Documente Academic
Documente Profesional
Documente Cultură
L?MNOMEPNRQ
SDT@UWVXYU[Z]\[T_^R`=T@acbc\Wd6ZeT@U[Ze^fahgZ8XYd6Z]af\Wd\Wd6T6i_djU[^:kml_nKaf\Wl_g[^cT@UZW^#l@k=o_d6T@np^jq4`=Xraf\sah\tdu\tdPvrwel@kyxfg[^fa
l_UWd;z{Z8djafd|b}afl_n6~8ah\Wd\8gWT@Udjo_d_=#gWn:dPo_d6^:ncd6^fw[l_UZaflnpT_Z8X
T@afXrl_Uli_djn#T;i_djnconcd6^fafncX
b}afd6ZnpT@UW_dRl@k
H//W kmncl_iXrl_vYdja:ahlncd|Z
_ UW( _ UW _# __# O
¡
¢gWn:djo_d|^:T@ncd £y¤¦¥§W=¨©¥«ª¬=t[ªt£®¤ q8vrd|T_Z¦XYUWahlaf\WdKZ8d}¯[UWXrahXrl_Ul@kah\td*T@_UWXrahgZ8d_
°²±´³#µ8¶¸· vYl_8¹»º¼[gW½R¾¿bpl_U^faÀT@U8a ¶
ÁÂ\Wd*b}l_U[^facT@U8aÂX
^Â^fdja#^fl;af\[T@aÂah\td*ÃWncXr_\8ahd6^faÂ^faÀT@np^ÂT@ncdT@Ã[l_gWaÂÄjdjncl;T@_UWXrafg[Z8d_Åaf\Wd#k«T@XrU8ahd6^fa=^facT@np^`Âd
bPT@U^fdjdÆT@ncdT@Ã[l_gWa#Ç@ah\(T@_UWXrahg[Z¦dql_n¢TÈk«T_b}afl_nÉT@Ãl_gWa µ_· k«T@XYU8afdPn6
Û ±ÝÜ4Þ
í U*af\WX×^4lÎZ¦gWvYd®`Âd^f\[T@vrv_vYll_V¢XrUKafgWncU*T@a/i@T@ncXYl_g[^wW\[T_^fd6^l@kWah\tX
^T_^fahncl_UWl_X×bjT@vncdji_l_vrgWahXrl_U/af\Wd
l_Ã[^fdjnciT@ahXrl_U[T@vahd6bc\tUWX
ÚÌgWd6^jq_af\WdÂwW\8oÎ^fX×bjT@vwWncl8bpd6^c^fd6^XrU8i_l_vYi_d6Z4q@T@U[ZKaf\WdÂUWdj`ÓT_^fafncl_UWl_X
bjT@vZ8X×^cb}li_djncXrd|^
ah\[T@ancd6^fgWvrahd6Z4ÁF\Wd;ncd6b}l_udjU[Z¦d|Zîahdj½ahÃll_VeX×^ï[ðñí»ò;z ^fafncl_UWl_;o(ÑÁF\Wduð®i_l_vYiXYUtHç:UWXri_djnp^fd~
«ä@af\Ód6Z8XrahXrl_U/Å:óÓXrvYdjo jØ_Ø ¡ ÝSl_ncdZ8djacT@XYvrd6Zô^fgW__d6^fahXrl_U[^Èkml_nl_wWahXrl_U[T@v:Ã[T_bcV_ncl_gWU[Zencd6T_Z8XrUWe`=XrvYv
_XYi_djUT@a=ncdjvYdjiT@U8aÂwWv×T_b}d6^ÂXrUÍaf\WdUWl_ahd6^jqWT@U[Z(T@a=af\WddjU[Zl@kaf\Wdul8Z8gWvrd_
µ
ßncl@k=ñ/T
`=ncdjU[b}dÍÁÂ\tdÍì8bpXrdjU8ahX ¯èbìa»ovYd ¡ XYU8afnclÎZ8gbpd6Z]af\Wd;wWncl_wdjncahXrd6^Ol@kÂaf\WdjncÒT@v:ml_n /ª
£¬FÖ ¡
npT_Z8X
T@afXrl_UÒT@aafdPwdjnpT@ahgWncdOÁRq_kml_nÂ`:\WX
bË\ah\Wd#djUWdjnc_ouZ¦X
^fafncXYÃtgWahXrl_U;`:XYaf\kmncd6Ú_gtdPUbpo Þ X×^_XYi_djUÃ8o;af\Wd
ª
£®¤W¨ :«
'& µ)(
¹
Þ %$
* Þ
é
Þ !#"]³ +
¡ ± µ_Ü kml_n¢^fT@vYv Þ
, -
`=\Wdjncd Þ ¡ _Xri_d6^:ah\WdT@l_gWU8a=l@kÂdPUtdPnc_o(ÃdPXrUWdjXrahafd6ZXrUDgWUWXra#npT@UW_dKl@kkmncd6Ú_gtdPUbpo_q/T@UZ ( X×^
. 0/21)354)176
1)896;::<48=1?>@:A>,1)B%CED:,1)B,3FD:,8HGCEDJIA1)BA>32KML ON
bPT@vrvrd|Z l_vYafÄjT@UWU/~¸^Ob}l_U[^faÀT@U8aj [ l_gs^f\Wl_gtv
Z
\Wl`Âdji_djnVUWl@`ôaf\[T@aaf\WdjncÒT@vnpT_Z8X×T@ahXrl_U;wd6T@VÎ^T@a`=T6i_djvrdjUW_ah\QP
RSUTV W Y
X
Ô ±Óµ Ø__ ò Ô [Z
ó XYdjU/~¸^ñ/T
`
T@U[Zeaf\[T@aKah\tdahl_acT@vÂdjUWdjnc_oeX
^wWncl_wl_ncafXYl_U[T@v®ahl J\0X ]. [Z
ìafdpk«T@UtÏ l_vrafÄPT@UWUñ/T
` DÁÂ\Wddj½wT@U[^fXYl_U l@k
^`_
ßv×T@U[bcV~ ^Fv
T
` ^f\Wl@`¢^Fah\T@a Þ
X #
^ Ë
b [
\ @
T p
n _
T }
b h
a j
d c
n ×
X f
^ h
a ×
X #
b @
l kah\td*vrl`:ϧkmncd|ÚÌgWdjU[b}oD§vYl_UW_Ï`=T
i_djvYdjUW_af\ ¡ z¸aÀT@Xrv«~
ba
l@kÉaf\WdjncT@vnpT_Z¦X
T@afXYl_Uq`=\tX
bË\¿b}gWaÀ^Èl ´T@ÃWncgtwWahvroÑT@a\WXr_\skmncd6Ú_gtdPUbpXrd6^jq`:\Wdjncdah\tdÍwW\tl_ahl_UsdjUWdjnc_o
dP½8b}dPd6ZW^:af\Wda»owWX×bjT@v4ah\tdPncT@v4djUWdjnc_o (
ÁÂ\Wd]dPUtdPnc_oÓZ¦X
^fafncXYÃtgWahXrl_UÓl@k;T_^fafncl_UWl_X
bjT@v#l_ÃÎxfd6b}aÀ^(vrXrV_d _¤ X
^Cl_UWvroYL%c)cdB<1befCg>,L)8h:AKjiýT
ÃWv
T_bËVÃlÎZ8o(bpgtnci_dJ:
k dji_djncah\Wdjvrd6^c^jqWXra=X×^¢T_llÎZ(ncgWvrdOl@kaf\8gW;Ãafl^cT6oCah\[T@a:ah\tdPncdX×^¢Tb}vrl^fd*ncdjv×T@ahXrl_U
Ã[dja»`ÂdjdjUah\Wd#afdjuwdjnpT@ahgtncd=l@kyTRÃ[l8Z8ouT@U[Zaf\Wd=`=T
i_djÃ[T@U[ZÈ`=\Wdjncd=XradPXraÀ^®l^fal@k/XraÀ^®djUWdjnc_oÁF\8g[^jq
^faÀT@np^\[T
i_dH^fgtnfk«T_bpd(afdPwdjnpT@ahgWncd6^Èl@kÆTîkmdP`?af\Wl_g[^cT@U[Zôò T@U[ZôT@vYv:djuXra;l^fal@k*ah\tdPXrndjUWdjnc_o¿XrU
l
%W
ah\Wd :¥«t¥ «]8Ê¢ªt8¤ :¨ ¡ æ:l`Âdji_djn6q8^fl_gWnpb}d|^T@ai_dPnco\WXr_\ul_nÂi_djncovYl@`ôahdjwdPnpT@afgWncd6^
`Âl_gWv
ZÈUWl_aÃ[d:wWX×bcV_d6ZKl_gWakmncl_ël_Ã[^fdjnciT@ahXrl_U[^XrUuaf\Wd:iX×^fXYÃWvrdÂncdj_XYl_Uyð®½ÎT@wWvrd6^/l@k4af\WX
