et al.(2003).

3.2.Ambiguous Radio Counterparts to Submm Sources There are a number of other issues associated with the identi?cation process which need to be considered.In par-ticular,we note that the radio positions of a handful of sources lie very near(<2′′)to an obviously low-redshift galaxy.These galaxies are unlikely to be related to the submm-emitting source,once the complete range of the galaxy and submm source properties(radio luminosity and colors)are considered.Instead we believe that these low-redshift galaxies are in fact lensing the background submm sources.Detailed discussions of two of these systems can be found in Chapman et al.(2002c)and Dunlop et al. (2004).There are?ve such galaxies in our total sample and we have not include these in the catalog of73submm sources with robust redshifts as we believe that the fore-ground galaxies are not the correct identi?cations for the submm sources.In three of these cases there is evidence from near-infrared imaging of faint K-band galaxies ly-ing closer to the radio position than the galaxy which was spectroscopically observed.These submm sources are par-ticularly di?cult to study(or even target)optically,and may have to wait for blind CO lines searches at millimeter wavelengths to determine their redshifts.For complete-ness,we include these galaxies in our identi?cation table, but?ag them as probable lenses and do not include them in any of our subsequent analysis.

We note that there are examples of low redshift(z<1) galaxies where the radio emission is coincident with the spectroscopically-targeted counterpart(within the relative astrometric errors),and we consider these to be the cor-rect identi?cations and include them in our sample and analysis.These galaxies have inferred dust temperatures which appear cold relative to similarly-luminous,local IRAS galaxies(Chapman et al.2002d;Blain et al.2004b). In addition to the ambiguity in the small number of cases where radio sources lie close to bright,foreground galaxies,Ivison et al.(2002)have demonstrated that roughly10%of radio-SMGs have more than one radio source within their submm error circle(8′′diameter,de-rived based on Monte Carlo simulations,and through com-parison to the robust identi?cations of SMGs in Smail et al.2002).Taking a well-studied example from the litera-ture:SMM J09431+4700(Ledlow et al.2002)represents a striking case of an SMG with two probable radio coun-terparts(denoted H6and H7).Ledlow et al.(2002)ob-tained a spectroscopic identi?cation of z=3.35for the optically-brighter radio source,H6.Nevertheless,both ra-dio sources were detected in1mm continuum from the IRAM PdB(Tacconi et al.2005)–con?rming that both contribute to the measured850-μm?ux.A search for CO emission in the system at the redshift of H6failed to detect any molecular emission from this galaxy,but did detect a massive gas reservoir in the optically fainter sources,H7, con?rming that both galaxies lie at the same redshift(Neri et al.2003).

We?nd eight examples of multiple radio coun-terparts to submm sources in our sample(noted in Table3).In these cases we have taken spec-tra of all of the radio sources.In three cases (SMM J123621.27+621708.4,SMM J123616.15+621513.7, and SMM J163650.43+405734.5)we obtained a spectroscopic identi?cation for one radio source but failed on the second.In three cases (SMM J105200.22+572420.2,SMM J123707.21+621408.1, and SMM J123711.98+621325.7)we con?rmed that both radio sources lie at the same red-shift to within1000km s?1.While in the?-nal two cases(SMM J105238.26+571651.3and

8

SMM J105225.90+571906.8),we found one radio source at high-redshift z>2,and the other at z<0.5.In these latter two cases,we assume that the high-redshift source is the correct identi?cation,since its signi?cantly greater radio luminosity suggests a dominant contribution to the submm emission.However,detailed multi-wavelength follow-up may reveal that the low redshift radio source also contributes signi?cantly to the submm luminosity. Even when only a single radio counterpart exists,the optical identi?cations are not always unambiguous–with two optical counterparts within the radio error-circle,or ～1′′o?sets from the radio to optical identi?cation.This a?ects only a small fraction of our sample(four SMGs or <～5%).Again,we have attempted to obtain redshifts for all components,but the sample is not complete in this respect(two possible counterparts remain to be observed spectroscopically).For the～10%of cases where the op-tical identi?cation is slightly o?set from the radio(noted in Table3),the optical source is sometimes extended and overlaps the radio source in lower surface brightness exten-sions.In these cases we are con?dent that we have identi-?ed the correct counterpart,although we have not neces-sarily characterized the source of the bolometric emission in the Keck spectrum.Chapman et al.(2004b)address some of these issues through higher spatial resolution ra-dio and optical imagery,while Ivison et al.(2005)have examined the detailed identi?cations of SMGs in a robust sample within one of our survey?elds.We have com-mented on individual objects which are subject to any of these identi?cation issues in Table3.

3.3.Redshift distributions

The redshift distribution of our submm galaxy sample (Fig.4)displays a marked peak at z～2.0–2.5,with an apparent dip in the distribution at z～1.5that is almost certainly a result of our failure to e?ciently identify red-shifts for SMG’s at z=1.2–1.8where no strong spectral features fall into our observational windows.The decline in the numbers of SMGs at low redshifts is due to a combi-nation of the submm selection function,and the intrinsic evolution in the population(e.g.,Blain et al.1999a).At high redshifts,our requirement of a radio detection(to lo-cate the submm counterpart)limits the maximum redshift detectable at our radio survey depths:a radio?ux limit of30μJy will yield detections of typical temperature and luminosity SMGs at z<3.5.

We describe the selection e?ects using an evolution-ary model which accounts for the dust properties of local galaxies with a range in template SEDs and that has been tuned to?t the submm/radio?ux distribution(Chapman et al.2003c;Lewis,Chapman&Helou2004).The model takes the local FIR luminosity function and evolves it in lu-minosity with increasing redshift following the functional form given in Blain et al.(2002).The functional form of the pure luminosity evolution function in this model is given as g(z)=(1+z)3/2sech2[b ln(1+z)?c]cosh2c.The values of b and c in this function which remain consistent with our observed redshift distribution(after?lling in the redshift desert as described in this section,and assuming

1σPoisson error bars on the N(z)histogram)are2.10+0.30

?0.40

and1.81+0.21

?0.37respectively.The range of parameters e?ec-

tively shift the z peak of the peak evolution function such that z peak=1.8+0.7

?0.3

.

Fig. 4.—The redshift distribution of our submm galaxy sample(red histogram).To interpret the likely e?ects of the sample selection on this distribution,we plot predicted model redshift distributions for submm galaxies with S850μm>5mJy (blue solid line)and radio sources with S1.4GHz>30μJy (green dashed line)based on the evolutionary models of Blain et al.(2002)and a family of long-wavelength SEDs tuned to reproduce the distribution of submm/radio?ux ratios(Chap-man et al.2003c;Lewis,Chapman&Helou2004).The ma-genta shaded region highlights the redshift range where no strong line features enter the observable wavelength range of LRIS.The blue shaded region identi?es the distribution of that proportion of the population we expect to miss due to our radio?ux limit.

We plot the predicted redshift distributions for all submm galaxies with S850μm>5mJy and all radio sources with S1.4GHz>30μJy.We expect to miss sources ly-ing between the submm and radio model curves due to our requirement of a radio detection to pinpoint the host galaxies.The galaxies in this region represent35%of the integral under the submm curve(Fig.4)in agreement with the proportion of radio-unidenti?ed SMGs(e.g.,Ivison et al.2002;Chapman et al.2003c).The combination of se-lection functions described by the model curves are clearly in good agreement with the observed redshift distribution. Contrasting the predicted and observed redshift distri-butions we would estimate that the dip due to spectro-scopic incompleteness at z～1.5would e?ect around20% of our parent sample–this is close to the26%spectro-scopic incompleteness estimated for the total sample in §3.3–suggesting that most of this arises from the so-called redshift desert.The in?uence of the radio selec-tion produces an increasing incompleteness compared to the original parent sample at z>～2.5.The model predicts that these missing galaxies will lie at somewhat higher red-shifts than our radio-SMG sample,but still overlap signi?-cantly.Under this model scenario,less than10%of SMG’s lie at z≥4where we have no identi?ed members of the population in our survey.The observed radio-SMG N(z)

Chapman et al.

9

Fig. 5.—Left panel:Comparing the N(z)for radio-SMG with obvious AGN signatures with those without(which we denote starbursts).The N(z)for a radio-selected sample of Quasars unbiased by extinction(Shaver et al.1998)is shown as a dashed curve. Right panel:The N(z)for radio-SMGs with those SMGs which were?rst identi?ed in the UV spectroscopy from a single-line (Lyαin all cases except one)highlighted as a separate histogram.The SMGs are compared with the Lewis,Chapman&Helou (2004)model for850μm sources,the Shaver et al.(1998)radio-selected sample of Quasars unbiased by extinction(comparable to the2dF Quasar evolution from Boyle et al.2000and Croom et al.2004),and the Silverman et al.(2004)sample of Chandra X-ray selected AGN.Models of Baugh et al.(2004)for>8mJy SMGs and>2mJy SMGs normalized to equal numbers of sources in each N(z)are also overlaid.

(accounting for incompleteness in the spectroscopic desert) can be reasonably approximated by a Gaussian with a cen-tral redshift of z=2.1and a sigma of1.2,while the model submm-only N(z)is?t by a Gaussian with a central red-shift of z=2.3and a sigma of1.3.

In§4.4and5,we will attempt to combine all the obser-vational information on the SMG population to con?rm the fraction of SMGs that could plausibly lie at very high redshifts.Nevertheless,we reiterate that our survey gives access to the majority(～70%)of the bright(～6mJy) SMG population and so allows us to derive representative properties for the bulk of this important population.

It is interesting to divide the SMG sample based upon their spectroscopic classi?cations.Quasars(Shaver et al. 1998;Boyle et al.2000;Silverman et al.2004)and LBGs (Steidel et al.1999;Lehnert et al.2003;Giavalisco et al. 2004)show very di?erent evolution histories from z=1–6. Similarly,Haarsma et al.(2000)and Cowie et al.(2004) have studied the evolution of lower redshift(z=0–1) faintμJy radio sources,?nding strong evolution and dif-fering evolution for subsets of this population depending on whether their restframe optical/UV spectra show AGN or star-formation signatures.As Fig.5a demonstrates, the redshift distribution of the SMGs does not depend strongly on the level of AGN activity apparent in their restframe-UV spectra.Moreover,the form of the redshift distribution for the radio-SMG population(Fig.5b)is very similar to that seen for Quasars(Shaver et al.1998;Silver-man et al.2004),selected either in the optical,X-ray or radio wavebands,and di?erent from than that seen for UV-selected galaxies.We have placed the data from Silverman et al.(2004)on Fig.5b by dividing each of their points by the appropriate comoving volume element in our adopted cosmology,and connecting the points with a spline?t. Only by adopting very contrived redshift distributions for the additional35%of the SMG population which are not currently identi?ed in the radio(Fig.4),is it possible to make the SMGs redshift distribution di?er radically from that seen for Quasars.We shall explore this point in more detail in§4.4.We also highlight in Fig.5b those SMGs which were?rst identi?ed in the UV spectroscopy from a single-line(Lyαin all cases except one),and note again that50%of these sources have been con?rmed at the cor-rect redshift through near-IR spectroscopy of the Hαor [O iii]λ5007lines.These single-line identi?cations re?ect the higher redshift tail of our measured N(z),consistent with the generally brighter optical continuum magnitudes of the z～2sources over the z～3sources.

It is also interesting to study the N(z)as a function of850μm luminosity.Model predictions of the N(z)for 850μm-selected galaxies have been presented by Baugh et al.(2004).In Fig.5b we have overlaid their model N(z) for SMGs with S850>8mJy and S850>2mJy,the lat-ter being dominated by the～2mJy sources because of the steep850μm counts.Their models are normalized to have the same number of sources in the N(z)integral.Baugh et al.(2004)?nd that the median redshift of the SMGs does not change signi?cantly over the2–8mJy?ux range.

We can divide our observed SMG sample into equal number bins with S850>5.5mJy and S850<5.5mJy. Our submm brighter galaxies(median S850=7.9mJy)lie preferentially at higher redshifts with a median redshift z=2.45(interquartile±0.35),compared to the submm fainter galaxies(median S850=4.7mJy)with median

10

z=2.01(interquartile±0.61).At face value,this re-sult disagrees with the Baugh et al.predictions.As we might expect from a radio-?ux limited survey,the radio properties of the submm bright and faint samples are in-distinguishable(submm bright:S1.4=74±27μJy,submm faint:S1.4=76±29μJy),suggesting that our radio selec-tion criterion is at the root of the discrepancy with the Baugh et al.model.For example,if the overall properties of submm-selected galaxies were similar for S850～2mJy and S850～8mJy samples(in particular the FIR-radio correlation)as in the models of Baugh et al.(2004),we would expect to miss more of the S850～2mJy sources at higher redshifts due to the radio detection criterion.As we will see in§4.1,these properties imply that the submm bright and faint sub-samples will coincidentally have indis-tinguishable distributions in dust temperature.Thus while at face value these results appear to imply that more bolo-metrically luminous SMGs tend to lie at higher redshifts, consistent with a strong luminosity evolution(see§4.2), strong selection e?ects are at present in our submm/radio sample,and we must exercise caution in our interpreta-tions of these trends.