^^fdjal@kè^fl_gWnpb}d6^
T@ncdnmpo
djncoÑ\tl_a;T@ahafdjnu^fg[bË\¿T_^KT_^^fahncd6T@^XrUãT]bpvrl^fdÃWXrU[T@ncos^facT@n6 r
mpoP q ¶ Õ UW Ht¢ÏnpT
oÎ^ ¡ q j)s òqwd|T@VXYUtÑT@a
Á \Wdl^fa/Xruwl_ncacT@UÎa/b}l_ncncd6b}ahXrl_Uahl#af\WdÂÃ[T_^fX×bX
Z¦d|T:l@k[^fll_af\Kah\tdPncT@v_djX
^c^fXrl_UOX×^_XYi_djUÃÎoRah\WdU[T@ncÏ
Â
ncl`kmd6T@afgWncd6^af\[T@aÂT@ncX
^fdFkmncl_,afnpT@U[^fXYafXrl_U[^ÃdPa `ÂdPdjUÍZ¦X
^cb}ncdjahd:djUWdjnc_o;vrdPi_djv×^XrUT@ahl_^T@U[Zl_vYd6b}gWvrd|^j
ÁÂ\Wd6^fd;bjT@UeÃ[d /¨ý¥t¥«£f¥f qWX kÂah\WdZ8l_XYUT@UÎa#^fl_gWnpb}dKl@kàdjX×^c^fXYl_U(X
^Tb}vrl_g[Zl@k=Z8X èg[^fdKT_^jq ja
l_n £®¤8ÊW¥£ëh¥f X{k`ÂdO\[T
i_dO^fl_udjaf\WXrUWKvYXrV_dOaf\WdjncT@v[djX
^c^fXrl_U^f\WXrUWXYUt*af\Wncl_gW_\ÍTb}vYl_g[ZÒl@k
}|
T_^j Wl_ndj½ÎT@wWvrd_q_ah\td=djX
^c^fXrl_U*l@k4af\WdOìgtUub}l_U8aÀT@XrU[^T@UÎoKT@Ã[^fl_ncwWafXrl_UKvYXrUWd6^l@`=XrUW#ahlah\td=af\WdjncÒT@v
vYXr_\8akmncl_XYac^ÂT@XrUH^fgtnfk«T_bpdRw[T_^c^fXYUt*af\Wncl_gW_\(XYac^Âl_gtahdjn¢T@afl^fwW\Wdjncd_
ìw[d6b}afnpT@v_vYXrUWd6^/T@ncdÂT@nc_g[T@ÃWvro#ah\Wdl^fa/Xrw[l_ncacT@U8a4wWXYd6b}d6^4l@kWXrU8kml_ncT@ahXrl_U*XrUKT_^fahncl_UWl_o_/ÁÂ\Wdjo
_XYi_d#Xruwl_ncacT@UÎaXrU8kml_ncT@ahXrl_UuT@Ãl_gWabË\WdjX
bjT@v[b}l_w[l^fXrafXYl_U/qÌahdjw[djnpT@afgWncd_q_wWncd6^c^fgWncd_q@i_djvYl8bpXra»o(×kmncl_
ah\Wdâ £Ê:Ê¢4¤eW=¥ f =~ , P P
Ô ± Ô ¡
Õ
mm
Gamma rays X-rays UV Infrared Radio
waves
-15 -12
Wavelength: 10 m 10 m 1 nm 1µ m 1 mm 1m
9.5 6.5
T (peak energy): 10 K 10 K 3000 K 3K non-thermal
Interstellar
Galaxy Supernova Cool Hydrogen
Sources: Stars
clusters remnants gas/dust
Active
Binaries galaxies
Thermal
Emission Cosmic-ray Synchrotron
Free-free Atomic lines
processes: interactions radiation
Molecular lines
mm
Gamma rays X-rays UV Infrared Radio
waves
Visible light
ÁÂ\WX×^Z¦X
T@_npT@ë^fgWT@ncXYÄjd6^ah\Wd:npT@UW_d=l@k4`=T
i_dPvrdjUW_af\[^T_bjb}d|^c^fXrÃWvrdÂahlKaflÎZtT6o~ ^T_^fahncl_Utl_udjnp^jq
ahl__djaf\Wdjn
m
`=Xrah\D^fl_d#l@kah\WdK^fl_gWnpb}d6^ÂT@U[ZCwWnclÎb}d6^c^fd6^Âaf\[T@aÂT6owWncl8Z8g[b}dOaf\Wd*npT_Z¦X
T@afXYl_UC`ÂdOl_Ã[^fdjnci_d_
m 7P
ó T
i_djvYdjUW_af\ T@U[Zkmncd6ÚÌgWdjU[b}o Þ T@ncdRncdPv×T@afd|ZCÃ8o Þ;± 7P y o
T@wWwWncl@½XrT@ahdjvro Õ Ñj Ó^ é ¡
¹
)
m ß\Wl_afl_U(dPUtdPnc_o ÜÞ X×^=T@wtwWncl½XYT@afdjvYo µ_· dPvrd6b}ahncl_Ui_l_vraÀ^Rmd ¡ T@a ± P %W
ßv×T@U[bcVDZ¦X
^fafncXYÃtgWahXrl_U¿ z¸ÃWv
T_bËVDÃlÎZ8o~ ¡
k
móÝí»ð 2Z8X×^fwWv×T_bpdjdjU8av×T6` ¡
RShTV
P
µ Ø__Ô 4q`=\Wdjncd X×^XYUsò?T@U[Z PeX
^XrU W
Ionization
free-free transitions
n=4
n=3
Pα
Pβ
Pγ
etc.
Hα
Hβ
Hγ
etc.
n=2
Paschen series (near infrared)
etc.
Ly α
Ly β
Ly γ
etc.
n=1
ground state o
mW|¢l_afn\Wahdjnpn¢T@T@U[af^fl_XrafXYl_^jUKqÎT@ÃU[djZCa»`Âd6dj^fdPw[Ud6b}a»X×`ÂT@l#vYvrodjUWdjncl__vYo*d6b}vrgWdj& vri_d6dP^jv×q^j\[q Ü4T
i_ÞdO± Ûl_nc dO³ b}l_Ûu wt± vYX×bjjT@Õ ahd6¶ Ç8ZC
dPÔ Ut° dPnc _³ o
vrdjÔfi_ dP v×^j¡ d
m ll_vrdjn=T_^:^fgWncncl_gWU[Z8XrUWO\tl_ahafdjnÉ^facT@& n T@Ã[^fl_ncwWafXYl_UCvYXrUWd6^OH ahnpT@U[^fXrafXYl_U[^ ¡
m W| ncdjdjÏ»^fw[T_b}d=dj½Îb}Xrahd6Zl_n:XYl_UtXYÄjd6ZT_^ djX
^c^fXrl_UvrXYUtd|^R ahnpT@U^fXYafXYl_U^ ¡
íUDTb}ll_v& T_^#l^fa¢T@afl_^OT@ncdXYU]_ncl_gWU[Z(^faÀT@afd_q ± K= k #. µ Ï»b}%npT_Z8XYlvYXrUWdKl@kàæÅ4^fdjd
Ã[djvrl` ¡ æ æ XrUHb}ll_vl_vYd6b}gWv×T@n=bpvrl_g[ZW^Rm_XriXrUWul_ncdbpl_wWvrdj½^fw[d6b}ahncgt ¡
Y[Z]\_^`\*a@^`\cbed[fhgifNjlkm\6nNo]\_jlgi\6k[f@pqf@rtsuj]v;aNkmbw^`fNjxfNqhsyg_a@zlfN{km\|^`r@a@bwsufNjlkbmfhzufNj]vN\_^=d=a
rN\_z}\_j]vNbmZlk_~
K;
]#yK4
]4t sykj]fNbbw^a@jlkma@^`\|j8bla@b/a@zuzNd=a
rN\|zu\_j]vNbmZlk__bwZx\|^`\a@^`\ 4
]Q]
m
Q
=