3.4.Optical Spectroscopic Characteristics of SMGs With73spectral identi?cations,we have the statistics to begin exploring the range of galaxy types in the SMG population(Table6).Eighteen of our galaxies(25%of the sample)show clear AGN signatures in their spec-tra:three of these are broad-line AGN,with the remain-ing?fteen being narrow-line AGN.A larger fraction of SMGs(30/73)have restframe UV spectroscopic charac-teristics similar to those of star-forming galaxies,with-out any identi?able AGN signatures.The remaining25 spectroscopically-identi?ed SMGs in our sample have red-shifts which are identi?ed primarily through a strong Lyαline,and do not have the continuum signal-to-noise to rule out weak AGN features.However,we note that the lim-its on their Lyαto[C iv]λ1549line ratios are generally consistent with those expected for starbursts.We stress that the relatively high completeness of our survey relies in part upon the surprising strength of the Lyαemission from these supposedly highly-obscured galaxies.The?ux of the Lyαemission line indicates that Lyαphotons can readily escape from submm-selected galaxies and suggests that they may have very patchy and inhomogeneous dust distributions(see Neufeld1991;Chapman et al.2004b). The high fraction of UV-emission line galaxies in our sample(38/56of those at high enough redshift to detect Lyα,or68%)is striking as compared with restframe UV-selected galaxies at z=2–3,given the similarity of their broadband colors(discussed below).The Lyman-Break Galaxy population(LBGs,Steidel et al.2003)at z～3, show strong Lyαemission in only25%of cases(Shap-ley et al.2003),and their lower-redshift counterparts,the BX/BM galaxies,show an even lower fraction with Lyαemission–(Steidel et al.2004).In part,however,this might be a a selection e?ect resulting from the di?culty of measuring absorption-line redshifts for the fainter SMGs (from the total spectroscopically observed sample of98, there are38/81or47%Lyαemission-line systems in the total sample of spectroscopically-observed galaxies,again excluding the spectroscopically identi?ed SMGs at red-shifts too low to detect Lyα,z<1.7).If we only con-sider the subsample of SMGs with R AB<23.5(21galax-ies)where our spectroscopic identi?cations are e?ectively complete,Fig.1,10/21are at z>1.7to measure Lyα. 8/10of these have Lyαin emission.

In addition to the73SMGs with robust redshifts there are25of the optically-faintest radio-SMGs for which we have spectroscopic observations,but failed to measure a redshift.None of these sources appear to exhibit strong, narrow emission lines in the observed UV/optical wave-bands.There are two possible redshift ranges in which this population could reside.Firstly,they may lie at z～1.2–1.8,where no strong emission or absorption features fall in the sensitive range of the LRIS spectrograph.Alter-natively,they may lie within the redshift distribution of our spectroscopically-identi?ed sample,but have spectra which are characterized by absorption lines,in particular absorbed Lyα.As these tend to be amongst the faintest sources in the sample(Fig.1),the signal-to-noise ratio in their spectra is often lower than for our successful spectro-scopic identi?cations.

To summarize,from the98radio-SMG for which we ob-tained spectroscopic observations:18%show obvious AGN characteristics;31%are apparently star-forming galaxies; 25%are di?cult to classify,but remain reasonable can-didates to be star-forming galaxies;and26%are spectro-scopically unidenti?ed,and so are unlikely to be strong AGN at z<1.2or z>1.8(but could be star-forming galaxies with weak/absent emission lines at almost any redshift).Assuming that the redshift distributions for the AGN and starburst populations are comparable,includ-ing those spectrally-unclassi?ed SMGs at z=1.2–1.8with AGN spectral signatures is unlikely to increase the frac-tion of such galaxies in the total sample above25%(see §3.5).This is a lower AGN fraction than suggested by either the identi?cation of X-ray counterparts with QSO-or Seyfert-like luminosities(>～36%,Alexander et al.2003) or Hαline widths of≥500km s?1(46±14%,Swinbank et al.2004).We conclude that perhaps a third of SMG’s identi?ed as apparently star-forming or unclassi?ed based on their restframe UV spectral properties are likely to host unidenti?ed(most likely obscured)AGN.

3.5.Optical Photometric properties of SMGs

In the deepest ground-based optical images the majority of SMGs are detected in most or all wavelengths observed: this is particularly true in the HDF-N and Westphal-14hr/SSA22?elds for which extremely deep UBR imag-ing from Subaru/SuprimeCAM and Kitt Peak/MOSAIC imaging(Capak et al.2004),and deep U n,g,R s(hereafter UgR)images(Steidel et al.2003)exist respectively.

In Fig.6we show the(U?g)–(g?R)color-color diagram for those radio-SMGs with robust spectral identi?cations and redshifts z>1.5.We compare these to the z～3 and z～2selection criterion presented in Steidel et al. (2004).Color transformations were determined between ?lter bands by matching the catalogs of z～2and z～3 galaxies lying in the extended-HDF region,and rederiving the colors from the Capak et al.images.The UBR AB-magnitudes of the Steidel et al.(2004)BX/BM galaxies in the HDF region were?rst measured using the SUPRIME and MOSAIC images of Capak et al.(2004).A new BX color-selection box was then de?ned empirically for these UBR?lters.The median o?set in magnitudes U?U n,

Chapman et al.

11

B ?g ,R ?R s (0.22,0.13,0.04)was then applied to our measured UBR -mag for the

SMGs.

Fig.6.—The observed frame (U ?g )–(g ?R )color-color dia-

gram for radio-SMGs in the HDF-N and Westphal-14hr ?elds,where deep photometry exists.Circles depict SMGs with red-shifts,squares show those galaxies with AGN spectra,and small squares denote radio-SMGs without spectroscopic iden-ti?cations.Most of these latter category have only a lower limit on their (U ?g )colors.The color regions corresponding to z ～3(upper region)and z ～2(lower region)selection are shown (Steidel et al.2004).All SMGs lying within the denoted z ～2,z ～3regions have spectroscopic redshifts con-sistent with their colors.All colors are on the AB-scale.

Many of the SMGs (～65%)lie well within the z ～2color selection region from Steidel et al.(2004),in accord with their spectroscopic redshifts.Approximately 30%of the galaxies are too red in (g ?R )to be selected by the restframe-UV criterion,possibly because of dust extinc-tion.Smail et al.(2004a)present a complete study of the extinction properties of SMGs using near-IR photometry.The relative classi?cation of AGN and SB populations of SMGs suggest that the strong emission-line AGN tend to-wards redder g ?R colors with a median g ?R =0.5±0.1,compared with the non-AGN sample g ?R =0.4±0.1.We also show the colors of sources for which we failed to obtain spectroscopic redshifts.Many of these are detected near the limit of the photometry in the BR passbands,but are undetected at U ,and are thus shown as lower limits in (U ?g )in Fig.6.Many of these galaxies also lie within the z ～2color selection region,but with much larger uncer-tainties.This provides some additional evidence that the radio-SMGs for which we were unable to obtain spectro-scopic identi?cations likely span a similar range in redshift to those with robust identi?cations.

While the colors of many SMGs appear consistent with the BX-selection,a signi?cant fraction (～50%of the to-tal radio-SMG sample,and ～30%of the spectroscopically identi?ed radio-SMGs)are too faint to be selected in typ-ical BX/LBG samples (typically R s <25.5–Steidel et al.

2004).All SMGs lying within the denoted z ～2,z ～3re-gions have spectroscopic redshifts consistent with their col-ors.SMGs lying outside these boxes typically have z ～2but redder g ?R than BX galaxies,or have colors which are very blue for their redshifts,likely a result of their AGN nature.Of the SMGs depicted in Fig.6(having z >1.5and lying in the HDF,Westphal-14,SSA22?elds),the me-dian and average quartile distribution of the total sample is R =25.3±0.8,while the distributions for the spectroscopic successes and failures are respectively,R =24.9±0.9and R =25.7±0.4.

This exercise suggests that deep optical imaging may provide reliable photometric redshifts,immune from the temperature uncertainty that plagues simple ra-dio/submm photometric redshift estimates (see §3.6).Us-ing the hyper-z software (Bolzonella et al.2000)for this sample and their UgRiK magnitudes,we derive photo-metric redshifts with a median error of ～30%(see also Smail et al.2004a and Pope et al.in preparation).

3.6.Submm/Radio indices and redshifts

In Fig.7,we show the ratio of submm to radio ?ux as a function of redshift for our SMG sample.We also assess the locations of the eight radio-identi?ed submm galax-ies from the literature with robust redshift identi?cations.The variation of the submm/radio ?ux ratio with redshift is the basis for the Carilli-Yun redshift indicator (Carilli &Yun 1999,2000).However,this diagram also depicts the variation in SED properties spanned by the SMGs,al-though these are subject to considerable selection e?ects.This ?gure is in essence a depiction of the joint evolution in the FIR/radio correlation and the range in dust properties in luminous galaxies (see §4.1).

We overlay the tracks of two galaxy classes on Fig.7,representative of quiescent spirals,such as the Milky Way,and ultra-luminous galaxies (L bol ～1012L ⊙),such as Arp220.We also show the track derived by Yun &Carilli (2002)using a compilation of low and high redshift galaxy samples.This latter track,and other similar ones from the literature (Dunne et al.2000;Barger,Cowie &Richards 2000),have been widely used to estimate SMG redshifts.If the SMGs were to represent a population with a nar-row range of spectral energy distributions,then we would expect them,on average,to trace a well-de?ned track in the CY diagram.As they appear to be widely scattered at a ?xed redshift (an order of magnitude range in the ?ux ratio at z >～1),a natural interpretation is that they span a range in SED shapes (characterized by T d ,§4.1).We note that our radio selection criteria may bias us towards identifying objects with enhanced radio-to-FIR emission (similar to the local IRAS galaxy,Mrk 231,whose AGN contributed a comparable radio luminosity to the starburst in this galaxy).The preferential inclusion of these systems in our radio-identi?ed sample would lower the typical ?ux ratio of the high redshift SMGs in our S 850μm /S 1.4GHz di-agram by about 0.3dex.Yun,Reddy &Condon (2001)suggest that such galaxies could make up a few to 10%of submm galaxies,implying at most a small bias.

12

Fig.7.—A plot of S 850μm /S 1.4GHz versus redshift for the

radio-SMGs with spectroscopic redshifts from our sample and from literature sources not lying within our ?elds (Smail et al.2002,2003b;Chapman et al.2002d).The shaded region shows the ±1σenvelope of the rms dispersion calculated in three redshift bins containing equal numbers of SMGs.The predicted variation in ?ux ratio for three SEDs are overlaid:Arp 220,a quiescent spiral galaxy (such as the Milky Way)and an empirically-derived track from the literature (Yun &Carilli 2002).

To quantify the dispersion of S 850μm /S 1.4GHz we divide the sample in redshift into three bins containing equal numbers of SMGs and plot the 1-σdispersion envelope on Fig.7.Clearly a large dispersion in SED proper-ties is spanned by our SMG sample,with a preference for the warmer templates (the Arp220,rather than quies-cent models).The trend of submm/radio ?ux ratio with redshift is much ?atter than the rapidly rising template tracks,and can be parameterized by:

S 850μm /S 1.4GHz =11.1+35.2z,

with an average RMS dispersion in the range z =1–4of σ(S 850μm /S 1.4GHz )～40.The shallow slope of this trend greatly increases the error in the redshift estimate for a given submm/radio ratio.This behaviour can be partially explained by the fact that lower redshift sources tend to be lower luminosity and cooler,following the weak local cor-relation between temperature and luminosity (Chapman et al.2003c).However,this should be only a slight e?ect.The selection criteria in the submm/radio are mainly re-sponsible for the narrow range of submm/radio values ob-served,and almost no discrimination is achieved for indi-vidual source redshifts,where ?z ～1(see also Blain et al.2004).It should be noted that a purely submm-selected sample should show an even wider range in submm/radio ?ux ratios than our sample –which already demonstrates that the range of SED properties in the SMG population render simple photometric redshift estimates too imprecise to be useful for predicting redshifts for individual galaxies

(e.g.,Aretxaga et al.2003;Efstathiou &Rowan-Robinson 2003;Wiklind 2003).

The Carilli &Yun redshift estimator has been exten-sively employed (Smail et al.2000;Barger,Cowie &Richards 2000;Chapman et al.2001,2002b;Ivison et al.2002),but until now it has not proved possible to critically compare it to a large sample of SMG’s with precise red-shifts.Perhaps surprisingly,the median redshift predicted for the SMG population using S 850μm /S 1.4GHz (z ～2.5)has turned out to be remarkably close to the value we de-rive.In part this is because,on average,the SMGs have SEDs similar to local ULIRGs and so the average prop-erties derived for the population are not too far wrong.However,this e?ect is further aided by the relatively nar-row redshift distribution seen for SMGs –which limits the intrinsic dispersion in the median redshift for the popula-tion.

4.DISCUSSION

4.1.Dust temperatures and Luminosities

Having constrained the redshift distribution and basic observable properties of the SMGs,our next goal is to study the distribution of their SED properties and bolo-metric luminosities.

Studies of local and moderate redshift galaxies in the ra-dio and submm wavebands suggest that variations in both the dust properties and the empirical relation between the FIR luminosity (L FIR )and 1.4GHz radio (Helou et al.1985;Condon 1992)will a?ect the observed radio ?ux from a SMG.The dust properties in luminous IR galax-ies such as SMGs are potentially very complicated,re-quiring many parameters for a complete characterization.However,Blain et al.(2002)have demonstrated that the characteristic dust temperature (T d –the best single tem-perature grey-body ?t to the SED)is the main parame-ter in?uencing the ratio of submm to radio emission and hence the observed radio ?ux.Since we only have a red-shift and two photometric points in the relevant region of the SED (850μm and 1.4GHz)from which to disentangle the dust properties,we will thus concentrate only on the dust temperature,?xing the remainder of the dust proper-ties at canonical values for local Ultra-Luminous Infra-Red Galaxies (ULIRGs).In the absence of additional informa-tion for most of our sources,we adopt a dust emissivity of β=1.5.

We therefore proceed by translating our observables (submm and radio ?uxes,and redshift)into intrinsic phys-ical properties,T d and total infrared luminosity L TIR (de-?ned as the integral between 8μm and 1100μm,and deriv-able from FIR with a small color-correction term –e.g.,Dale et al.2001).We calculate T d explicitly by adopting a suite of dust SED templates from Dale &Helou (2002),spanning an equivalent range of T d from 15–90K (e.g.,Fig.8).

Chapman et al.13

Fig.8.—Redshifted SED templates for galaxies from Dale

&Helou(2002)that are consistent with the radio and850μm

?uxes of16representative SMGs from our sample.The galax-

ies span a range in dust properties,translating into charac-

teristic temperatures from～20K to～60K if the FIR to radio

correlation holds at high redshift.We mark on the wavelengths

corresponding to observations with SCUBA at850μm,VLA

at20cm,SHARC-2at350μm and Spitzer/MIPS at24μm.