{l\_bd[\_\_jZ]\6a6ra@{lkmfN^`]bmsufNj{la@jl ]kf@pQ¡#¢@£¤a@jl CfNbmZ]\_^¥gifNjlkmbmsubwo]\_j8b`kc¦§bwZ]\6km\K¨©ªa@{lkmfN^`x«
bwsufNj¬{la@jl ]kca@^`\K^`\6kmfNjlkmsu{]z}\p®fN^SbwZ]\C¯°vN^`\_\_j]Z]fNokm\*\i±4\²gib_³µ´`¶·\6kmb*¨©-kms}bm\6kca@^`\;Z]suvNZ/¸4 8^`qfNo]j8b`a@s}j
bwfNlk_¸¹\N¶ºv]¶D»¼a@o]jla½\6a¾¦§¿tÀNÁNÁ(q´csuj¡a
d=a@sus¶Ã±Ä\6gibwsurN\(^a@j]vN\Tp®fN^KvN^`fNo]jl 8«Å{laNkm\6 D¨e©ÆaNkmbw^`fNjxfNq;
sÇkKÈxÉ¥È_Á@Ê4qW¦§d=subwZ¼vta@lki´Ë¶[»Ìs}zuz}suq\_bw^`\_«Åd=a6rN\#bm\_z}\6k`gifNl\6k_¸xz}suÍN\*bmZ]\KÎ8a@qh\6k¥ÏQzu\|^`Í(»Da@Ðtd[\_zuz/Y\|zu\6k`gifNl\N¸
a@zÇkmfTs}j¡#a6d=a@sus¸8xsÇgÑÍhoxbmZ]\Kkml\6gibw^`oxqÒa@vta@sujCp®^`fNqÔÓ Á@Ê4qÆfNj8d=a@^ ]k_¶
Õ su^gi^a
p®b¦ËÈ_À:ÍqÖS´Äg_a@^`^`ckmqa@z}zu\_^4bw\_zu\6k`gfN\6k{]o]b4vN\_b/qTolg`Zc{\_bwbm\_^bw^a@jlkma@^`\|jg#a@{lf@rN\qhftkmb
@f p/bwZ]\#¡#¢@£×~Ø ÕÙÚÕ syk{]o]suzy 8s}jxv Ù £cÛ¨ Õ ¦ Ù bw^a@bmftkm]Z]\_^`syg:£{km\|^`r@a@bwfN^`Sp®fN^[Ûla@^`«¨Åjp®^a@^`\² Õ kmbw^`fNjxfNq;]´
Éha*À8¶ «Åq,bm\|zu\6k`gifNl\#sujXaSÎo]q;{fcÎ\|b_¶Ü\_^mp®\6gib¦§{]o]b\_Ðt\|jkms}rN\@´kmÍt;bm^a@jlkmla@^`\_jlgiSsÇka
rÝa@s}zya@{]zu\[p®^`fNq
kmlaNgi\²gi^a
p®b_~\¶Þv]¶=¨© ÕÙ ¦Å¨Åj8p®^a@^`\6 Õ kmbw^`fNj]fNqT Ù a@bw\_zuz}subm\Ý´[d=aNkFzÇa@o]jg`Z]\6 Cs}jßÈ_àNáNÓa@jl Colkm\6 aâNÁ@«ãgiq
bw\_zu\²k`gifN\cbmf;qa@bmZ]\cd=ZxfNz}\ckmÍta@b[d=a
rN\_z}\_j]vNbmZlk#È_Á@Ê4q É¥È_ÁNÁ@Ê4q¶Y[ZxsÇksyk[kmbwsuzuzlbmZ]\c{\6kmb=kmo]^`rN\_ha@b
bwZ]\6km\Sz}fNjxv;d=a
rN\|zu\_j]vNbmZlk_¶
_ äÂå¬lQæmç
]Å f@p[èj]\hkmbm^`olgibwox^a@z \|b`a@suzé 8\6gi^`\6aNkm\6k*aNkëêìsujlgi^`\6aNkm\6k_¶hYFZ]sÇksykc{\6g|a@okm\hzusuvNZÚb
vN\|b`k:{]zuo]^`^`\6 haNk:sub[sÇk c
§
íKt/x
tl bmZ]^`fNo]vNZbmZ]\Sbw\_zu\6k`gfN\Ka@l\_^`bmo]^`\N¶Y[Z]\SÍN\_p®fN^`q;o]zyaSsyk_~
©:\6kmfNz}oxbwsufNjDîKa@^g_km\²gifNjl xk:ï¤À Á8¸°ÁNÁNÁñðò¦§d=a6rN\_zu\|jxvNbwZî;a@\_^`bwo]^`\@´
þ ÿTÿ8þ
þ
Y[Z]\^`\_vNs}fNjc{\_bd[\_\_j;¨©¾a@jl ^aN 8suf=d=a
rN\|zu\_j]vNbmZlkd=aNk4fNj]\f@p]bwZx\QzyaNkmb4bwf#{\Q\_Ð]z}fN^`\6 {8*aNkmbw^`fNjxfNqh\_^k_~
¢
^aN 8sÇa@bmsufNjhsykvN\_j]\_^a@z}zu;d[\6a@Í(¦ØS¶ù·#¶8{]zÇaNgÑÍt{fÚ 8lo]Ð 8\_jlkms}beõê ´`¸]a@bwqftkm]Z]\_^`\:sÇk[a@{lkmfN^`{xs}j]v]¸N 8\_«
bw\6gibwsufNjKbm\²gÑZ]j]fNzufNvN#sÇk 8s gio]zub|¸6a@bbwZx\{lfN^ \|^`zusuj]\{l\_bd[\_\_j*fNxbwsyg_a@zta@jl #^aN 8suf=bm\6g`Z]j]synNox\²k_¶!#sÇk`gif@rN\_^`
f@pqa@j8DsujÚbm\_^kmbw\_zuzÇa@^Szus}j]\6kp®^`fNq æÅlç=æÅl sujøbwZ]\¬È_àNâNÁNkS\|jgfNox^a@vN\² ¬gifNjlkmbm^`olgibwsufNj¾f@pckml\6gisÇa@z
bw\_zu\²k`gifN\6kSp®fN^KbmZ]\qq(îÝkmo]{]«Åqq%^`\|vNsufNj/
¶ "submZßbmZ]\ 8syk`gifÝrN\_^`f@p#z}suj]\6kKa@bKfN^K{l\_zufÝdöêÂïÈTqqH¸
bw\_zu\²k`gifN\6k;{\_vta@jÂbwfü{l\ 8\6kms}vNjx\² üd=submZÂrN\|^`üZ]s}vNZkmox^mpaNg\(aNg|gio]^aNgi¾bmf¾d[fN^`ͬ#\ gsu\_j8bwzu¾sujÂbwZ]syk
^`\|vNsufNj/¶
»ftkmb]^`\6gisykm\DsykbmZ]\È «Åq Î]Ï[»DYW¦§ÎÚa@q\6kÏz}\_^`Í»¼a@Ðtd[\_z}zSY/\_zu\6k`gfN\@´hfNj2»Da@o]jlaø½\²a8¸
¡ a
d=a@sus¶#Yfp®fÚgiolk:^aN 8sya@bwsufNj\#gsu\_j8bwzua@bSêï¤ÓNÁNÁ@Ê4q¸lkmo]^mpaNgi\KaNg|gio]^aNgi(f@pQbmZ]\;À%$Ýâ@«Åla@j]\_z4q\|b`a@z
^`\&\²gibmfN^QZaNkbmfK{l\('cÀNÁ@Ê4q¸NfN^Q{\_bwbm\_^[bwZa@jìÈ
îÚÈ_ÁNÁNÁcsujlgÑZKÉ;^`\_qa@^`ÍÝa@{]zu\[\|jxvNs}j]\_\_^`suj]v]¶ Õ ]xz}s}j]vSbmZ]\
a@j]vNo]zya@^^`\6kmfNzuo]bms}fNjSp®fN^`qTo]zÇa:bmf*bmZ]\:Î]Ï黼Yc¸@d[\=km\_\=bmZ]\=\_jla@zubcp®fN^d[fN^`Ítsuj]vca@bzufNj]vN\_^d=a
rN\|zu\_j]vNbmZlk_~
a@bcêïòÓNÁNÁ@Ê4q~ À 8 Á ¸°ÁNÁNÁñðßÁ8¶°ÓîÚÈ ¸ùÁNÁNÁï a@^g_km\6gifNjl ]k
a@bcêïòÓKqq~ À Á8¸°ÁNÁNÁñðßÓîÚÈ ¸°ÁNÁNÁXï Á@a ^g|km\6gifNjl ]k
â
¦`È
´1 æÅ/3 2Qä#Q5 4õ cÅ4
]Å p®^`fNq s}j8bm\_^kmbw\_zuzÇa@^kmfNzusÇ ëla@^`bwsygiz}\6k[¦e¯µ 8olkmb_³ÝfN^:³ºvN^a@sujlk_³ ´Ë¶/Y\_ql\_^a@«
bwo]^`\:^a@j]vN\6kp®^`fNqÔÈ_ÁNÁNÁS½õrN\_^`;gzuftkm\=bmf;kmb`a@^k_¸NbwfÈ_ÁS½õs}jbmZ]\#s}j8bm\_^`s}fN^f@p 8\_jlkm\= 8okmbw«giz}fNo ]k_¸