The templates assume the median value of the local

FIR/radio relation(Helou et al.1985)allowing a one-to-

one mapping of S850μm/S1.4GHz?ux ratio to T d.These

template?ts to a representative selection of galaxies from

our sample are shown in Fig.8.We also assume the

low-redshift FIR/radio relation to calculate L TIR for our

sources.Radio luminosities are calculated by?rst K-

correcting the observed?uxes to the restframe using a

synchrotron spectrum with an index ofα=?0.75.Our

adopted spectral index is supported by the small number

of galaxies in our sample which have measured radio spec-

tral indices(Ivison et al.2002;Richards2000).In this

parametrization,the location of a galaxy with a measured

redshift in Fig.7is uniquely described by its value of T d.

In Fig.9we plot T d versus L TIR for our SMGs;the me-

dian T d is36K(12K interquartile range)and the median

L TIR=8.5+7.4

?4.6×1012L⊙,but note the strong correlation

between observed luminosity and temperature.With our assumption of the FIR–radio relation,and?tting to the observed S850μm/S1.4GHz distribution for our SMGs,the dust temperature is approximately proportional to

T d∝

1+z spec

14

450μm ?uxes (in parallel with our 850μm observations),our 450μm data is typically of insu?cient quality (due to the weather conditions)to usefully constrain the SEDs,failing to individually detect the vast majority of SMGs.However,stacking our 450μm observations does verify that the FIR/radio and 850μm points appear to predict T d reasonably well,although calibration uncertainties at 450μm make this a di?cult measurement (see also Dunne et al.2002).The average (inverse variance-weighted)mea-sured ?ux is S 450μm =32±6mJy for our SMG sample,while the median FIR–radio ratio predicted S 450μm =41±12mJy.As we have noted,we do not have su?cient signal-to-noise on an object-to-object basis at 450μm to determine how tight the correlation

is.

Fig.9.—Characteristic dust temperature (T d )versus the log

of the total IR luminosity for radio-SMGs with spectroscopic redshifts from our sample.We see a tight trend of inferred tem-perature with IR luminosity –but propose this mostly arises from the selection criteria for our sample.The Chapman et al.(2003c)derivation of the median and interquartile range of local IRAS galaxies from the 1.2-Jy 60μm catalog are shown,linearly extrapolated to 1013L ⊙.The average T d for z =0.3–1μJy radio sources from Chapman et al.(2003a)are shown and agree well with the trend from IRAS distribution.The submm ?ux limit precludes detection of sources in the shaded region (shown for 3mJy,the weakest source ?ux in our sample).The average error bar for galaxies in the SMG sample is shown at the median temperature of the sample,o?set arbitrarily in luminosity.

At somewhat shorter wavelengths,Kovacs et al.(2004)have detected more than 10of our sample using the 350-μm SHARC-2camera on the Caltech Submillimeter Ob-servatory.These results suggest that our T d predictions are accurate to ～20%(without considering systematic cal-ibration e?ects of SHARC-2),corresponding to an uncer-tainty in the total IR luminosity of ～85%(L TIR ∝T d 3.5for z ～2.2sources).This implies that the FIR/radio cor-relation can be used to predict the dust temperature of a typical SMG with reasonable precision,and therefore that

radio observations alone could be used to estimate FIR lu-minosity of a star-forming galaxy for many purposes,once the spectroscopic redshift has been measured.

Fig.9provides a complementary view of the S 850μm /S 1.4GHz –redshift diagram in Fig.7,with the red-shift information now hidden in the luminosity and dif-ferences in S 850μm /S 1.4GHz manifested as changes in T d .Fig.9does however allow for a direct consistency check with the properties of local and moderate redshift IRAS galaxies,of which our radio-SMG population could rep-resent high-redshift analogs.To provide a stepping-stone to the high-redshift SMG population,we have also cal-culated the T d values for z =0.3–1μJy radio galaxies (Fig.9)from Chapman et al.(2003a)in the same manner as for our radio-SMG sample,using the S 850μm /S 1.4GHz ratio.These allow us to con?rm that the form of the T d –L relation from the local IRAS samples appears to hold out to z ～1.Note however that these lower redshift radio sources are not individually detected in the submm,and so the error bars are dominated by measurement errors,rather than a true distribution of the T d values.The av-erage submm ?ux densities of these galaxies are ～1mJy,and so they fall within the shaded submm ?ux selection region of the diagram.

Fig.9demonstrates that the distribution of the radio-SMG population appears inconsistent with the expected local T d –L relation characterized by Chapman et al.(2003c)(see also Blain et al.2004b).The median T d of our SMGs (36K)is lower than the locally predicted T d for galaxies of these luminosities (at L TIR =1013L ⊙,the local median is 42K).

As we noted in §3.6,our radio detection criterion might be expected to bias us towards galaxies with SEDs simi-lar to Mrk 231,with radio excess over the local FIR/radio relation,which make up 10%of local IRAS galaxies (Yun,Reddy &Condon 2001).Such galaxies would be o?set to hotter temperatures in Fig.9,in the opposite sense to the o?set we see and are therefore not likely to rep-resent a large fraction of our SMGs.Nevertheless,the de-creased median T d (relative to the local prediction)could still be a selection e?ect,since we have potentially missed luminous galaxies with both hotter and colder dust tem-peratures through our combined radio/submm selection function.Chapman et al.(2004b)have uncovered a sam-ple of apparently hot and luminous galaxies (lying above our submm selection boundary)whose inclusion would in-crease the median T d signi?cantly.In contrast,Ivison et al.(2005)have suggested likely identi?cations for submm galaxies in our sample that lack radio detections;these galaxies would appear on Fig.9with colder T d than our radio-SMGs.Both of these missing populations of lumi-nous galaxies suggest that the true T d -scatter in high-redshift,luminous galaxies is larger than observed locally.The scatter in the observed temperatures of our SMGs (12K interquartile range)is already larger than observed locally (8K interquartile range).While we note that some broadening of the local distribution is suggested for the highest luminosity galaxies,any additional hotter or colder SMGs in the high-redshift sample would only increase the this di?erence in the dispersion of the two populations.On their own our spectroscopic redshifts and ra-dio/submm photometry for radio-SMGs can not provide a complete picture of the range in SEDs spanned by the

Chapman et al.15 most luminous galaxies in the Universe.However,the

available evidence suggests that some caution should be

exercised when assuming that the T d–L properties of lu-

minous,high-redshift dusty galaxies are identical to those

at low redshifts.

4.2.The bolometric luminosity function

Estimates of radio-SMG dust temperatures using the

FIR–radio correlation allow the calculation of bolometric

luminosities,and we can use the e?ective volume of our

submm redshift survey to construct a bolometric luminos-

ity function in two redshift ranges at z=0.9±0.3and

z=2.5±0.5(Tables4,5).

The850-μm?uxes of the galaxies in our sample span a

factor of5range,translating into a similar range in submm

luminosities.By including the temperature information

we can estimate bolometric luminosities,which span a

slightly larger range,～10×.We note again that our

bolometric luminosity calculations assume that the T d–

L properties of luminous,high-redshift dusty galaxies are

similar to those at low redshifts,and some caution should

be exercised when interpreting the results.

Computing the raw volume densities of the radio-SMGs

is accomplished using the radio luminosity and an acces-

sible volume technique as described in Avni&Bahcall

(1980).We adopt a general form for the luminosity func-

tion

Φ(L)?L= i1

16

function based on FIR–UV-βrelation (Meurer et al.1997)applied to a survey of LBGs from Adelberger &Steidel (2000),and normalized to the e?ective volume contain-ing the z ～3LBGs (Steidel et al.1999).For the z ～2BX/BM population (Steidel et al.2004),stacking analy-sis of their radio and X-ray emission have suggested that the bolometric luminosity function of Adelberger &Stei-del (2000)is a reasonable representation (Reddy &Steidel

2004).

Fig.10.—The FIR luminosity functions at z =0.9±0.3and

z =2.5±0.5.L FIR is calculated by integrating under the tem-plate SED from the Dale &Helou (2002)catalog which best ?ts the submm and radio ?uxes of the galaxy at its measured redshift.For comparison,we show the local FIR luminosity function from Chapman et al.(2003c),constructed in a con-sistent manner with our z =2.5function using the Dale &Helou (2002)templates.Also shown are the submm detec-tions (>3σ)of LBGs from Peacock et al.(2000)for SFRs >1M ⊙in the HDF,and for those galaxies with inferred SFRs >100M ⊙from Chapman et al.(2000,2001b).For reference we plot the predicted Φ(L FIR )for LBGs from Adelberger &Steidel (2000)which assumes the FIR/UV–βrelation (where βis the UV continuum slope)of Meurer et al.(1997).

It is di?cult to determine whether pure luminosity evo-lution can explain the dramatic increase in volume den-sity of high-redshift galaxies with 1011–1013L ⊙(based on the UV-and submm-selected samples)over similarly lu-minous local galaxies.The high-redshift,FIR luminos-ity function remains poorly constrained in both low-and high-luminosity regimes.For UV-selected galaxies at high redshift it is currently very di?cult to accurately estimate L FIR (e.g.,Adelberger &Steidel 2000;Reddy &Steidel 2004).In addition,it is unclear what the completeness is in Φ(L FIR )for a UV-selected sample.The submm-estimated Φ(L FIR )at high luminosities is also poorly constrained due to incompleteness e?ects that are di?cult to characterize (as described above).We have already explored the extent to which the submm-selection itself may have signi?cantly underestimated the total volume of luminous galaxies at

high redshifts,as galaxies with hotter characteristic dust temperatures are missed by submm selection (Chapman et al.2004b;Blain et al.2004a).Similarly,our radio-pre-selection means we are missing a small fraction of colder SMGs without radio counterparts which may contribute to the number density of z ～2.5galaxies with L FIR >1012L ⊙(Ivison et al.2005).We anticipate the deep Spitzer obser-vations of all these high-redshift galaxy populations will shed additional light on this complex question.

4.3.The restframe-UV derived L FIR of SMGs By far the best-studied population of high-redshift star-forming galaxies is that identi?ed through their restframe UV-emission (Steidel et al.1999,2004).These galaxies have provided unique insights into the evolution of the star-formation density in the Universe and the correspond-ing formation of normal galaxies (Madau et al.1996).A key question for SMGs is to understand how they ?t into the framework de?ned by the UV populations –in par-ticular,how well does the recipe for deriving SFRs for UV-selected galaxies work on this restframe FIR-selected population?

We can use our multi-color optical data for a subset of the SMGs to investigate their restframe UV properties and derive SFRs in an analogous manner to that applied to LBGs.This analysis relies on estimating the luminosi-ties and spectral slopes at wavelength around 1500?A in the restframe.For galaxy populations at z ～2–3,this can be achieved using B -and R -band photometry for the HDF sources,and using g -and R -band photometry for the SSA22and Westphal-14sources.

As described in §2,we measured BR -band photometry for all SMGs in our sample.Since UV-derived luminosities and the corrections for dust extinction,are highly suscepti-ble to photometric errors (Adelberger &Steidel 2000),we need to isolate a sample of SMGs with well-measured pho-tometry.Unfortunately,most SMGs at higher redshifts are faint and as a result to obtain a reasonable sample we are required to use those galaxies where the photo-metric errors are only less than 0.1mags in both bands.We de?ne the subsample which includes all 33SMGs ly-ing in the HDF,Westphal-14,and SSA22?elds,with red-shifts z >1.5to allow for accurate measurement of the dust-corrected UV luminosity (the same sample used in Fig.6).We also consider the HDF subsample on its own (17SMGs),since its large size and contiguous areal cov-erage make it statistically representative on its own.The sample has a median R AB =24.9and a 1σrms of 1.1,and a median photometric error of dR AB =0.04.The dust-corrected luminosities are biased in a manner which is dif-?cult to quantify.The optically-faintest sources will have very small UV luminosities,but may in principle have very steep continuum slopes,with large implied dust correction factors (Adelberger &Steidel 2000).While we will calcu-late UV luminosities for all SMGs in this subsample,we identify those SMGs showing AGN spectra,and conserva-tively exclude them from the average properties calculated below.

To estimate the SFR from the UV,we follow the pre-scriptions of Meurer et al.(1997)and Adelberger &Stei-del (2000):the UV luminosities are ?rst corrected for line blanketing in the Ly αforest,and then corrected for dust extinction using the UV continuum slope derived from the

Chapman et al.17 (G?R)color,corresponding to wavelengths between rest-

frame1000?A and2900?A.The transformations from the

measured(B?R)to(g?R)in the HDF?eld are described

in§3.4.We correct the(g?R)color for the opacity of

the Lyαforest according to the statistical prescription of

Madau(1995).Values of(g?R)corr ranging from0.0to

1.0correspond to a UV spectral indexβ=?2to+0.6,

when the spectrum is approximated by a power law of the

form Fλ∝λβ.The(g?R)corr color is then mapped to

a color excess,E(B?V),from which the dust-corrected

UV luminosities are derived.Our medianβ=?1.5±0.8,

corresponds to E(B?V)=0.14±0.15for a Calzetti ex-

tinction law,very close to the distribution for LBGs pre-

sented in Adelberger&Steidel(2000),suggesting that the

UV identi?cations of SMGs do not distinguish themselves

from the general LBG population with signi?cantly redder

UV spectral slopes(c.f.Smail et al.2004a).

The dust corrected UV luminosity translates into a SFR,

following Kennicutt(1998):

SFR(M⊙yr?1)=1.4×10?28L1500(ergs?1Hz?1),

where the relationship applies to galaxies with continuous

star formation over time scales of108years or longer.For

younger stellar populations,the UV continuum luminosity

is still increasing as the number of massive stars increases,

and the above equation will underestimate the SFR.We

measure a median SFR=13M⊙yr?1from the dust cor-

rected UV.Finally,to compare these estimates with those

from the FIR we simply convert our UV-derived SFR di-

rectly into a FIR luminosity(Kennicutt1998).

From the dust-corrected UV,we predict a median FIR

luminosity of0.055+0.02

?0.04

×1012L⊙;this compares to the

submm/radio measurement of L FIR=5.6+2.1

?1.6×1012L⊙.

This corresponds to a median radio/submm-to-UV ratio in the derived L FIR of100,with a quartile range of29–168.We reiterate that the UV estimate has been cor-rected for dust extinction in the standard manner.The dust-corrected UV-estimated FIR luminosities are com-pared directly with the FIR luminosities measured from the radio/submm in Fig.11.The relative o?set between the two does not di?er signi?cantly if we use the more statistically reliable subsample from the HDF:L FIR,UV=

5.28+1.51

?3.40×1010L⊙versus L FIR,radio=5.79+1.96

?1.41

×1012L⊙.

In addition,we note that those sources in the HDF with the faintest apparent R-band magnitudes do not have signi?cantly di?erent UV-inferred L FIR from those with brighter R magnitudes.

As suggested above this procedure of estimating the FIR luminosity from the UV continuum slope is meaningless for the SMGs exhibiting strong AGN-signatures in their spec-tra,since the UV continuum is not necessarily dominated by stellar emission,and so we have not included these AGN-classi?ed SMGs in the median calculations above. We use di?erent symbols for the SMGs with AGN spectra in Fig.11to identify them to the reader.

Since the FIR–radio luminosity relation is essentially lin-ear except at the faintest extreme(Condon1992),the re-lation at high redshift would have to vary by a similar factor to our observed discrepancy(～100)in order for the luminosity estimates to be in accord.This is highly unlikely given the relatively well understood physics of the relation(Condon1992).However,direct measurements of the FIR-radio relation at high redshift are required to re-fute this possibility.(We can begin to rule out evolution of FIR–radio at the implied level～100×,based on previous discussion in§4.1–Kovacs et al.in preparation.) Clearly the dust-corrected UV luminosity using the stan-dard prescription rarely hints at the huge bolometric lu-minosities measured in the radio/submm,underestimat-ing the true luminosity by over two orders of magnitude. Three of the SMGs do have UV-inferred FIR luminosi-ties within a factor of three of that observed in the ra-dio/submm.One of these shows hybrid SB/AGN char-acteristics and complicated multi-component kinematics (SMM J163650.43+405734.5–Smail et al.2003a).The other two are apparently starbursts which reveal the true magnitude of their bolometric luminosity in the

UV.

Fig.11.—The FIR luminosity of SMGs(used as a proxy for their SFRs)as measured from the radio and submm?ux den-sity ratio,compared with the FIR luminosity estimated from the UV luminosity and spectral slope.The line is a simple equality,L FIR(radio)=L FIR(UV),the expected correlation if the dust-corrected UV luminosity is a reliable measure of the total star formation rate in these systems.Note the large o?sets of most SMGs from this line.Error bars are derived from uncertainties on the radio,submm,B-or g-,and R-band ?uxes.SMGs showing obvious AGN spectra are shown by large squares.There are no upper limits as we include only those SMGs in the HDF,Westphal-14hr,SSA22,and ELAIS-N2?elds which are detected in both the B and R bands. The very large disparity we derive for the UV-and radio-estimated luminosities of UV-detected SMGs is apparently at odds with the conclusion of Reddy&Steidel(2004) who?nd that radio(and X-ray)estimates of the SFRs for z～2BX galaxies on average match the UV-derived SFRs,with an average dust correction factor of just4.5. How do we reconcile this with our?nding that SMGs have UV-inferred SFRs which underpredict the radio-measured SFR by a factor of100on average?

One possible contributing factor is that there may be a wide range in dust obscuration in the high-redshift galaxy

18

population.Chapman et al.(2004b)used MERLIN radio images and HST UV images of SMGs at ～0.3′′resolution to study the di?erential emission between the two wave-lengths.They demonstrated that the radio emission (and by implication submm emission)is always more compact than the UV as traced by HST imagery.In addition they found that the radio highlights regions of low UV emission as bolometrically-luminous in ～50%of the SMGs.An in-creasing proportion of very-highly obscured activity may therefore be present in the more active systems.This has also been demonstrated through measurements of the neb-ular H αemission line for our SMG sample (Swinbank et al.2004).In this way the selection of ?ux-limited samples of galaxies in the restframe UV and FIR would give sig-ni?cantly di?erent mean obscurations.This is equivalent to stating that the UV luminosities and spectral slopes are measurable for only the least-obscured regions of the galaxies and hence are not representative of the bulk of the emission in these galaxies.It is not therefore surpris-ing that they indicate much lower bolometric emission.We therefore conclude that although many SMGs can be detected,and their redshifts estimated or measured in deep observations in the restframe UV,they cannot gener-ally be identi?ed as bolometrically luminous galaxies with-out the use of radio or submm observations.There is a second,related important point to make before we pro-ceed to discuss the evolutionary history of the luminosity density contributed by SMGs versus UV-selected galax-ies over the lifetime of the Universe.In the light of the vast mismatch in the derived bolometric emission for this population we will assume that the contribution of bright SMGs to the star formation density at high redshifts is not included in current UV estimates (e.g.,Madau et al.1996;Steidel et al.1999)–and so we need to derive the contri-bution from the highly obscured population independently and sum this with that from the UV to derive the total.

4.4.Luminosity and Star Formation histories Using the bolometric luminosities measured above for the SMGs,we calculate the following bolometric lu-minosity densities in redshift bins from our survey:

ρL (107L ⊙Mpc ?3)=2.2±2.3,8.0+4.9?3.2and 3.4+4.6

?2.5for redshift intervals of z =0.5–1.5,1.8–2.6and 2.6–3.5re-spectively.These luminosity densities are calculated for the entire radio-identi?ed SMG population,and corrected for spectroscopic completeness in the following way.For the 26%of radio-SMGs for which we did not obtain red-shifts,we corrected the clear de?cit of sources in the red-shift range z =1.5–1.8to result in a smooth distribution matching our normalized model (Fig.4).The remainder of the incompleteness (17%)was distributed uniformly over the entire N (z )uncovered by our robust redshift sample.Our justi?cation for this procedure is two-fold.First,the S 850μm /S 1.4GHz properties of sources with robust redshifts are indistinguishable from those where we did not obtain redshifts,suggesting a similar range in T d and z .Sec-ondly,the UBR colors of these two samples are similar,although with large uncertainties as many of the spectro-scopic failures are typically very faint,and are consistent with colors of z ～2star-forming galaxies (Steidel et al.2004):see §3.4.

In order to justify translating these luminosity densi-ties into star formation rate densities (SFRDs)we must

?rst clearly identify the signs of AGN activity in the indi-vidual galaxies,and then remove any contribution to the luminosities of these galaxies from the AGN,either direct contribution to the radio emission or through heating of dust by the AGN.

We begin by assessing the possible AGN con-tribution in individual submm galaxies from their UV spectral properties.Three SMGs are iden-ti?ed with QSOs in our radio-identi?ed sample (SMM J123716.01+620323.3,SMM J131215.27+423900.9,and SMM J131222.35+423814.1):these are the only ob-jects in which the optical luminosity is a non-negligible fraction of the bolometric luminosity.These sources have comparable optical and FIR luminosities and space den-sities to submm-detected QSOs at z >2(Omont et al.2001,2003;Carilli et al.2001).The fraction of optically bright QSOs in the SMG population could actually be slightly higher since ～20%of our SMGs were pre-selected with optically faint magnitudes.This implies that only ～4%(3/80,as a fraction of the total radio-SMG sam-ple observed spectroscopically,which weren’t pre-selected as optically faint)of optically-bright AGN have L FIR emission with comparable luminosity to that seen in the optical waveband (since our survey would have uncovered all the QSOs emitting strongly in the radio/submm).This fraction of radio-SMGs are therefore removed from consid-eration in determining the luminosity density evolution.There are several strong indications that star formation dominates the luminosity of the majority of the 98radio-SMGs in our sample.As discussed in §3.3,30%of our sample with the brightest UV continua exhibit stellar and interstellar absorption lines,implying that their continua are dominated by young,massive stars and not the non-thermal power law spectrum of an AGN.Another 25%of our sample have UV continua which are too faint to detect absorption features,but remain consistent with starbursts,while a further 25%do not reveal enough features to even identify a redshift.These statistics suggest that ～80%of radio-SMGs do not harbor an partially-or un-obscured,luminous AGN.

The presence of an obscured AGN is much more di?cult to determine from the restframe UV spectra and instead we have to turn to multiwavelength surveys of samples of SMGs.High spatial resolution radio observations of SMGs are one route for determining the morphology of their ra-dio (and by implication FIR)emission and hence search for the presence of a strong AGN contribution to the FIR (identi?ed as a radio point-source).Such studies reveal ex-tended morphologies on scales >1′′in 65%of cases (8/12galaxies in Chapman et al.2004b).These observations suggest that AGN do not dominate the FIR emission of most SMGs,as star formation is the most likely route to produce spatially-extended FIR/radio emission.Several of the galaxies with spatially-extended radio emission show AGN emission lines in their rest-frame UV spectra,un-derlining the fact that intense starbursts are likely to be energetically-important,even in the presence of AGN.A second route to search for obscured AGN is to em-ploy observations in the X-ray waveband.The proportion of the submm population detected in deep Chandra and XMM-Newton X-ray surveys,which are sensitive to even strongly dust-obscured AGN (Ivison et al.2002;Barger et al.2002;Almaini et al.2003;Alexander et al.2003,2005),

Chapman et al.19 at?ux densities signi?cantly greater than expected from

star formation alone is at most30%.X-ray spectral anal-ysis reveals that most of the SMGs are not Compton-thick sources with QSO-like luminosities,and instead suggest that the AGN in SMGs have modest X-ray luminosities (Alexander et al.2005).

Two more recently-pursued routes to search for the pres-ence of an AGN within an SMG are to use near-infrared spectroscopy and mid-infrared imaging.Swinbank et al. (2004)report on near-infrared spectroscopy of24SMG’s from our sample.Of the15SMG’s classi?ed as star-forming from their UV spectra,the emission line proper-ties(line widths and[N ii]/Hα?ux ratios)in the restframe optical support this classi?cation for60%,a further20% have intermediate classi?cations(Hαline widths of500–1000km s?1)and only20%are clear-cut AGN based on their restframe optical spectra.This broadly supports our UV spectral classi?cations and implies that only～30% of SMGs are likely to host luminous AGN.A similar rate of identi?cation of AGN-like SEDs(20–30%)is found in recent Spitzer mid-infrared photometric studies of SMG’s: 2/13from the combined sample of Ivison et al.(2004)and Egami et al.(2004)and2/7for the radio-identi?ed SMGs in Frayer et al.(2004).

We conclude that around20–30%of SMGs show de-tectable signs of AGN activity using a number of inde-pendent indicators.The broad agreement between this proportion based on such a wide-range of indicators gives us con?dence that it represents a true limit to the ex-tent of AGN activity in the population.We stress that this doesn’t indicate that30%of the FIR emission from these galaxies comes from AGN.Even when an obvious and strong AGN is present in an SMG,there are signs that it does not dominate the bolometric emission(e.g., Frayer et al.1998).The AGN contribution to the lumi-nosity of the whole SMG population may be as low as 10%(Alexander et al.2005).Using the results that<～30% of SMGs have properties indicative of an AGN,we will conservatively assume that70%of the FIR luminosity is derived from star formation in the total SMG population. The luminosity densities we estimated earlier can now be translated into SFRDs by scaling them down by30% to account for a maximal AGN-contribution and then us-ing the standard calibration of(1.9±0.3)×109L⊙(M⊙yr?1)?1(Kennicutt1998),we plot the resulting SFRDs in Fig.12.

An additional point has been plotted on Fig.12to repre-sent the～35%of the total SMG population which are not detected in our radio maps.Their redshifts are assumed to follow the radio-undetected fraction from our N(z)–normalized model(Fig.4),implying a signi?cant overlap with the redshift range probed by our radio-SMG sam-ple.The redshift range we show for the radio-undetected sample extends to z=5.1,at which point less than one SMG would lie in a total SMG sample of151galaxies(our targeted98radio-SMGs,plus an extra35%undetected in the radio).This uniform population model is the most in-tuitive representation of the radio-unidenti?ed SMGs,but we stress it is based purely on a model and it is possible that the radio-undetected galaxies have a very di?erent distribution to the one we predict,extending to higher redshifts,or even a bimodal distribution compared to our well characterized

radio-SMGs.

Fig.12.—The evolution of the energy density(parametrized by SFRD)in the Universe with epoch.Our new submm mea-surements(large squares,shown at the median value for each redshift bin)are compared to the published estimates from optical/UV surveys and radio/IR tracers of the star forma-tion density.The open square indicates the SMGs without radio identi?cation,at the median redshift derived from our modeling of Fig.4.The smaller symbols for the optical es-timates indicate dust-corrected estimates.A Gaussian?t is shown for the four submm galaxy points,tracing an evolu-tion comparable to luminous radio-selected Quasars(Shaver et al.1998).For the submm sources,the smaller points show a simple redshift-independent correction to the luminosity den-sity to match the submm extragalactic background down to 1mJy.The dashed line is the best?t for a simple parametric model constrained by the counts of sources in the FIR/submm and the spectrum of the extragalactic background(Blain et al. 2002).Other UV,mid-IR and radio derived points are from Giavalisco et al.(2003–highest-z circles),Steidel et al.(1999–z=3–4triangles),Connolly et al.(1997–z=1–2stars),Yan et al.(1999–z=1.3hexagon),Flores et al.(1999–z=0.3–1 circles),Yun,Reddy&Condon(2001–low-z solid circle). We can now estimate the evolution of the SFRD for all SMGs brighter than～5mJy at850μm.A solid curve is plotted in Fig.12,representing a Gaussian?t to the four SMG points(after redistributing the objects in the high redshift bins into two non-overlapping bins in red-shift).The?t is SF RD=1.26×exp[?(z?2.18)2/σ2], withσ=1.30.This evolutionary behaviour is quite simi-lar to that inferred for the luminosity density of Quasars (e.g.,Boyle et al.2000;see Fig.5).

We also plot in Fig.12the SFRD estimated from a number of UV-selected galaxy surveys at z=1–6(Con-nolly et al.1997;Steidel et al.1999;Giavalisco et al. 2003),and low redshift(z<1)radio and mid-IR obser-vations(Yun,Reddy&Condon2001;Flores et al.1999).