kmZ]s}\_zy 8\6 Tp®^`fNqÔkmb`a@^=^aN 8sya@bms}fNjbmZla@b#gfNoxzÇ CZ]\6a@b[bwZx\*vN^a@sujlk_¶
¦ÅÀ´6 3 798itÂ/ ]ÅQ p®^`fNq-s}fNjxs;:_\6 vtaNkm«giz}fNol xk=¦e¯ ¡#¨m¨^`\_vNsufNjlk_³ ´{fNq;{la@^ \² ë{8;\_j]\_^`vN\_bwsyg
]Z]fNbmfNjlkp®^`fNq NfNo]j]v]¸NZ]fNbkmbËa@^k_¶Y[Z]syksyka@zykmfcÍtj]f@d=jaNk ä:t/N
8t:æwç:Qú É=<#\_^`qa@jSp®fN^
¯°{]^a@Ís}jxv[^aN 8sÇa@bmsufNj/³}¶Y[Z]syka@^`sÇkm\6kp®^`fNq \|zu\6gibw^`fNjlk{\_s}j]vcaNg_gi\_z}\_^a@bm\² S{8cbmZ]\\_z}\6gibw^`ftkmb`a@bmsÇg¹èl\_zy
f@pj8olgiz}\_s]suj(z}f@d=« 8\_jlkmsub;giz}fNo ]kf@pZ]fNb[sufNj];s :_\6 vtaNk#¦e¯°xzÇaNkmqa8³ ´Ë¶YFZ]\*km\²gibm^`o]q f@p^aN 8sya@bms}fNj
p®^`fNq bmZ]syk]^`fÚgi\6k`kzuftfNÍÚkFrN\_^`X sµ±4\_^`\|j8b¹p®^`fNqÆ{]zyaNg`Í«{fÚ ;^aN 8sya@bwsufNj/s}b[sy> k 4a@bc¦k`a@q? \ lo]Ða@b
a@z}zp®^`\6nNox\|jgsu\6ki´Qo](bmfXaTqa@Ðtsuq;oxq xZ]fNbwfNj(\_j]\_^`vNN¸8d=Z]syg`Z(sÇk:f@pfN^ 8\_^[bmZ]\KbtxsÇg_a@z4bmZ]\_^`qXa@z
\|zu\6gibw^`fNj(\_j]\_^`vNHA¦ @,B¥´Ë¶
¦ÅÓ´DC ]
8t¹ææmmQ/ ]ÅQ ¦®fN^/a@{kmfN^`]bwsufNj4´8{8suj8bw\_^kmbm\_z}zya@^lqfNzu\6goxz}\6k4a@jl Sa@bwfNqk_¶ Õ bwfNqsyg
kms}vNjla@bmo]^`\6k[a@^`\p®fNo]jl (sujbwZ]\KfNxbwsyg_a@z®îFE?G,a@jl (¨©S¶lÃkml\6gisya@z}zusuqhfN^`b`a@jÚb:sÇk tlÂä#wH7
] cÅ4
]Å ¸Nd:Z]\_^`\=bmZ]\ckm\|a@^a@bw\:\|zu\6gibw^`fNjlk¹bmZla@bégifNjlkmbmsubwo]bm\cacxzÇaNkmqaSgfNq\:bwfNvN\_bmZ]\_^
bwfqa@ÍN\ha@jøa@bwfNq ¶ I4fNjxvN\|^`«Åd=a
rN\_z}\_j]vNbmZD\_qsÇk`kmsufNjìgifNq\6k#p®^`fNqWqfNzu\²gio]zu\6k_¸4f@p=d=ZxsÇgÑZñ¡#¢syk
bwZ]\Kqftkmb#suqhfN^`b`a@jÚb_¶*YFZ]\6km\KZla
rN\Kbd[f^`\6aNkmfNjlkFp®fN^*\_qsubwbms}jxv]~gÑZla@j]vN\6k#sujDbmZ]\_s}^c^`fNb`a@bms}fNjla@z
fN^[bwZ]\_su^=rs}{]^a@bmsufNjla@zl\_j]\_^`vNkmb`a@bw\N¶Q©:fNb`a@bwsufNjla@z\_j]\_^`vNhzu\_rN\_zÇk:a@^`\qfN^`\*gizuftkm\_z}kmlaNgi\6 4¸8a@jl
bw\_jl ¼bmf]^`fÚ 8og\Tqs}zuz}suq\_bw^`\K\_qsÇk`kmsufNj/¸d=Zx\|^`\6aNkrtsu{]^a@bms}fNjla@zgÑZla@j]vN\6k*a@^`\aNk`kmfÚgisya@bw\6 ¼d=subwZ
s}j8p®^a@^`\6 xZ]fNbwfNjk_¶
¦§¿8´ ì å /
]J8i K§twL4¥
tt km\_\|jCbwZ]^`fNoxvNZXgizufNol ]k¹f@pgiftfNzsujÚbm\_^kmbw\_zuzÇa@^vN^a@sujlk_¶
Y[Z]\TvN^a@s}jlkSa@^`\NM%ÈiÊ4q sujß 8sya@q\_bw\_^Ka@jl ¬a@{lkmfN^`{DrsÇkmsu{]zu\hzusuvNZÚbq;olgÑZDqfN^`\hkmbm^`fNj]vNzuDbmZla@j
¨© ¦km\_\T{l\_zufÝd´Ë¶;»Ìftkmb*kmb`a@^k_¸ls p[qhfN^`\KbmZla@jÂÈ_ÁNÁzus}vNZ8bm«ÅN\²a@^k#a
d=a6N¸4kmo8±4\_^*kmfNq\;a@{lkmfN^`]bms}fNj
¦®b]syg|a@zuzu;Zla@z paSqa@vNj]submol 8\=\_^*È_ÁNÁNÁSz}suvNZ8bw«ÅN\6a@^ki´`ÝfNbmZ]\_^k[kmsubQsujlkmsy 8\#¯ gif8gffNjlk_³f@p 8olkmb[a@jl
a@^`\*nNo]subm\*suj8rtsykms}{]zu\#s}j(fN]bmsÇg_a@z^aN 8sÇa@bmsufNj{]o]b:\²aNkmsuzuhfN{km\|^`rN\6 Cs}j(bmZ]\רe©*¶¯ Ù \_\_s}jxv]³8fNbwZ]\_^`d:sÇkm\
s}j8rtsykmsu{]z}\ckmb`a@^ksÇk:aTqX+a OmfN^[{l\_j]\ièlb:f@p¹¨e©ªaNkmbw^`fNjxfNq;N¶
Z\[
]_^a`R^UbD^\c_dfebg`R^¾å C IfÝd:«\_j]\_^`vND¦hM,È\iG*´:]Z]fNbmfNjlk:f@p¹¨e©2^aN 8sÇa@bmsufNjHg_a@jD^`\_«
z}\6aNkm\[\_z}\6gibm^`fNjlks}jTbwZ]\=gi^`8kmbËa@z8zya@bwbmsÇgi\:f@pÄkmfNq\ékm\_qsÇgifNjl olgibwfN^k_¸@kmolg`ZKaNk¨Åj Ù {¦®sujl 8suo]q a@jÚbmsuqhfNjxsÇ 8\@´`¶
Y[Z]\6km\p®^`\|\S\_z}\6gibm^`fNjlk:s}jlgi^`\6aNkm\cbmZ]\S\|zu\6gibw^`syg_a@zgifNjl 8olgibms}rs}bef@pQbmZ]\*qa@bm\_^`sÇa@z/a@jl (bmZ]\S^`\6kmo]z}bmsuj]vhgio]^`«
^`\|j8( b lf@d g_a@jÂ{\q\6aNkmo]^`\6 4¶ìYFZ]\6km\ 8\_bm\²gibmfN^k;a@^`\km\_zu\6gbmsurN\FfNj]z}¬]ZxfNbwfNjlkSf@pckm) o gis}\_j8b*\_j]\_^`vN
¦kmZ]fN^`b[\_j]fNoxvNZDê´ég_a@j(^`\|zu\6aNkm\
tQæx
8 ¶
i jlkÝå bgmn%åam C Y[Z]\6km\Da@^`\{laNkm\6 fNj¤¦§\|Ðl\_jlkmsurN\@´ha@^`^a
ÚkTf@ph¦§bt]syg_a@z}zu]´À â¼ð¾À â
8\_bw\6gibwfN^km«ÅfNj]«a@«g`Zxs}/¸]a@ja@z}fNvNfNolk:bmfbmZ]\CÏ=Ïo-a@^`^a6 8\_bm\²gibmfN^k#olkm\6 sujDfNxbwsyg_a@zaNkmbm^`fNj]fNqT¶SY[Z]\6km\
fN{]bËa@suj]\6 TbwZx\[èl^kmb¨© ¯°]sygibwox^`\²k_³utà 8sujÚ{xo]^`vNZ/³ùk/¨©¥Ï Õ »Æd=aNkèl^kmbbmfKvNf*suj8bwfS^`\_vNo]zÇa@^km\_^`rtsygi\N¶¨©a@^`«
¢ ¢
^a68k4o]Sbwf;È_ÁNÀ@¿ ]suÐt\_zÇkj]f@dì\_ÐsÇkmb_¸Na@jl *ÀNÁ@¿tá 8\_bm\²gibmfN^ka@^`\\_Ðt\²gibm\6 ;kmftfN j p Ù Ítr
qY\|zu\6k`gifNl\N¸6Îo]j]\
È_àNà t s ¶=YFZ]\KkmfNz}sy 8«kmbËa@bm\cbm\6g`Z]j]fNzufNvNsyk:qhfN^`\\|ÐfNbwsygSbmZla@jbwZx\ Ù suzusÇgifNjolkm\6 Cs}jfN]bmsyg|a@z¹Ï=u Ï ck_~[gfNq«
qhfNjgifNq;{]sujla@bms}fNjlka@^`\=¨Åj Ù {¦Å¨Åjl s}o]q Õ j8bwsuqfNj]sÇ \Ý´4fN^Q¡:v8Ï[ ]Y\c¦»Ì\|^gio]^` É]ÏéaN 8qsuo]qTÉÚY\|zuzuoÚ^`suo8q(´Ë¶
Y[Z]syk[syk:{l\6g_a@olkm\SbwZ]\S]Z]fNbmfNj\_j]\_^`vNs}\6k:a@^`\*qTolg`Zzuf@d[\|^:bwZa@js}jbmZ]\SfN]bwsyg_a@zxbwZx\*\_j]\_^`vNj]\_\6 8\6 (bmf
\|Ð8gis}bm\;a@j\_zu\6gbm^`fNj(s}j Ù suz}sygifNjsyk[bmftfzya@^`vN\cbwf 8\_bm\6gb#¨© ]Z]fNbmfNjlk_¶
i v 1 ^\wr^Un%`¼å C YFZ]\6km\;a@^`\×aXÍtsujl (f@pQbmZ]\_^`qhfNq\_bm\|^6¸xd=Z]syg`Zd[fN^`ÍC{l\6kmbca@b#bwZ]\KzufNj]vN\_^
¨© d=
a rN\|zu\_j]vNbmZlk_¸éÈ_Á@Ê4qWfNj8d=a@^ ]k_¶(Y[Z]\_Dq\6aNkmo]^`\bmZ]\sujlgi^`\²aNkm\Ts}jø\_z}\6gibm^`sÇg_a@z[gox^`^`\|j8bK 8o]\bmfDbmZ]\
$
Z]\6a@bwsuj]vS\±4\6gibQf@p/a@{lkmfN^`{\² K¨e©Â^aN 8sya@bwsufNj/¶»f8 8\_^`j;{fNz}fNq\_bw\_^kokm\km\_qsÇgifNjl 8ogbmfN^kkmolg`Z;aNk5<#\~x<ca
¦®vta@zuz}suo]q« 8fN\² (vN\_^`qa@j]suo]q(´Ë¶c¨e© 8\_bw\6gibwfN^ka@^`\;qfNo]j8bw\6 (sujìa L 4
t4
¦giffNz}suj]v(rN\²k`km\_z§´èlz}zu\6
d=subwZzusÇn o]sy Ø ¢ y¦ $L$K½K´fN^=zusÇn o]sÇ (¡:\h¦§¿½K´`¸8bmfh^`\6 8og\j]fNsÇkm\S 8o]\cbmfbwZ]\_^`qa@z/a@vNsubËa@bms}fNj¶
i \
bDå {_^føS c_kFb C "s}bmZìkmog`ZD 8\_bm\²gibmfN^k_¸l¨© aNkmbm^`fNj]fNqTh{\6gifNqh\6k:qfN^`\*zusuÍN\KfN]bwsyg_a@zaNkm«
bw^`fNj]fNqTN¶éØ:fNb#fNj]zuHg_a@jD 8su^`\6gib=suqXa@vN\6kFf@pQbwZx\;kmÍt{\KfN{]b`a@s}j]\6 4¸ 8s ±Ä^aNgibms}fNj]«ÅvN^a@bms}jxv 8]
8t_7
]
tt g_a@j{\olkm\6 øbmfßkmbwo 8ߨ©%km\6gbm^a8¶õ¡=f@d[\_rN\_^6¸QsubsÇkXs}qfN^`bËa@j8bKj]fNbbmf¾s}tj Om\²gibo]j8d=a@j8bw\6
{laNg`ÍvN^`fNo]jl K^aN sÇa@bms}fNjTs}j8bmf*bmZ]\c 8\_bm\²gibmfN^6¶Y[Z]syk¹qh\6a@jlkbwZla@b[a@zuzlbmZ]\gifNqfNj]\_jÚb`k¹f@pÄbmZ]\#s}jkmbw^`o]q\_j8b
Zla6rN\bwf{\*qfNo]j8bm\² Xs}jlkmsy 8\cgi^`NftkmbËa@b`k_¸]giftfNzu\6 (l\_^`Zla@lkbmfz}synNoxsÇ (¡:\|zusuo]qÆbm\|q\_^a@bwo]^`\6k#¦§¿½K´`¶
Ù Nf q\;f@pbmZ]\{aNkmsÇgë]^`s}jgsu]zu\6kcsuqlfN^`b`a@j8bcs}jübwZ]\¨©Òd[\|^`\T\6kmbËa@{]zusykmZ]\6 ìn o]subw\\6a@^`zuDfNj/¶¨Åj
È_áNÁNÁ8¸ Ù su^ "q ¡#ÃQ© Ù Ï[¡Ã>IÂ 8syk`gifÝrN\_^`\6 ¨e© ^aN 8sya@bwsufNj{8xzÇaNgisuj]vbwZ]\_^`qfNq\_bw\_^k:{l\_NfNjl ÌbwZx\;^`\6
\|j Sf@plbwZx\ Ù oxj/³°kkml\6gibm^`o]q¶¨ÅjÈ_á â8¸Ïìܨ Õ H ¨ Ù » cY:¡¦ÅÜ^`f@p®\6k`kmfN^f@p Õ kmbw^`fNj]fNqTca@bà 8suj8{]o]^`vNZ4´
olkm\6 ìa@jü\|zu\6gibw^`syg_a@z騩- 8\_bw\6gibmfN^ha@bKa(Z]s}vNZ¸/ 8^`ìkmsubm\hsuj¾Y\|jx\|^`s p®\bwfq\6aNkmo]^`\TbwZ]\¨© ^aN 8sya@bwsufNj¬f@p
bwZ]\;»ÌftfNj/¶:¡=\TlfNsuj8bw\6 CZ]syk \|bm\6gibwfN^*a@b#bmZ]\;»ffNjDa@jl (bmZ]\_jDa@bbmZ]\;gizu\²a@^ckmÍj]\6a@^#bmZ]\×»ÌftfNj/¸suj
fN^ 8\_^=bmfhgifNql\_jlk`a@bm\cp®fN^:bwZx\ר© ^aN 8sÇa@bmsufNjp®^`fNqÆbmZ]\×Ãa@^`bmZ/³°k[a@bmqftkm]Z]\_^`\NtbmZ8olk[Zx\*]sufNj]\_\_^`\6 kmÍt
kmo]{]bm^aNgbmsufNj/¶
¦ËÈ
´=¨©2^aN 8sya@bms}fNj¼¯ km\_\6k:bwZx^`fNo]vNZ/³]rN\_^` \|jkm\×gisu^goxqXkmbm\_z}zya@^_î6sujÚbm\_^kmbw\_zuzÇa@^ çQ
ñæÅQç= ¶ Õ
mk bw^`suÍtsuj]vK\_ÐÚa@q]zu\#sÇkFbwZ]\Sg\_j8bm^`\f@pfNo]^<ca@zÇa@ÐH¦§bmZ]\*»suzuÍtS"¾a6x´Ë¸8d:Z]sÇgÑZsyk[a;gifNq]z}\_Ða@jl ^a@]sy 8zu
\|rNfNzurtsuj]v^`\_vNsufNj/¶ Õ bKrtsykms}{]zu\;d=a
rN\_z}\_j]vNbmZlkbwZx\Xa@{kmfN^`]bwsufNj¬bwf@d=a@^ ]k#bmZ]\<ca@zyaNgbmsygÏ\_jÚbm^`\sÇk ÓNÁ
¢
qXa@vNjxs}bmol 8\6k¦ÅapaNgibwfN^f@pÈ_% Á fN^4fNj]\qsuz}zus}fNj#qsuz}zus}fNj´Ë
bwZ]5
\ <*Ïsyk/gifNq]z}\_bm\_z}#sujÚrsÇkmsu{]zu\N¸|a@jl fN^ 8sujla@^`