A dust-correction of a factor of5×for the z～3LBG population(Pettini et al.2001)has been applied to the high-redshift UV-selected populations to give their dust-corrected estimates.Recent analysis of the SFRD evolu-

20

tion via C II?λ1335.7in damped Lyαabsorbers suggests a total SFRD comparable to the dust-corrected UV esti-mates(Wolfe et al.2003a,b).

This analysis allows us to bring together the various high redshift populations.Fig.12highlights the rela-tive importance of di?erent classes of high-redshift star-forming galaxy,critical for a full understanding their rel-ative importance galaxy evolution.We see that although the bright,radio-detected SMGs presented here represent only20%of the850-μm background,the estimated star-formation density at z=2–3is within a factor of two of that inferred from restframe UV observations(Steidel et al.1999;Adelberger&Steidel2000).Including a contri-bution from the more numerous,less-luminous SMGs with 850-μm?uxes below～5mJy would result in their SFRD matching or even exceeding that seen in the UV.Moreover, as we argued in the previous section,the SFRD contribu-tion of this population is e?ectively missed by UV-selected surveys,and hence we must sum the bright SMG and UV estimates to determine the total SFRD.

We also need to account for the large fraction of the submm galaxy population which is below our～5mJy ?ux limit at850μm.We chose only to integrate down to 1mJy as this is the?ux scale where the estimated SFR of typical UV-selected galaxies become comparable to the FIR sources.There is a correction factor of～2.9×which is required for the submm points to account for the SMGs integrated down to1mJy,comprising～60%of the ex-tragalactic background in the submm waveband(Smail et al.2002).We apply this correction to Fig.12assuming that these fainter SMGs share the same redshift distribu-tion to the brighter SMGs whose distribution is explicitly measured in this paper.The SFRD measurements cor-rected in this manner would suggest that SMGs are the dominant site of star formation activity in the Universe at z～2–3.However,the discussion below suggests that our assumption of a similarity between the redshift distribu-tions of bright and faint SMGs is likely to fail,even by?ux densities of S850～1mJy(see also Lacey et al.2004).

A second important point to draw from Fig.12concerns the relative evolution of the UV-and submm-selected pop-ulations.Our redshifts show that SMGs are coeval and en-ergetically comparable in a volume-average sense with the population of UV-bright,star-forming galaxies detected at z～2–3(Steidel et al.1999).However,the SMGs with S850μm>5mJy and UV-selected galaxies(which have a median850-μm?ux of<～0.5mJy,Adelberger&Steidel 2000;Reddy&Steidel2004)clearly don’t evolve in the same manner.SMGs appear to evolve more strongly than the UV-selected population out to z～2(C03)and in-deed seem to behave in a manner very similar to lumi-nous Quasars and X-ray selected AGN,whose luminos-ity density peaks at z～2.3(Boyle et al.2000;Croom et al.2004;Silvermann et al.2004).As Fig.12makes clear,this is in stark contrast with the individually-less luminous,UV-selected galaxies whose comoving luminos-ity density is approximately constant out to at least z～5 (Lehnert et al.2003;Giavalisco et al.2004).This sug-gests that the properties of the bright submm population is more closely linked with the formation and evolution of the galaxies or galactic halos which host QSOs,than the more typical,modestly star-forming galaxies identi?ed from their restframe UV emission.Nevertheless,there is likely to be an intermediate luminosity regime where the UV-and submm-selected populations overlap signi?cantly, and therefore where their evolution becomes similar–we propose that this likely occurs at sub-mJy levels.

4.5.Contributions to the FIR background

Using our the long-wavelength SEDs which we?tted to the submm/radio observations of the SMGs we can calcu-late their contribution to the extragalactic background at other wavelengths in the far-infrared.The measured FIR background(FIRB),and the contribution per unit wave-length from our SMGs with redshifts and well-?tted SEDs are shown in Fig.13.At850μm this is simply the sum of the?ux measurements for our SMGs normalized by the ef-fective survey area(assuming a radio-detected fraction of 65%).At all shorter wavelengths,the curve represents the sum of the best-?t SEDs,examples of which were shown in Fig.8.We see that the SMG sample we are studying contributes around20%of the background at>600μm and a diminishing fraction at shorter wavelengths.

To provide a simple baseline comparison we also illus-trate the contributions to the FIRB if all of the SMG’s lie at z=2.3(the median redshift of our model-corrected spectroscopic sample)and have SEDs matching that of Arp220,again normalized to our total850μm?ux.Note that the actual peak of the background due to the SMGs is signi?cantly broadened by the dispersion in their redshifts and dust temperature,compared to a single T d～45K-like Arp220SED at z=2.3.

The contribution of SMGs to the peak of the FIRB varies on a galaxy by galaxy basis by a factor of～10. Galaxies with hotter characteristic dust temperatures con-tribute more,but because they typically lie at higher red-shifts in our submm-selected sample,their contribution is scaled down.Splitting our sample into high and low redshift bins,we assess the relative contribution to the emission around the peak of the FIRB at～200μm as a function of redshift.SMGs at z=2.8±0.3contribute three times less to the FIRB than SMGs at z=2.2±0.3. This implies that the increase in characteristic T d with redshift is not fast enough to counteract the diminuation of restframe200μm?ux from the K-correction.If we in-clude plausibly-identi?ed SMGs without radio detections (Ivison et al.2005),which are predicted to have colder dust temperatures,we?nd that their contribution to the FIRB is much smaller(by a factor～5)compared with the warmer T d radio-SMGs.Moreover,the radio-unidenti?ed SMGs are spread throughout both redshift bins,they do not substantially a?ect the relative FIRB contributions as a function of redshift.

We can also estimate the contribution of the bulk of the submm galaxy population to the FIRB by applying a cor-rection factor(×2.9)to our composite SMG template to include galaxies down to1mJy,comprising～60%of the Fixsen et al.(1998)850μm background.This extrapo-lation suggests that galaxies selected at850-μm are sig-ni?cant contributors(responsible for>～30%of the total emission)to the FIRB at wavelengths of>～400μm,assum-ing that modest extrapolations up the luminosity function are composed of galaxies similar in characteristics to those SMGs in our sample.However,the>1mJy SMG popula-tion probably contribute no more than6%of the emission at the peak of the FIRB at～200μm,which is dominated

战神的挑战2——图文流程攻略

战神的挑战2——图?流程攻略 【游戏封?】 游戏名称：P u z z l e Q u e s t2 中?名称：战神的挑战2 游戏发?：N a m c o N e t w o r k s A m e r i c a,I n c. 游戏制作：I n?n i t e I n t e r a c t i v e 游戏语种：英? 游戏类型：R P G??扮演 发??期：2010年8?13?

%{p a g e-b r e a k|游戏介绍|p a g e-b r e a k}% 野蛮?的技能表(括号中学得该技能所需的等级)： 连捶(1)：战场上有多少红宝?，则造成1/2数量的伤害值，需要4的红魔法。 部落的标记(2)：消掉战场上所有的红宝?，每4颗红宝?增加1点头?伤害奖励，该效果?直持续。 狂怒(3)：在战场上随机位置?成14颗红宝?。 部落的守护(4)：消掉战场上所有的绿宝?，没颗绿宝?增加2点防御，效果?直持续。 猛击(5)：减少敌?每种颜?的魔法值各10。 头?粉碎者(6)：消掉所有头?，敌?损失的回合数为1+1/5消掉的头?数。 野蛮咆哮(7)：敌?防御减少75%，持续3回合，本回合未结束。 破坏狂(8)：消掉所有的绿宝?，增加相等数量的红魔法。 反冲(10)：使?后该回合内通过武器攻击造成的伤害增加50%。 战争狂嚎(12)：随机位置增加3颗红?头?，如果当前红魔法值不少于25则本回合未结束。 最后的反击(15)：使?后每损失25点?命值增加1点头?伤害，该技能

践踏(20)：选择?颗宝?，消掉周边3*3区域的宝?并获得所有这些宝?的效果，对敌?造成8的伤害。 残暴(25)：在后6回合内使头?伤害加倍。 猎头者(30)：消掉最上2?的宝?，获得它们的效果。对敌?造成10的伤害。 嗜?(35)：接下来的6个回合，对敌?造成伤害值的25%转化为??的?命值。 毁灭者(40)：增加50点武器攻击点数，本回合未结束。 迷幻之怒(50)：选择?种宝?，战场上所有该颜?的宝?转变为头?，红魔法不少于25则本回合未结束。 战?的?些分析：消宝?增加该颜?魔法，六个?格为装备的武器，拳头形状表?武器攻击点数，消掉战场中的拳头增加该点数，满?武器使?需求后点击武器?格可以释放，图中我?3、3的数值分别标?了?前的头?伤害加成值和防御值，前者增加消掉头?后对敌?的伤害值，后者标?有该数值%的机会使任何敌?对我?造成的伤害减半。左下?为现在装备的五个技能，注意战?中只能装备五个所以需要在战?前在技能选项中先选择好最有利于战?的技能。 野蛮?战?的?些?得： 消4个或以上的宝?能奖励?个回合，单回合连续消掉?批宝?会有更多奖励(?如稀有武器)。 消宝?时注意看看消去后是否会给对?造成机会，如果没有什么好的

《战神3》图文流程攻略Chapter 11 -奥林

《战神3》图文流程攻略Chapter 11 -奥林匹亚要塞 跟随赫尔墨斯的脚步，抵达木桥上，桥会被迎空飞来的火球砸断，并且背后会不断燃烧，被火追上便会坠崖而亡，中途桥会更改一起方向，不用搭理摇杆向上推一路按×注意面前的平民会阻挡你的道路，用攻击键杀死他们而不是圆圈，跨上平台，平台上有许多平民，抓住处决可以回血，他们的生杀大权就掌握在你的手中。 一路上连跑带揍也没能追上神使，不过最终这个罗嗦的家伙还是被逼在一个锁头上，去左边触发剧情，不用被赫尔墨斯所干扰，掌握自己的节奏一路上利用跳跃点抵达安全区域，上面拼命、小恶魔、狗仔龙蛇混杂，推荐使用太阳神的头颅△蓄力镇住他们，然后L ?清理。

接着向前走，赫尔墨斯把正门关了没办法，我们只能绕远路了，从门左边的楼梯左右开工跃上房顶，发现点赫尔墨斯已经跑到很远的地方等奎托斯了，无奈ing到圆台下方右侧拉动摇杆，腾出缺口跳上平台，拉动巨型投石车，利用QTE跟随巨石抵达赫尔墨斯所在的雅典娜巨像上，酷毙了～ 跟随赫尔墨斯的血迹到达与赫尔墨斯决战的平台，这家伙的速度非常快，天天跑路可不是 盖的，也许一开始会适应不来，用L ?来攻击命中率会比较大，或者直接重攻击的收尾重击的伤害非常大看准时机可以给予赫尔墨斯很大的伤害，途中会有QTE对抗，无外乎是正转或者是反转半圈摇杆，几下重击周后赫尔墨斯被击倒在地，显然他累了，这家伙临死之前还唧唧歪歪拿斯巴达人的荣耀来说事。

已经没有力气的赫尔墨斯坐在地上任你鱼肉，按下圈就开始华丽的表演，非常非常的血腥……未成年人绝对禁止观看…以上以及以下的图片纯属欣赏，完事之后获得“赫尔墨斯之靴”这件物品可以让你跑得飞快。

PSP《战神-奥林匹斯之链》图文攻略流程

【首发】PSP战神Ω奥林匹斯之链图文流程攻略! https://www.doczj.com/doc/8310426443.html,-自由自在游戏社区奉献 《战神Ω奥林匹斯之链》（God of War:Chains of Olympus）是以PS2上大畅销的《战神》（God of War）系列为题材的PSP原创新作，本作由 Ready at Dawn研发，充分发挥PSP所具备的 3D处理性能，呈现出不亚于PS2版壮阔的游 戏场景以及细腻角色。此外，在日前SCEA旗下 工作室Ready at Dawn高级制作人Eric Koch也 通过官方博客PlayStation Blog对外确认，自 2007年起就备受注目的PSP游戏《战神Ω奥 林匹斯之链》已经顺利宣告完工，游戏已经交厂 压碟准备上市，现在这款PSP超级大作将于2008 年3月4日在北美地区发售。 游戏承袭相同的故事背景与游戏类型，玩家将再 度扮演斯巴达战士奎托斯，继续施展“浑沌双刃” 等致命武器，面对全新的难关、敌人与奥林匹斯 诸神的试炼，体验与不同于本传的原创故事剧 情。游戏承袭PS2系列作的特色，在PSP上 重现了广受欢迎的电影运镜式精采游戏画面与 刺激的动作战斗系统，并加入全新的招式、关卡、 机关与怪物，以及以希腊神话为基础的全新剧 情。玩家将扮演斯巴达战士奎托斯，从人间的雅 典一直到冥王哈帝斯的冥府之门，一路朝向地狱 的深渊迈进，体验希腊神话中黑暗且残暴的世 界，并对抗各种传说中的怪物以及强大的头目角 色。 游戏名称：战神Ω奥林匹斯之链/战神Ω奥林帕斯之链 英文名称：God of War:Chains of Olympus 代理发行：SCEA/Ready At Dawn Studios LLC.游戏类型：ACT-Action Game（动作游戏） 制作厂商：SCEA/Ready At Dawn Studios LLC.对应主机：Play Station Portable（PSP） 语言版本：英文（美版）发行日期：2008年03月04日 载体容量：UMD×1参考价格：$:39.99美元 介绍网站点击进入

《战神的挑战2》图文流程攻略

《战神的挑战2》图文流程攻略 事情还没有结束，在更深的地下有人知道Laurella母亲的情况，打开通往下面的门离开地下墓穴的世界我们进入了地精实验室。 这扇巨大的金门在其他3个地方上锁了，分别拉下这3个地方的闸门。