]Z]fNbmfNvN^a@]Zlk#kmZ]f@d¤fNj]zu^`\_zÇa@bmsurN\|zuDjx\²a@^`{8kmbËa@^`«gizufNol ]k_¶K·o]b*a@bKÀ8¶°À@Ê4q%bmZ]\a@{lkmfN^`]bmsufNjsÇkfNj]zuDÓ
¢
qXa@vNjxs}bmol 8\6kS¦a@{kmfN^`]bwsufNj(paNgibwfN^È_tÁ TïÆÈ_Á ûÒÈ_â´#a@jl bmZ] \ <SϪsyk#^`\6aN 8s}zu(qXa@xl\6 DkmZxfÝd:s}j]v
s}j8bm\|jkm\*^aN 8sya@bms}fNjDkmfNo]^gi\6k#a@jl gifNq]z}\_Ð(rN\_z}f8gsubms}\6k_¶*YFZ]\;·^aNg`ÍN\_bmbwA« ¼z}suj]\Kf@pQZ8Ú ^`fNvN\|jDa@bmfNqk#g|a@j
{l\TfN{lkm\_^`rN\² ¸Äa@j bmZ] \ #fN]]zu\_^*kmZ]s p®bK¦yp®fN^`qTo]zÇa8
~ ê Nê¼¢ ï ¡)F£
´=f@p=·¤^ ¬vNs}rN\6k#rN\_zufÚgisubwsu\6k_¶h¦Åg_a@j 4ç
d[fN^`ÍhfNoxb=bmZ]\cd=a
rN\|zu\_j]vNbmZf@pbwZ]\K·^aNg`ÍN\_bmbwA« zus}jxF\ ¥@´`¶
á
¦ â´YFZ]\[\²a@ÍK^`\_vNsufNj;f@plbmZ]\#kml\6gibw^`oxq f@pgiftkmqsÇg¹qsÇgi^`f@d=a6rN\{laNgÑÍtvN^`fNo]jl ^aN 8sya@bwsufNj;a@bQÀ8¶x$½
sÇk#q\6aNkmo]^`\6 (sujDbmZ]\;qsuz}zus}q\_bm^`\*^a@j]vN\N¶cYFZ]sÇk¥^aN 8sÇa@bmsufNj/¸Z]s}vNZxz}kmsuvNj]s èÄg_a@j8b[p®fN^cfNo]^o]jl 8\_^kmb`a@jl 8suj]v
f@pbmZ]\S\6a@^`z}Z]sykmbwfN^`f@pbmZ]r
\ E:j]surN\|^km\N¸8syk=gifNjlkmsy 8\_^`\6 s}jaTzÇa@bm\_^=zu\6gbmo]^`\N¶
LMONPEQ MO/RSH
TVUXWZY\[UX]_^a`b[Xcd[Xe;fD]_ghU@ch]Oijke lnmogogpq`*pqchWr[@s=eetvu=U6wXp6]yxz^_[D{X|(eH}~[XcoXpq`u=U6wXpq~pQcdX^ah]W8[(co[X^
ghpqcopq^_`kU@^apr^_op
RU@`b^_/]Y\[Xcd[]_goopq`bpX,T#UXW8Y\[^apq~p]bjy[Xgp6]jQU@c]_pqpD^_o`b[XmoXjy~\[XmhWd]omd^U@`bpDeOmhjb
U
SYjy^ap6W8feU@cotveUXW8pOY\c8^_pQ`_spq`bpqchjypXU@~\^_o[XmoXjypq`b^U@Ycjy`bmhjyYU@~zU@chWo]KU@`bp]_mogog[]_p6Wr^a[rpXpqgo^
jk~p6U@`6
i {X{8/*U@`b~X8#¡ 8¡¢£lv¤pq~\~o¥pq~\pqgod[Xcop¦/U@h]yx4W8pq^ap6jy^ap6W`kUXW8Y[¡§co[XY]_pXQU@^¨X|=©Pª:«Vl¬®¯iq°Fexb
U@goghU@`bpqc8^a~fjy[XeY\coSs`b[Xe%^aop©Y~\f±
U
f
ldpKu=UX]#~\[[XYcoCs[X`*]_[Xmo`kjyp6]=[@s#§Y\c8^_pQ`_spq`bpqchjypX²xb*©PU6³_[X`
W8pqwXpQ~[Xgoepqc8^K[@s*`kUXW8Y[´UX]_^_`b[Xco[XeOf´WZYWco[X^[µjqjymo`moc8^aY~VU
s^_pq`±r[X`b~W¶±
U@`¨8zomdY\~W8Yco[Xc£`kUXWoU@`
pQ·ghpq`b^_Y]_pX¸Y[Xcopqpq`k]#Ychjy~\mhWZpWT¢¹¦ºl»¼U@e;d`bYW8Xp@xb/¦½:¾¹¦/¦ln[8W8`bpq~\~¤zU@codxª#¢%lv©DU@~wXpq`bc4x
U@chW¸4¿± µÀ¢ÁlÂ:mh]_^_`kU@~\YUxb
Ã:coY^=[@s`kUXW8Y[Ähmo·h#iµU@c]_f®Åiq|8ÆhÇbÈF±£eÆhÇqª:«XÆ4É
¥ oY]u=UX]^_op¡uzp6U@Xp6]_^z]_Y\XchU@~W8pq^ap6jy^U@d~\p¡8fOpU@`b~f;`kUXWZY\[K^_pq~\p6]bjy[Xghp6]q}Y\^zY][@s^_pQc(mh]_p6WOY\cC^aopÊËTU@chW
z
[Xgo^aYjqU@~4co[@u*oU@~^ao[XmdXH]_eU@~~\pq`zmocoY^b]z]_mhjbUX]zefU@chWCÌ4fU@`bp¡e[X`bp¡gd`kUXjk^_YjQU@~4^_op6]_pKWoU6f8]q
]_[X~mo^_Y\[Xc¶Y]O^_[jy[Xe;oYcop^_pQ~p6]bjy[Xghp6][@wXpq`;~[XcoDhUX]_pq~Y\cdp]OUX](çaèéê4ëìbê¿ëíRîïêoéê¿ëðo£¥Fop
æ
]_pQgU@`kU@^aY[XcåñÙhpq^Ëuzpqpqc^_op=^_pq~\p6]bjy[Xghp6]go`b[8W8mhjyp6]RUÝgohUX]_p#W8Yóòôpq`bpqchjyp=pq^ËuzpqpQc^_op¡]_Y\XcU@~]ÀY\c^_op#^Ëuz[
`bpjypqYwXpQ`k]q¤fjy[Xcocdpjy^_Y\cd^aopO^Ëuz[^_pQ~p6]bjy[Xghp6]Ýu=Y^aU(jqU@o~p^_opqY\`;]_YXchU@~]¡jqU@cpUXWoW8p6W4Êns=^_op
gohUX]_p¡W8Yóòôpq`bpqchjyp#hpq^Ëuzpqpqc^_op¡^_pq~\p6]bjy[Xgp]FY]zUÙe;mo~^_Y\go~p#[@s{XõX|ö68^_op*]_YXchU@~]UXWoW(jy[Xch]_^a`bmjk^_YwXpQ~fh}Yós
^aopOgohUX]_pKW8Yóòôpq`bpqchjyp*Y];iq÷X|öÒ^_op;`bp6]_mo~^bU@cµ^:Y]:«qpq`b[o*ÊvcPXpqcopq`kU@~ÂZs[X`*U]_[Xmo`kjyp;U@^#pq~\pqwÒU@^_Y\[Xc
ø4^_op
u=U6wXpqtns`b[Xc8^]À`bpqY\c8s[X`kjyp¡YósRñRjy[]oøÙ®£ù¬ôdu=opq`bp*ñOY]z^_opK]_pQgU@`kU@^aY[Xc[@s^adp*Yc8^apq`_spq`b[XepQ^_pq`zpq~\pqepqcµ^b]q
V]À^_op¡]_[Xmo`kjyp=e[ÒwXp6]UXjy`b[]b]^_op¡]_fX^adp¹jy[XeOoY\cdpWK]_YXchU@~o^_opq`bpys[X`bpzÄhmjk^_mhU@^_p6]RUX]øwÒU@`bYp]q
XY\wY\coO`bY]_pÝ^a[UjbhU@`kUXjy^_pq`bY]_^_Yj¹ghU@^a^_pq`bcHjqU@~~\p6WPçaèéê¿ëúìbê¿ëê/èûhêïìyëoçaèüÚêðo¥FoY]z[µjqjymo`k]:[Xco~fY²s^_op
U@coXmo~U@`/]_Y\«qp[@sZ^_opz]_[Xmo`kjypY]4~p]b]¿^aU@c¬hàÒñln`kUXW8YU@c]yx}@YósZY^/Y]4~U@`bXpq`6q^adps`bYcoXp6]s`b[XeÅWZY²ò4pq`bpQc8^4ghU@`b^b]