石器时代的食人魔，有铁头功技能：造成20伤害并震晕一回合，附加当前对方蓝魔法数值的伤害。

来自希腊神话故事的人身牛头怪守卫第一道闸门的开关，战斗时你的武器会被替换成一件“斗牛士的披风”(需要8行动点数)，它的作用是防止被神牛冲倒。神牛的冲刺技能：造成10的伤害并击倒对方5个回合。另一个技能Gore：造成5的伤害，如果对方处于击倒状态则附加25点的伤害。 [pagesplitxx][pagetitle][/pagetitle] 这只枭守卫着南方闸门的开关，没有什么特殊的能力。

西方闸门开关由著名的美杜莎驻守，除了石化技能之外，它的毒武器造成14*8(回合)的伤害并给自己增加10的防御。其他生命和防御都较低，不难对付。拉下第三个开关，任务更新：逃离地精实验室。

看起来暴乱已经激怒了堕落的精灵，精灵boss和它的两个手下以及一个巨大的钢铁侠一起出现在这里。一番嘘寒问暖之后站在最前面的精灵族战士手持双刀前来切磋，解决掉开胃菜之后，对上身后那位女性战法师。她有一个Mirror Shield技能：将对方的防御值全部吸到自己身上，持续10+1/8自身蓝魔法的回合，而她初始防御就有47，所以使用该技能后基本上防御值超过100，注意使用减防技能。另一个Dark Blast技能如果紫色魔法多的话伤害也会很高，而且只消耗3的红色魔法。

战神的挑战1代全捕捉攻略

Arboreth 死亡腾蔓 第1列的骷髅头向右移7次 第7列的紫星向左移6次 第3列的红宝石向下移 第6列的绿宝石向下移 Arkliche 大巫妖 第3列的绿宝石向右移 第6列的绿宝石向左移 第2列的紫星向右移 第7列的紫星向左移 第2列的骷髅头向上移 第7列的骷髅头向上移 Catapult 攻城机 第2列的黄宝石向左移 第7列的绿宝石向右移 第4列的蓝宝石向右移 第4列的紫星向右移 第3、6列的闪光骷髅头向上移 Centaur 半人马 第3列的蓝宝石向右移 第6列的蓝宝石向左移 第2列的红宝石向右移 第7列的红宝石向左移 第4列的骷髅头向左移 第5列的骷髅头向右移 Dark Dwarf 黑暗矮子 第8列的绿宝石向上移 第6列最上方的骷髅头向右移 第3列最上方的绿宝石向右移 第1列的绿宝石向上移 第3列的红宝石向上移 第6列的绿宝石向左移 第4列的红宝石向上移 第6列的金币向下移 Doom Knight 末日骑士 第２、４、６、８行最上方的骷髅头向下移 第１（最上方）、３（最下方）、５（最下方）、７（最下方）行的骷髅头向下移第３、７行的骷髅头向上移

Dragon Spider 龙蜘蛛 第２列的骷髅头向右移 第７列的骷髅头向左移 第４列最上方的黄宝石向上移 第５列最上方的黄宝石向上移 第４列最下方的红宝石向左移 第５列的红宝石向右移 Elven Guard 精灵守卫 第２列最下方的黄宝石向右移 第７列最下方的黄宝石向左移 第２列最下方的绿宝石向右移 第２列的绿宝石向右移 第２列的骷髅头向左移 第７列最上方的绿宝石向下移 第７列的骷髅头向右移 Fire Elemental 火元素 第５列最上方的骷髅头向下移 第３列最上方的骷髅头向下移 第６列最上方的骷髅头向下移 Fire Giant 火巨人 第８列的金币向左移 第２列的闪光骷髅头向右移 第５列的骷髅头向下移 第１列的金币向右移 第５列的红宝石向右移 Flame Dragon 火焰龙 第２、４列最下方的骷髅头向左移 第２列的骷髅头向下移２次 第４列的黄宝石向左移 第４列最下方的红宝石向上移 第６、８列最下方的骷髅头向左移 第６列的骷髅头向下移 第８列的骷髅头向左移 第６列的骷髅头向下移 第８列的黄宝石向左移 Frost Dragon 霜龙 第３列上方第２个骷髅头向右移 第６列最上方的骷髅头向左移

《战神3》图文流程攻略

《战神3》图文流程攻略 听完奎爷讲述一下自己在战神1以及战神2之后，玩家跟随奎托斯坠入冥河，顺着河流游，途中会被怨灵吸去生命/魔法以及武器能力，不用挣扎，这是剧情需要，上岸之后失去神力的奎爷疲倦不堪，拖着疲惫的步伐向前走吧，向前不用走多远便可以触发剧情，宙斯的姐姐-智慧女神雅 典娜，并且给失去光芒的混沌双刃注入了新的能力，变成了“流亡双刃”，拿起它，踏上冥界的征途。 向右走，通过跳跃点到达对面远处的平台，不要忘了下方有重要物品-两个红魂箱子以及一个装有“美杜莎之眼”的宝箱，拿完之后就可以上去继续征途了，一路向前解决一波杂兵，适应一下实力大不如前的奎托斯，抵达第三个存档点，此时有两条路，左边一条是通往蕴有四个红魂箱子的平台，右边则是继续流程。

喜欢搞基的-冥王哈迪斯在和你对话了，你明白不明白他话中的话呢……几句话之后获得新魔法“斯巴达军团”(怎么来的?有点莫名其妙～)接下来出现的一些敌人是给你练魔法的;面前有“地狱藤蔓”暂时还开不了，通过右方的锁链一路到对面，途中会有一些杂兵干扰，从锁链上下来之后补充一下生命以及魔法，继续赶路，跳下低处，别忘了背后有两个红魂箱子，然后向前推开大门，首先是一大堆被折磨的灵魂再被美杜莎虐待，上前解决掉两只美杜莎后大门便会打开，往左继续流程。 [pagesplitxx][pagetitle][/pagetitle]

一路向前到达“受折磨的皮尔埃托斯”处，开始一个小解谜，以皮尔埃托斯头上的平台为跳板，进入到里面的房间将“易燃的荆棘”推到对面地狱犬牢笼处，再回头跳上皮尔埃托斯头前方的平台，然后到达“易燃的荆棘”上方飞上上方的平台，触发机关放出地狱犬，几下攻击之后，触发QTE，骑上地狱犬自由的奔驰吧，清理掉路上的杂兵，驾驭地狱犬靠近皮尔埃托斯喷火，让他脱离苦海，随后处决胯下的地狱犬，去获取“阿波罗之弓”。 (使用方法是：按住L2锁定敌人，右摇杆切换目标，?射箭，按住?可以蓄力，带有穿透效果，杀伤力很可观。) 用阿波罗之弓燃烧掉场景中部上方的藤蔓，然后将“易燃的荆棘”推到如上图所示的位置，站在上面即可飞进地图中上方场景内奖励一个红魂箱子以及一个“凤凰羽毛”(增加魔法最大值物品)

[PSP]《战神》详细完美图文全攻略及心得

[PSP]《战神》详细完美图文全攻略及心得 游戏介绍 《战神Ω奥林匹斯之链》（God of War : Chains of Olympus）是以 PS2 上大畅销的《战神》（God of War）系列为题材的PSP 原创新作，本作由 Ready at Dawn 研发，充分发挥 PSP 所具备

的 3D 处理性能，呈现出不亚于 PS2 版壮阔的游戏场景以及细腻角色。 游戏承袭相同的故事背景与游戏类型，玩家将再度扮演斯巴达战士奎托斯，继续施展“浑沌双刃”等致命武器，面对全新的难关、敌人与奥林匹斯诸神的试炼，体验与不同于本传的原创故事剧情。游戏承袭 PS2 系列作的特色，在 PSP 上重现了广受欢迎的电影运镜式精采游戏画面与刺激的动作战斗系统，并加入全新的招式、关卡、机关与怪物，以及以希腊神话为基础的全新剧情。玩家将扮演斯巴达战士奎托斯，从人间的雅典一直到冥王哈帝斯的冥府之门，一路朝向地狱的深渊迈进，体验希腊神话中黑暗且残暴的世界，并对抗各种传说中的怪物以及强大的头目角色。 战神故事背景 人类不断制造战火，加上战神阿瑞斯在凡间不断的挑动纷争，使凡间陷入一片混乱。于是战神的香火日渐鼎盛，势力也一天比一天强大，终于招致了众神的不满和嫉妒。众神的矛盾不断加剧，阿瑞斯更开始了向别的神公开宣战。 斯巴达战士奎托斯，是斯巴达的军队统领，他体格过人，英勇善战，但为人残酷，贪婪，嗜战。某一年，他率领大军和北方的日尔曼人决战，尽管他使出浑身解数，斯巴达战士视死如归，

仍然无法挽回己方的败局，残酷的战斗中，斯巴达军渐渐处于下风，越来越多人被杀，领袖奎托斯更被日尔曼首领逼到绝境。 在敌人的刀快要刺下来，奎托斯生死的瞬间，英勇的斯巴达人向天怒吼：“阿瑞斯！！！！我乞求您的帮助，让我打败我的宿敌吧！！当我达成心愿，我的灵魂将属于您！！”话音刚落，一道闪电从天而降，击中奎托斯。一副带锁链的利刃紧紧地缠在了他的手上，日尔曼首领被奇迹惊得呆住了。奎托斯乘他不备，挥动了手上的双刀，那硕大的躯体在一瞬间就被砍得四分五裂，只剩下那头颅圆睁着充满困惑和不解的眼睛。阿瑞斯接受了斯巴达人的交易，战场上日尔曼战士都被无形的力量宰杀，好一场血腥的屠戮！战斗过后，阿瑞斯告诉斯巴达人，这刻起，你就是我的仆人，灵魂永远属于我，必须为我永世效劳，手上那不能取下的双刀就是契约的证明。 从此，奎托斯整天陷入无休止的战争中，不断地屠杀着，征服着。直到一天，他把战火烧到了一个村落，他肆无忌惮地杀戮着，见人就剁，还吩咐部下烧村。神庙旁的老妇极力地劝说他放过庙里的人，那是神圣的地方啊！可是奎托斯已经毫无人性了，他闯进庙中，见人就杀，不久，人们都倒在血泊中。杀红了眼的奎托斯却惊讶地发觉妻儿已经死在他眼前，而凶手正是他本人！这一刻，后悔，懊恼，愤恨同时涌上他的心头，人性和良知渐渐回到他的脑中。他认为这是阿瑞斯的陷阱，是让他彻底地为战神

战神1和2全文完整攻略

战神1图文攻略 冒险开始于停放在爱琴海的一个港口的一艘船上，首先，你会遇到一群怪物，干掉它们，然后打开地板上的舱盖进入船内。附近有绿魂，前面的路被一堆桑物挡着，用手中的刀砍掉后继续往前走，一直遇到一个九头蛇怪。它在攻击你前，会抖动它的耳朵，注意防御。在它死前，它上方会出现“圆环”标志，接近按“O”键发动“死忘之舞”，干掉它后，继续前行来到船的甲板上。甲板上，有一群会飞的怪物在攻击人群，干掉它们，（你可以杀死这些人而得到绿魂）。 之后，遇到一个九头蛇怪，站在它的“右前方”快速连击，小心防御。如果不小心被它抓住了，赶紧按“O”键挣脱。

干掉它后，打开你右边的一扇门得到一些“红魂”，然后跳进甲板上的洞口，往前游一段距离后，顺着一张网往上爬来到甲板的另一部分。在横梁中央有一个分贫口（往右边走能得到些“红魂”）。 在甲板的另一头的平台上有一群弓箭手，你需要在其下面放一个木箱子，才能上到那平台上，注意弓箭手可以毁坏你的箱子。当上了平台干掉他们后来到另一个甲板。 之后，你需要攀登一个梯子和一张网同时要干掉些敌人，当你到达桅杆的顶部时，你可以通过那个横梁往走得到些“红魂”之后顺着绳子荡到另一艘船上。

破坏掉三个木板墙出现一条通道（会得到一个“红魂”和一个“蛇发女怪”）顺着路进入船内，开始过场动画。动画过后，继续前进，在角落处有一些蓝魂生命绿魂。顺着一张很长的网来到甲板上，遇到九头蛇怪家族，经过一场战斗，干掉它们后，（详细战斗请读者自行体会），爬进最大的那只的喉咙里去“船长的关键”.返回甲板，跳上板条箱来到右侧的平台上，（在平台上得到蛇眼）顺着绳子荡到之前的那个甲板上。然后爬到之前你来过的平台上，穿过门口，进入船长门。（用“船长的关健”） 一开始有一个“可爱”的迷你游戏，你可以得到些“红魂”。 之后，爬上船另一尽头的梯子来到甲板，右边有和条通往船坞的通道，上了船坞往左走，干掉些敌人后来到一升降平台，跳过其右边的堵破墙后，再跳过一个

PSP战神斯巴达之魂图文流程攻略

PSP战神斯巴达之魂图文流程攻略 剧情交代完后,开始可操控,第一战是在船上,这很容易让人想起家用机上的一代.不过这次的敌人是人鱼. 消灭掉一批敌人后,看到远处的友船被章鱼腕搅碎.接着进船舱.继续虐人鱼.此时,可以使用 L+O键使出一招冲向敌人然后将敌人扑倒扔出去的招数.出船舱后会看到一只章鱼腕来敲你,注意用翻滚(L+R+摇杆)躲避.攻击挡路的2个章鱼腕.把尽头的一个蛇形章鱼腕砍死后开始剧情. 然后遇到大鱼,大鱼要打个2,3次.这是第一次.它会掌击等,掌击伤害较大.大鱼受到一定攻击后会换个地方出来.反正继续抽它.出现O后,需要用边上的钩子把它吊起来,然后就是QTE,之后来到亚特兰蒂斯岛上. 跃入水中,来到第一个记忆点处. 从右侧沿墙边攀岩到一处平台,出现敌人.击破一处墙壁后出现通路.继续前进.利用一个光点做秋千点来到一个平台,平台上有一个牛头人,新手可能比较难对付.注意用翻滚保持距离. 此时大鱼又来骚扰你,砍断它的一条腕子吧.之后对付牛头人.接着拉动机关.上方的栅栏打开. 攀岩上去.经过一条羊肠小道后来到一处墙壁前,需要WALLCLIMB. 爬到顶上后来到一处流水的平台,这里有敌人在猎杀平民,貌似是"斧头帮"的. 就算是斧头帮,奎爷也找抽. 之后开门,用石头压在一处开关上,在瀑布处利用光点跳跃即可继续前进. 经过一阵攀爬,来到thanatos之庙的DEATHGATE.而普通人是无法穿越这里的,需要用KERES的头骨才行.于是奎爷调头找其他路去了.此处有个记忆点. 继续前进,进入一个房间后,拉下机关,圆形平台开始上升,敌人开始掉下来. 经过水道后来到亚特兰蒂斯城.进城前,先享受下滑水,注意按X键 来到一个大平台上,出现独眼巨人.攻击力,速度都不错,注意回避,别硬磕. 来到一间有雕像的地方,按住O一段时间后蓄力踢开雕像,可以潜水通过. 在到达大平台前的桥梁会倒塌,注意不要掉下去死掉.这里有个记忆点. 接下来是波塞冬神殿. 进入神殿后,奎爷看到了儿时的情景 之后看到了自己的母亲.继续剧情 在母亲说出奎爷的父亲是谁后,就变成了一只怪物.最后奎爷将其击倒,母亲也恢复正常,并获得解放. 之后得到callisto''sarmlet. 紧接着得到魔法:亚特兰蒂斯之眼.可以放出直线型光线攻击. 之后就被大鱼带走.来到海底火山附近.此处有记忆点. 通过爬墙壁来继续前进,途中有会吐火球的怪,爬墙时机动性不好,注意下. 经过2个光点跳跃来到caldera.这里的圆形平台上有火鸟,按O用投技比较好解决.