[@s=^aop]_[Xmo`kjypOu=Yghp[Xmo^^adpågU@^a^_pq`bc/( [(^aop^_p6jbocoYýXmdpåXYwXp6]¡Yc8s[X`beU@^_Y\[XcrU@h[Xmo^Ýþhcop]_^a`bmjk^_mo`bp
Y\cï^aopD]_[Xmd`kjkpXomo^YcÿU
]_[X`b^O[@s*^bU@coX~\p6Wãs[X`beHÊv^Y]Ogh[]b]_Yo~p^_[W8Y]_pqc8^U@coX~p(^aopYc8s[X`beU@^_Y\[Xc
U@h[Xmo^^_op]_fd`bY\X8^_cop6]b]K8f
mh]_Yco
U
jy[Xegomo^_pQ`^_[ãjy[XeOoY\cdp^adps`bYcoXpYc8s[X`beU@^_Y\[Xc^bU@XpqcÿU@^
W8Y²ò4pq`bpqcµ^Yc8^apq`_spq`b[XepQ^_pq`Ú]_ghUXjyYco
] ÙU@~^_o[XmoXKpq·go~U@YcoY\cdVd[Òu¶^adY]^_`bYj Kuz[X`bµ]Y]co[X^pUX]_fu=Y^_o[Xmo^
eSU@^_opqeU@^aYjq]q
|8âiq°X÷(ßr{X|X|8Þ|X|X|gO
Þb|X|X|(e£]qÆ4É6x¹Y\wYW8Yco(8f´Uª:modo~\pjy[Xch]_^bU@c8^*[@s¡õX°e£]qÆ¿Ék©g4jÒÆ¿É6^_op
W8Y]_^bU@chjypKY]*U@gogd`b[Ò·Y\eU@^_pQ~f
{X|(©g4j@;¥zoY]#Y]#^_op;cop6U@`bp6]_^¡[@sz^adp^t o½¹]q4¥FopqfhU6wXpKpqpqcDs[XmdchW
[Xmo^=^_^ [ w&y°U@chW(]_pQ`bwXpUX]zgo`b[Xp6][@s^_opKW8Y]_^U@c8^=U@chWp6U@`b~fÃ:coYwXpQ`k]_pX
)mlPz-δ×HÓP+Ð Ô Q
Îãչδ×HÓPÐ@× ×PÕ*ÔÐ ±rp*co[@u£`bp6U@~Y]_p#^_hU@^zuzp6U@Xpq`=ýXmhUX]bU@`btv~Y\Xp§cµmdt
kj ~p6U@`VUXjy^_Y\wY\^ËfY]FY\cCsÂUXjy^VýXmdY\^_p;jy[Xeåe[XcYc(cop6U@`bµfU@~U@·Yp6]qt#mhUX]bU@`k]uzppQ~YpQwXpXoU@`bp`kU@`bp¡]_moghpq`bt
gh[@uzpq`_smo~µ#jy^aYwXpV¡U@~UXjk^_Yjz=mjk~pqY4lco[Òu:cåUX]¡Ýx¤[X^aýXmhUX]bU@`k]U@chWÙcop6U@`bµf¡{¹ hU6wXpzU@~~Z]_[X`b^b]
[@s¼[X^_opq`*pq·^a`kU@tv[X`kW8YchU@`bfgo`b[Xgpq`b^aYp6" ] ^aopqfPpQeYV^ |#tv`kU6f8]¡U@chWDÊT,U@chWÛmo~\^_`kU
wY[X~\pq^¡~Y\X8^q4U@chWP^aopqf
ÄhYj Xpq`66Ä4U@`bpVU@chW sÂUXW8p:s`b[Xe,WoU6fK^_[KWoU6fX¸pq`bhU@g]Ú]_^_`kU@coXp6]_^[@sÚU@~\~vX¾¡¦/¤zÊeSU@g]hU
wXp=]_o[Òu:c§o~[Xh]q
[@seU@^_^apq`#U@goghU@`bpqc8^a~feå[@wYcosÂUX]_^_pq`V^_hU@c(~\YX8^Q[ å^_oY]zY]=YcCsÂUXjk^#U@c[Xgd^aYjqU@~ Xpq[Xepq^a`bYjqU@~4Y\~~\m]_Y\[Xc
åomo^Y^=jQU@c([Xco~f;uz[X`bOY²seSU@^_pq`bYU@~Y]hpqYcoKpq·gpQ~~pWC^_[Òu=U@`kWo]Àmh]wXpq`bfjy~\[]_p#^_[;^_op*]_gpQp6W[@s~\YX8^Q
± hU@^#jQU@cPpq·go~U@Y\c^_op6]_p;]_^_`kU@coXp*gd`b[Xghpq`b^_Y\p6]U}
[XepK]_[X`b^=[@s=§²jypqcµ^_`kU@~/pqcoXYcopXoY]:go`b[µWZmhjyY\co
£
µmdXpU@e[Xmoc8^][@s¡pqcopq`bXf£lv]_[XeåpC[@s¹^_opýXmhUX]bU@`k]OU@`bpUX]O~\mdeåYco[Xmh]UX]iq|X|P^ËfgoYjqU@~zU@~U@·Y\p6]yxKU@chW
pQ·ghpq~~\Yco;³_pq^b][@sVgo~UX]_eU8åª:[@uzpQwXpq`K^_opWoU
~ f ^_[ µWoU6f(ÄYj XpQ`bYcoep6U@ch]Ý^adY]Y]*U@~~[µjqjymo`b`bY\cdYc
U
wXpQ`bf]_eU@~~ô]_ghUXjy p co[;dY\XXpq`z^_hU@c[Xmd`V]_[X~U@`=]_fµ]_^_pQ e zlv:c[Xµ³_p6jy^=jQU@c ^=j hU@coXp*jy[Xopq`bpqc8^a~fKsÂUX]_^_pq`
^ahU@cCY\^=jqU@cjy[Xee;mocdYjqU@^_pVUXjy`b[]b]Y^b]zghU@`b^b]q}Xomo^z^_op:sÂUX]_^_p]_^z^_oY]zjqU@chU@gogpqcY]^_op¡]_gpQp6W[@s~\YX8^Q
[
Yós*fX[Xm£]_pqpwÒU@`bYU@^_Y[Xch]O[XcÿUr^_Y\ep6]bjqU@~\p(^q=fX[Xmco[Òu?^_op[Xµ³_p6jy^jQU@cdco[X^hpoYXXpq`^ahU@c£]_Y«qp
T ® hm x¥Fop#hp6]_^=UXjqjypQgd^ap6WCpQ·go~U@chU@^aY[XcY]^_hU@^=UK]_mogpq`beUX]b]_Y\wXpzo~UXjbOo[X~pVY]YcµwX[X~wXp6W(lgpq`bhU@gh]
iq~
| 1k
x(¥zop~p6jk^_mo`bp[X c |#tv`kU6fDUX]_^_`b[Xco[XeOfu:Y\~~¼WZY]bjymh]b]#d[Òuº^_[Dpq·^_`kUXjy^*pqcopq`bXfPmh]_Y\cdD§X`kU6wY\^f
gh[@uzpq`6\}8]_pqp U@~]_
[ #`#ª:pU
wXpqch]q8~pjy^_mo`bp6]z~U@^apq`:^aoY]z^_pq`beâQâqâ
)+[-Ø ÕKÖRÎ¶Í¶Ö Êvci õ
[µjypq~\fc¤z¦/¦ïlÂU@^*W8Yc8omo`bXs[X`¡e[]_^:[@sR^adpi ÷X|X]yxFWZY]bjy[@wXpQ`bp6W
U `kUXWZY\[;]_[Xmo`kjyp¡u:Y\^_r§o~\Ygh]q[@sVi@ {X{
]_pjy[XchWCgpQ`bY[µW¿RFs^apq`¡dpjy^_Yj]_gpjymo~U@^_Y\[XcU@h[Xmd^V¦,¡© l~Y^a^_~\p
;
X`bpQpqcepqc/âqâqâyxo]_op¡U@chWC¥/[Xc8fªVÀ±ÿÊ µªU@^_^a`bYomo^_p6W^adp*]_Y\XcU@~]À^a[UK]_goYcocoYco*cdpQmd^a`b[Xc]_^bU@`å^_op
þhchU@~~\fKjy[X~\~U@gh]_p6WK`bpqechU@c8^[@s¿U]_mogpQ`bcd[ÒwÒU¹pQ·go~[]_Y\[Xc/¥zdpQ[X`bfKUXWKgo`bp6W8Yjy^ap6WK^_op#pq·Y]_^apqchjyp:[@s¿]_mhj
[Xµ³_p6jy^{] ÛU]_[X~U@`¡eUX]b]¡jy[Xego`bp6]b]_p6W(Y\c8^_[´iq|@eá`kUXW8Ymh]q4jypQc8^_`kU@~W8pqch]_Y\^f iq|É qRjye Æ úå:[X[µW8f
hUXW`bp6U@~~\fpq·gp6jy^ap6W^_[]_pqp¡[XcopX