战神 斯巴达之魂 图像攻略

游戏名称 战神Ω斯巴达之魂 游戏原名 God of War: Ghost of Sparta 游戏制作 SCEA 游戏发行 Sony Computer Entertainment. 游戏平台 PSP 游戏类型 ACT 游戏人数 1人 游戏发售 2010.11.05 游戏载体 UMD/下载 游戏版本 欧版

操作设定 方向.↑魔法-北风神号角 方向.←魔法-厄里尼厄斯降罪 方向.→魔法-亚特兰蒂斯之眼 方向.↓斯巴达武装/雅典娜之刃切换 摇杆控制奎爷移动 方块轻攻击 三角重攻击 圆圈投技 叉子跳跃.可二段 R按住发动希拉之祸 L防御.敌人攻击前防御可发动弹反start游戏菜单.查看装备道具.升级装备selec t 系统菜单

片头剧情过后.船甲板上和雅典娜神像对话后.进入教学战斗.就是基本熟悉下操作.进入船舱后 出口边上打破门有红魂宝箱.

出门后会遇到海怪斯库拉的触角.和其身上类似寄生虫一样的小怪物.首先要攻击缠绕在甲板上的两个大触手.这里需要注意的是第二个触手抬起后如果不迅速通过的话.触手会继续缠绕船体来挡路.之后QTE搞定带嘴的触手后进入BOSS战 战神斯巴达之魂详细图文攻略已完结 2010年11月09日15:20腾讯游戏我要评论(9) 字号：T|T

和系列一样.战神的第一个BOSS战都是气势十足的.海怪斯库拉.也是很有气势的.不过这东西总感觉和PSP上一代战神第一个BOSS类似.斯库拉战斗分三个阶段.第一阶段.会用触手三连击.可以防御.有时会用爪子重击地面.不可防御.只要注意回避重击即可.猛打头部后其会从场景另一边出现.这次会用嘴进行攻击.还会放小寄生虫.搞定后再次从初始位置出现.注意的是这次出现是附带一次大范围攻击判定的.搞定后QTE结束.

战神奥林匹斯之链攻略

战神奥林匹斯之链攻略 战神：奥林匹斯之链》（God of War : Chains of Olympus）是以PS2 上大畅销的《战神》（God of War）系列为题材的PSP 原创新作，本作由Ready at Dawn 研发，充分发挥PSP 所具备的3D 处理性能，呈现出不亚于PS2 版壮阔的游戏场景以及细腻角色。此外，在日前SCEA旗下工作室Ready at Dawn高级制作人Eric Koch也通过官方博客PlayStation Blog对外确认，自2007年起就备受注目的PSP游戏《战神Ω 奥林匹斯之链》已经顺利宣告完工，游戏已经交厂压碟准备上市，现在这款PSP超级大作将于2008年3月4日在北美地区发售。 游戏承袭相同的故事背景与游戏类型，玩家将再度扮演斯巴达战士奎托斯，继续施展“浑沌双刃”等致命武器，面对全新的难关、敌人与奥林匹斯诸神的试炼，体验与不同于本传的原创故事剧情。游戏承袭PS2 系列作的特色，在PSP 上重现了广受欢迎的电影运镜式精采游戏画面与刺激的动作战斗系统，并加入全新的招式、关卡、机关与怪物，以及以希腊神话为基础的全新剧情。玩家将扮演斯巴达战士奎托斯，从人间的雅典一直到冥王哈帝斯的冥府之门，一路朝向地狱的深渊迈进，体验希腊神话中黑暗且残暴的世界，并对抗各种传说中的怪物以及强大的头目角色。 游戏开始之前还是让大家熟悉下游戏的操作： □：轻攻击，武器升级后可长按，配合L键打出不同的招式。 △：重攻击，同样，武器升级后可长按或配合L键打出不同的招式。 ○：抓，使用物品，开箱子，开门等等。 ×：跳，可二段跳。 L：防御，可以配合攻击键使用。 R：配合△可施放魔法。 类比摇杆：移动，同时按住L，R，可以进行回避动作，回避动作中暂时无敌。

战神1__完美图文攻略

冒险开始于停放在爱琴海的一个港口的一艘船上，首先，你会遇到一群怪物，干掉它们，然后打开地板上的舱盖进入船内。附近有回复剂，前面的路被一堆桑物挡着，用手中的刀砍掉后继续往前走，一直遇到一个九头蛇怪。它在攻击你前，会抖动它的耳朵，注意防御。在它死前，它上方会出现“圆环”标志，接近按“抓”键发动“死忘之舞”，干掉它后，继续前行来到船的甲板上。甲板上，有一群会飞的怪物在攻击人群，干掉它们，（你可以杀死这些人而得到回复剂）。 之后，遇到一个九头蛇怪，站在它的“右前方”快速连击，小心防御。如果不小心被它抓住了，赶紧按“投”键挣脱。 干掉它后，打开你右边的一扇门得到一些“红球”，然后跳进甲板上的洞口，往前游一段距离后，顺着一张网往上爬来到甲板的另一部分。在横梁中央有一个分贫口（往右边走能得到些“红球”）。

在甲板的另一头的平台上有一群弓箭手，你需要在其下面放一个木箱子，才能上到那平台上，注意弓箭手可以毁坏你手帕箱子。当上了平台干掉他们后来到另一个甲板。 之后，你需要攀登一个梯子和一张网同时要干掉些敌人，当你到达桅杆的顶部时，你可以通过那个横梁往走得到些“红球”之后顺着强子荡到另一艘船上。 破坏掉三个木板墙出现一条通道（会得到一个“红球”和一个”Gorgon Eye）顺着路进入船内，开始过场动画。动画过后，继续前进，在角落处有一些魔法生命回复剂。顺着一张很长的网来到甲板上，遇到九头蛇怪家族，经过一场战斗，干掉它们后，（详细战斗请读者自行体会），爬进最大的那只的喉咙里生进“Captain’s Key ”.返回甲板，跳上板条箱来到右侧的平台上，（在平台上得到Gorgon Eye）顺着绳子荡到之前的那个甲楹上。然后爬到之前你来过的平台上，穿过门口，进入Captain 门。（用“Captain Key”）

《战神3》图文流程攻略

《战神3》图文流程攻略 看来奎托斯也对刚刚那一役动魄不已，回到火神之处对着回身大吼“你想害死我不成!”火神慌忙解释，奎托斯还未等火神解释完一把将“石脐”甩给火神让火神铸造武器。 片刻之后“复仇凶鞭”诞生，在火神将其交给奎托斯的同时，火神处于对女儿的疼爱不惜豁出性命对奎托斯痛下杀手但战神岂能让他这么容易得手，一阵QTE打斗之后尖锐的铸造台贯穿了火神的腹部，火神在临死之前还不忘祈求奎托斯放过他的女儿，获得武器“复仇凶鞭” 接着老规矩出现一批敌人让玩家熟悉新武器，“复仇凶鞭”的空中追击能力很强，熟悉完了之后去右边存个档然后飞上平台跨过空间门，回到阿布罗狄忒的寝宫，这回又可以再搞一次，这个淫荡的女人自己的男人都死了还有心情在这里和她的杀夫仇人逍遥快活…此时房间内的物品可以再次拿包括红魂箱子以及牛角。

拿完房间内的东西后向下走继续流程存个档抬起大门，出去之后习得魔法“复仇女神盛怒”出来一堆杂兵此时进入魔法无限状态，不挺得使用魔法吧，不过感觉这个魔法远不够带劲的，继续向前活到“上层庭院”，在之前坏掉的控制杆处用“复仇凶鞭”攻击坛子给其充能，即可联接道路。 图1 [pagesplitxx][pagetitle][/pagetitle] 图2

图3 图4 图5 向右行走飞过沟壑，补充完生命后抬起大门向前迈进，跟随镜头熟悉房间构造之后便开始搜刮工作，将弩炮调整并且射击如图1所示; 攀爬绳索下来获得“美杜莎之眼”，接着回到地图右下角搬动拉杆将平台放下(如图2)，将弩炮调整向左射击，顺着绳索向左一直爬到达如图所示位置在边缘处利用飞行绕过铁栏杆拔住最高的绳索一路向左下来可以获得一个牛角和红魂箱子(如图3、4); 然后先到右下角搬拉杆将平台生气，回到弩炮位置将弩炮位置摆正中央先射一箭将铁门射开

PSP 《战神斯巴达幽灵》详细流程攻略

《战神斯巴达幽灵》详细流程攻略 流程开始(XD选择的HARD难度)： 此流程并不包含全宝箱的收集，只是重点讨论强制战斗与BOSS战的打法以及解密。关于全宝箱的收集，巴士速攻会在后期为各位玩家献上。 ★强制战斗：难度★☆☆☆☆ New game，看玩剧情后开始第一场强制战斗。敌人全是鱼人，全部用圆圈键投技一次搞定.几乎没有难度。 ★流程：

之后开门进入船舱。里面有一些敌人。这里系统会提示L+圆圈的组合攻击。可以将一些敌人扑倒然后进行追加攻击。船舱对面有一个被障碍物堵住的宝箱。然后上楼梯来到船舱外。 ★强制战斗：蜘蛛怪若干，触手若干= =难度★☆☆☆☆ 刚到船舱外面会出现一些蜘蛛怪，左边有一个大触手，时不时会进行攻击。路中间还有一个挡路的触手。蜘蛛怪用投技就可以搞定，左边的触手攻击时需要一点预判然后紧急回避。我们可以不用管他们，直接打中间挡路的触手，把他打下海。然后继续前进，攻击第二个触手。这个触手不会被打下海，但是攻击到一定程度它会抬起来，然后赶快通过就可以了。 最后就剩下一只触手和蜘蛛怪了。蜘蛛怪依旧轻松解决，触手的攻击也能够用L键轻松防御。攻击方面直接跳起来砍他，不久即可打出QTE。 ★BOSS战：难度★☆☆☆☆ BOSS攻击方式： a.触手三连击：用触手攻击。前两下是用触手敲地面，最后一下是用触手横扫地面。前两击防御不能，最后一击可以防御。 b.拍地：BOSS用手掌拍击地面。防御不能。 c.吐蜘蛛怪：吐出一些蜘蛛怪。此招只会在BOSS出现在左边的时候才会使用。 d.咬击：用嘴向前咬，防御可能。此招只会在BOSS出现在左边的时候才会使用。

e.甩头攻击：甩动头部的大范围攻击，防御可能。此招只会在BOSS出现在左边的时候才会使用。 对于a.b攻击最好是看准时机紧急回避，a攻击的最后一击可以防御。当BOSS没有攻击的时候赶快跑上去跳起来用到砍，不久可以把他打跑。之后BOSS会在场地左边出现。 BOSS一出来就会使用c，蜘蛛怪基本没什么威胁，投技解决之。之后他还会随机使用d 和e。果断防御，没什么威胁。在左边打了BOSS几次之后，它会再次回到右边。再次重复第一阶段。直到把BOSS打到QTE位置。可不要按错了哦XD~ ★流程： BOSS被虐之后逃跑。我们沿着石头跳到上面。利用宝箱上面的绳索滑倒下面。落入水中，潜入水中可以发现一个宝箱，然后上岸存档，继续前进。 沿着墙壁爬到对面，一直向上来到一个平台。玩儿过试玩版的朋友都知道，这里会得到一个魔法。但是正式版里却不是这样。只有几个敌人，多用投技或者L+圆圈的扑倒攻击，把前面的那堵墙弄垮。接着继续前进。用勾索点跳到对面，遇到牛头怪。 ★强制战斗：牛头怪*1;触手2只;鱼人*3 难度★☆☆☆☆ 刚进入战斗不久，奎爷就被两只触手缠住。不断连打L和R可以帅气地解决两只触手。三只鱼人就用投技虐之，不过还是要注意一下牛头的攻击，最好离他远一点。最后和牛头单挑就简单了。打出牛头的QTE，用到捅爆它的头~

战神2完美图文攻略

作者：David203 游戏名称战神2 游戏发行SCEA 游戏平台PS2 游戏价格49.99美元 游戏类型ACT 游戏版本美版 游戏人数 1 游戏发售2007.3.13 游戏载体DVD9 点击进入 游戏官网 Chapter 1

消灭初期杂兵→开门到走廊→巨像打破墙壁→爬上梯子到平台处→BOSS战 BOSS战：等其手臂砸下后猛攻，3次后他会闭上眼睛，之后跳上左侧平台，按R1放出石块，再把奎托斯弹到BOSS身上，按提示出入4次□即可

□键潜水（右侧木门可以打坏，里面是成人小游戏）→R1撞开栅栏→爬上梯子存档→拉动旁边机关跳到对面→消灭3杂兵继续前进→连按○抵挡巨像的脚→爬上左侧墙壁（左侧有红魂箱×2）→与巨像第2战。 巨像第2战：先跳到右侧平台，等其右手拍下后跳起攻击右眼（随便攻击手臂也可以），受创3次后他会趴在码头一侧，过去按○触发QTE系统，反复两次搞定。

前进到存档点右侧房间→拉出石台→拉到开关上按R1+×将石台踢到第2个开关处→拉动石台到右侧通道尽头→打坏石台上雕像→跳上取得两箱红魂和盖亚之壶→再拉回石台压住开关继续前进。 拉起升降平台→攻击巨像露出的眼睛→剧情后爬上右侧墙壁→消灭杂兵穿过独木桥（L1+□对付成群的杂兵效果明显）（图4中位置跳到对面可取得两箱红魂）→存档后沿铁链滑到下方→消灭敌人拉开铁门→巨像将桥砸断→与巨像第3战。 巨像第3战：把雷剑拔出要分3步：1、把他的断手仍到他身上（将其击晕后拔剑红魂被吸光）2、到他身前按L2发动魔法猛攻（击晕后拔剑蓝魂被吸光）3、到他身前跳起猛砍（击晕后拔剑HP减半）。之后拔出雷剑按L1+△攻击巨像，击晕后按○进入其体内。 沿路前进，到中间破坏第1个动力源；之后爬上右边的网继续上行，到尽头处跳起抓住绳索到达对面，再走过独木桥到达第2个动力源，这里要先用L1+□攻击一次才能破坏；继续上行到第3个动力源（贴着围墙移动时按R1可以下降、再按×键上升），攻击两次后再按R1破坏（从后边的网<第6图>爬上可取得美杜沙之眼）；继续上行，爬上梯子拉动机关降下钟摆，攻击钟摆使其晃动，再上去拉回机关，钟摆就会挡住光线，抓住绳索爬到对面，攻击3次后，按R1破坏最后的动力源，之后从巨像口中跳出。

战神1攻略

战神1攻略 走了不远就要先来一场小boss战. 水怪(我猜应该是神话里面的leviathan吧= =)攻击力不强,可以上去硬砍.当它的血砍到一定程度后,画面就提示你输入相应的按键来干掉它. 这个类似于小游戏的设定在游戏会经常出现,各位要多多注意了 打完水怪后一直前进,途中要注意得小心翼翼得走平衡木.如果不幸掉下去的话可以立即按X 来抓住木头就可以了. 来到船长室,由于没有门的钥匙,所以还不能打开. 路上还要打一次水怪,对手实力不强,上前拼就可以了. 先继续前进,来到一个很多弓箭手的地方.接下来要进行推箱游戏,目的就是把箱子推到弓箭手的下方,以便使得自己能够跳上去. 这里有点难度,因为箱子被弓箭射中大概4次后就会损坏.要多练习几次才可以. 一路向前,来到学习第一个技能的地方. 获得POSEIDONS RAGE,按L2就可以施放,提高等级后还会有其他追加技能. 过后利用绳索来到boss战的地方, 路上记得要拿GORGON EYE,拿够六个GORGON EYE就可以提高体力上限 BOSS 三头水怪,两边是小的,中间是大的. 左右两边有箱子给你跳上去,中间可以爬绳子上去.先打两边的,打到一定程度后它会倒下. 立即利用箱子跳上去,用钉钉住就可以了. 然后是大的那只,也是把它的血打到一定程度,画面会提示进行按键(连打圆圈)的小游戏. 每次按键成功都会将水怪的头撞向木杆,三次成功后,木杆就会折断,再p它就可以打死了击败后从它的嘴里进去假救人(其实是杀人= =),拿到船长室的钥匙. 从右边的木箱上去,利用绳索回到船长室.途中可以拿到第二个GORGON EYE 开门后就结束序章了 其实游戏现在才正式开始,一段精美的CG后... The Gate Of Athens 从船里出来一直走, 先别拉开关,走另外一边拿PHEONIX FEA THER,这个拿够6个就可以提高魔力上限的. 从船里出来一直走, 先别拉开关,走另外一边拿PHEONIX FEA THER,这个拿够6个就可以提高魔力上限的. 对付它可以等到画面提示按键时连打圆圈来搞定,而且这样也容易得到绿魂 上来后又有按键的小游戏杀怪,大家借此可以多熟悉一下. 有些宝箱可以是可以选择补血或者补气的,主要看你在它是什么颜色时打开它. 之后进去一个很多柱子的地方, 有些是木箱在最底下的,先把底下是木箱的打掉就可以跳上去了. 留意靠近下方处有一个GORGON EYE和大量的红魂,如图

战神1详细图文功略

SCE超人气作品《战神》完美图文攻略 01 Aegean Sea 冒险开始于停放在爱琴海的一个港口的一艘船上，首先，你会遇到一群怪物，干掉它们，然后打开地板上的舱盖进入船内。附近有回复剂，前面的路被一堆桑物挡着，用手中的刀砍掉后继续往前走，一直遇到一个九头蛇怪。它在攻击你前，会抖动它的耳朵，注意防御。在它死前，它上方会出现“圆环”标志，接近按“抓”键发动“死忘之舞”，干掉它后，继续前行来到船的甲板上。甲板上，有一群会飞的怪物在攻击人群，干掉它们，（你可以杀死这些人而得到回复剂）。 之后，遇到一个九头蛇怪，站在它的“右前方”快速连击，小心防御。如果不小心被它抓住了，赶紧按“投”键挣脱。 干掉它后，打开你右边的一扇门得到一些“红球”，然后跳进甲板上的洞口，往前游一段距离后，顺着一张网往上爬来到甲板的另一部分。在横梁中央有一个分贫口（往右边走能得到些“红球”）。

在甲板的另一头的平台上有一群弓箭手，你需要在其下面放一个木箱子，才能上到那平台上，注意弓箭手可以毁坏你手帕箱子。当上了平台干掉他们后来到另一个甲板。 之后，你需要攀登一个梯子和一张网同时要干掉些敌人，当你到达桅杆的顶部时，你可以通过那个横梁往走得到些“红球”之后顺着强子荡到另一艘船上。 破坏掉三个木板墙出现一条通道（会得到一个“红球”和一个”【Gorgon Eye 1】）顺着路进入船内，开始过场动画。动画过后，继续前进，在角落处有一些魔法生命回复剂。顺着一张很长的网来到甲板上，遇到九头蛇怪家族，经过一场战斗，干掉它们后，（详细战斗请读者自行体会），爬进最大的那只的喉咙里生进“Captain’s Key ”.返回甲板，跳上板条箱来到右侧的平台上，（在平台上得到【Gorgon Eye 2】）顺着绳子荡到之前的那个甲楹上。然后爬到之前你来过的平台上，穿过门口，进入Captain 门。（用“Captain Key”）

psp游戏-战神：奥林匹斯之链(图文全攻略)

图文流程攻略： 作者：Levelup - Fnix 操作说明 □ ：轻攻击，武器升级后可长按，配合L键打出不同的招式。 △：重攻击，同样，武器升级后可长按或配合L键打出不同的招式。 ○：抓，使用物品，开箱子，开门等等。 X ：跳，可二段跳。 L ：防御，可以配合攻击键使用。 R ：配合△可施放魔法。 类比摇杆：移动，同时按住L，R，可以进行回避动作，回避动作中暂时无敌。 SELECT：系统菜单。 START ：武器，能力升级的界面。

界面和之前发布的DEMO版一模一样，奎秃的头像。有点毛骨悚然。 鉴于这次战神的流程很短，所以小生选择了HARD难度进行游戏。 开头的CG动画，比起PS2上前辈们，算是很大程度上的缩水了。 THE SHORES OF ATTICA 开始游戏后，是和DEMO里一样开场，杀掉一些小兵后，推动角落里的弩炮，摧毁敌人的战船。

之后，地面被砸开一个大洞，跳下去，杀掉小兵。这里有一个绿色的箱子，是补血的，先别急着开，等下对付小BOSS的时候没血了再开。 清完小怪来到正前方的门面前，连按O开门。 突然门被一只独眼巨人砸开了。

发生拼刀场面，连点O推开独眼巨人。 紧接着出现一直硕大的怪兽，将独眼巨人活吞了。 这个怪兽就是我们在这款游戏中第一个要面临的小BOSS。他被用来让玩家联系回避操作。 攻击模式很简单，头微微缩进去就是要咬人了，口中就火光就是要喷火球了。及时回避是这里要学习的。 将它打到1/4血的时候出现QTE，按提示操作，解决掉它，最后举起独眼巨人留下的木锤给它最后一击。虽然没能杀掉它，不过也打伤了它一只眼睛。 从破开的门出来，出现CG动画，刚才的怪兽爬到城里去作乱了。

战神1完美图文攻略教学提纲

战神1完美图文攻略

冒险开始于停放在爱琴海的一个港口的一艘船上，首先，你会遇到一群怪物，干掉它们，然后打开地板上的舱盖进入船内。附近有回复剂，前面的路被一堆桑物挡着，用手中的刀砍掉后继续往前走，一直遇到一个九头蛇怪。它在攻击你前，会抖动它的耳朵，注意防御。在它死前，它上方会出现“圆环”标志，接近按“抓”键发动“死忘之舞”，干掉它后，继续前行来到船的甲板上。甲板上，有一群会飞的怪物在攻击人群，干掉它们，（你可以杀死这些人而得到回复剂）。 之后，遇到一个九头蛇怪，站在它的“右前方”快速连击，小心防御。如果不小心被它抓住了，赶紧按“投”键挣脱。 干掉它后，打开你右边的一扇门得到一些“红球”，然后跳进甲板上的洞口，往前游一段距离后，顺着一张网往上爬来到甲板的另一部分。在横梁中央有一个分贫口（往右边走能得到些“红球”）。

在甲板的另一头的平台上有一群弓箭手，你需要在其下面放一个木箱子，才能上到那平台上，注意弓箭手可以毁坏你手帕箱子。当上了平台干掉他们后来到另一个甲板。 之后，你需要攀登一个梯子和一张网同时要干掉些敌人，当你到达桅杆的顶部时，你可以通过那个横梁往走得到些“红球”之后顺着强子荡到另一艘船上。 破坏掉三个木板墙出现一条通道（会得到一个“红球”和一个”Gorgon Eye）顺着路进入船内，开始过场动画。动画过后，继续前进，在角落处有一些魔法生命回复剂。顺着一张很长的网来到甲板上，遇到九头蛇怪家族，经过一场战斗，干掉它们后，（详细战斗请读者自行体会），爬进最大的那只的喉咙里生进“Captain’s Key ”.返回甲板，跳上板条箱来到右侧的平台上，（在平台上得到Gorgon Eye）顺着绳子荡到之前的那个甲楹上。然后爬到之前你来过的平台上，穿过门口，进入Captain 门。（用“Captain Key”）

《战神3》图文流程攻略

《战神3》图文流程攻略 到达迷宫之巅之后发现潘多拉并无大碍，利用跳跃点攀上奥林匹斯之链，一直向上爬，通过螺旋阶梯去圣火之间，转动摇杆一共一周将迷宫拉到奥林匹斯之巅，此时圣火与潘多拉都已经准备就绪，可是奎托斯在潘多拉的身上仿佛看到了自己女儿的身影，不愿意要潘多拉去献身，这是战神系列的总BOSS“众神之神-宙斯”出现了。 第一阶段 BOSS的攻击方式①连续拳 ②QTE对抗 ③小型闪电火 流程：终于要和宙斯打了，第一阶段在一个房梁上展开，这样看起来很像一款格斗游戏，宙斯的每一招的杀伤力都很大切忌不要贪刀①连续拳如未打中奎托斯则是两下，打中则会出最后一

下闪电火伤害非常的高，QTE对抗只需要转摇杆即可以将宙斯反扔回去，宙斯在空中的小型闪电火可以用弹反返回攻击，火拼数回合之后，宙斯头上出现QTE标志， 第二阶段 BOSS的攻击方式①连续拳 ②QTE对抗 ③小型闪电火 ④闪电火撞击 流程：全程推荐使用拳套，和第一阶段一样，如果你弹反的技术够硬可以顶着宙斯打，他攻击你就弹他，如果不慎被连击，第三下记得用魔法抵消要不非常的痛，④闪电火撞击速度很快，加上宙斯又会闪烁经常不知道从哪里就冲下来了有点郁闷，没有什么很好的办法，和第一阶段一样打到宙斯头上出现QTE提示。

接着便会出现剧情，潘多拉为圣火献身，奎托斯打开魔盒之后发现里面是空的，在宙斯的嘲笑之下奎托斯变得怒不可遏，朝上一路走飞到圆型台上便可以和宙斯开始战斗，象征性的打就可以了，过了一段时间之后盖亚会出现添乱，剧情过后再次回到盖亚体内，抵达盖亚心脏使用“海格力斯之力”打掉晶体。 [pagesplitxx][pagetitle][/pagetitle] 然后攻击盖亚的心脏，之后宙斯便会再次出现，最终决战就此展开，这个阶段的宙斯有些烦但我有个很无赖的方法，首先武器选用的是拳套，在打宙斯以及弹反宙斯的时候的同时就站在盖亚心脏的旁边以L1 ?为主这样既可以打到宙斯还可以打飞宙斯的分身还可以通过盖亚的心脏获得生命值回复，然后一些具体比如宙斯吸血之类就自己随机应变吧~我实在是太困了这个BOSS不愿去研究怎么无伤了，也没想到T T。 最后利用豪快的QTE解决宙斯以及盖亚，万物都已泯灭，一切化为虚无，奎托斯的结局如果我就不剧透了，交给各位玩家自己去感觉吧;《战神》的确是一部让人值得敬佩的游戏作品，奎托斯永远活在我们玩家的心中，多玩《战神3》完美图文流程攻略至此--结束!(以下图片纯欣赏。)