=[Xmo^K°X|X|(gomo~]bU@`k]¡U@`bpco[@uºco[@u=c/(-spquÜ]_goY c y iq|X|X|C^_Y\ep6]ghpq`;]_p6jy[XchW4(¥zopqY`*`kU@^_p
[@s ]_~\[@u=YcoXtËW8[@u=c/U@cWjk[Xc]_pýúmopqc8^;~[]b];[@spQcdpQ`bXfX=jqU@cpepUX]_md`bpW¿£¥zopqfUXjy^SUX]C§²jy~\[8jbµ]qYc
g `b[ÒwYWZY\co#mh]_pysmo~8[Xh]_pq`bwÒU@^_Y[XchU@~8^ap6]_^b][@sôXpqcopq`kU@~8`bpq~U@^_Y\wY\^f pQpÝ f& ¥/pq~p6]bjk[XgpXo pQgd^apqe;pq`Vi X |
o
U@chW:go`bY~Ri X °8
)+[- ×DÑc Ö *%ÐÒ× *%ÐÒ×PÍD Ñ %v Î Ô Îr × kQ¶ÍD Ñ kÏ ÍÎãÏÐ6ÎÓPÐ@ Ñ Êv^u=UX]SpQ·t
g p6jy^ap6Wr^_hU@^;U@^K]_o[X`b^¡u=U
wXpq~\pqcoX^_h]Klvjyex#^_op`kUXW8YU@^_Y[Xcs`b[Xeá^_opU@`b^a/]U@^ae[]_goopq`bpX/U@chWP^_hU@^
h
s`b[Xeº^aop#©Y~\fK±
U6flu%¬xXuz[Xmd~WÙhp:uzpU@]_[^ahU@^À^aop:`kUXW8Y[¡^apqegpQ`kU@^_mo`bpz[@s4^adpV]_f;U
u=U
f¡s`b[Xe
^aop©Y~\få±
U
få]_o[Xmd~WOv p ÿ|KOo¤mo^zY\cri õX°K¸R hÊËÝ 2±Ê˦ o½¡ºl³_mh]_^z~Y\Xp#µ#Ý µÝ¢K8^aopqf
uzpQ`bp#¤pq~\~h¥pq~pQgdo[Xcop=¦/U@Xomo^^_op6]_p=uzpq`bp#UX]_^a`b[Xco[Xepq`k]4^_`bfYco^a[Km]_p=U*]_ghU@`bp=U@c8^_pqcochU#^_[KeU@g;^_op
©Y~\f;±
U6fdx/s[XmdchWU@c[ÒwXpq`kU@~~8hUXjbX`b[XmochW4XU@~~h[@wXpq`R^_op¡]_fX8U@^¼U@[Xmo^:¬®
â °Kjyeµjy[X`b`bp6]_g[XchW8Yco
^a[d~UXj [µW8f`kUXWZYU@^_Y\[XcDU@
^ Cg {OhÊ ^VY]#W8Y Cjymo~^=^_[epUX]_md`bp*mocdY²s[X`beÁhUXj X`b[XmochWC`kUXW8YU@^aY[X c
^aopq`bpY]Fco[§pqego^Ëfho]_f^_[åjy[XeghU@`bp¡Y^=u=Y^_/
©p6UX]_mo`bpqepQc8^b]#U@^;]_o[X`b^apq`u=U6wXpq~pqcoX^a]¡hU
wXp;^_[DpeUXW8pOs`b[XeáoYXotnÄhfYcoU@Y\`kjy`kU
s^q/U@~\t
~\[[Xch]q4[X`K]_ghUXjypjy`kU
s^q/p6jqU@mh]_p;[@szwÒU@`bYU@o~\pU@h]_[X`bgo^_Y\[XcDU@cWDpqeY]b]_Y\[XcYcr^aopU@`b^a ]¡U@^_e[]_goopq`bpX
¥zopqf(hU6wXpjk[XcZþh`bepWC^_hU@^#^aopK]_gpjy^_`bmoe [@sR^_op»z©D¤jy~\[]_pq~fþh^b]VUOo~UXjbtvh[8W8fjymo`bwXpKU@^¹¨8
;
¥zopz`kUXW8YU@^_Y\[XcY]oY\Xd~\f¡Y]_[X^_`b[XgoY[
j *^_op=]bU@epRYc;U@~~ZWZY\`bp6jy^aY[Xch]q/¥FoY]4þ^]^_op=jy[Xchjypqgo^[@sh`bpq~\YjzW8Y~\md^ap
`kUXW8YU@^_Y[Xc;jk[XeYco#s`b[Xe2^adp=go`bYe[X`kW8YU@~@þ`bpQU@~\~Y\cO^aopzwXpq`bfKp6U@`b~f*pq·gU@ch]_Y\[XcK[@s4^_op=Ã:coY\wXpq`k]_pX¥z8mh]
^aopO»¼©P¤ïY]:XpqfpqwYW8pqchjypÝs[X`V¤Y¤zU@cojk[]_e[X~[XXfDlv]_pQp~U@^apq`#~\p6jy^_mo`bp6]yxU@chWY^b]VWZY]bjy[@wXpQ`bfu=UX]:U
eSU
³_[X`=^_`bY\moego[@s`kUXW8Y[åUX]_^_`b[Xco[Xe;fX
iq{
?A@CB@8? @';D:k?.<IFG<.H J
[XmochWÛpj o[Xtv~\[8jqU@^aY[XclÂ]_[XchU@`qx:Y]U(^ap6j ocoYýXmopOmh]_p6WD8fDU(c8moeOhpq`#[@sVjy`bpU@^_mo`bp6]q/]_mhj UX]hU@^b]U@chW
W8[X~\gdoY\c]q/T#UXW8Y\[Ýpj o[Xtv~\[8jqU@^aY[Xc;u=UX]WZpQwXpq~[Xghp6WÙY\cO±[X`b~WK±
U@`z¨8U@chWÙ`kUXWoU@`pj o[p6]s`b[Xeº^adpV©[[Xc
uzpQ`bp[Xh]_pq`bwXp6Wr^_opqc/D¥FopU
wXpQ`kU@XpW8Y]_^bU@chjypå[@s#^adp©[[Xc
Y]K{X÷X|8|X|X|(e]_[^aopC[Xmo^atËU@chWZtÂUXjb
pj o[åWZpQ~U
fåY]=¨SßDlÂ{X÷X|8Þb|X|X|àÒ{X|X|8Þb|X|X|x®£¨8â°å]_p6jy[XchWo]q
¤ pjqU@mh]_pP^adpDYcµwXpq`k]_pqtË]býXmhU@`bp(W8p6jk`bp6UX]_p[@s;]_Y\XcU@~¹]_^_`bpqcoX^_ U@gdgo~\Yp6]O^a[[X^a^_opD[Xmd^aX[XYco
gomo~]_p*U@chW(^_op*`bpq^_mo`bcoYco;p6j o[oo`kUXWoU@`:p6jbo[p]#U@`bp¡gd`b[Xgh[X`b^_Y\[XcU@~4^a[lvW8Y]_^bU@chjyp@xÆ7# [UO~U@`bXp*`kUXWoU@`
^apq~p]bjy[XgpY]Ýcopqp6W8p6W4¥zopDiq|X|X|@tns^*:`bp6jyY\[`kUXW8Y[W8Y]_UX]¡pqpQcpQ·^apqch]_YwXpq~\fPmh]_p6W(s[X`Kgo~U@copq^bU@`bf
`kUXWoU@`6Ú¥Fop=e[]_^W8Y]_^U@c8^`bp6UXjbhU@d~\pz[X8³_pjy^Y]F µU@^amd`bc/}XU@^RY^]jy~\[]_p6]_^U@gogo`b[UXj ;[@s/÷8 °#]_^a`b[Xcd[XeåYjqU@~
Ã=cdY\^b]:^aopÝpj o[åWZpQ~U
fY]:¨ßlÂ÷8â°CßÛiq°X|8Þb|X|X|8Þb|X|X|x_àÒ{X|X|8Þ|X|X| ®ÿ÷X°X|;]_p6jy[XchWo]q
T UXWdU@`4eSU@gdgoY\cdz[@sd©Ûpq`kjkmd`bf¹U@cW¡¾pqc8mh]¿hUX]4pqpQc;jqU@`b`bYp6W¡[Xmo^8f¹]_gUXjkp6jy`kU
s^q/¾pqc8mh]q6]_mo`_sÂUXjyp
V
Y]#gpq`bghpq^_mhU@~~\f(oYWoW8pqc8fW8pqch]_p;jy~\[XmWo]:[@s=»=½ Ç hdmo^#hUX]#pQpqcP`kUXWoU@`btveSU@gdghp6WYcDX`bp6U@^¡W8pq^U@Y~8f
^aopD©P ¡R¦¦/#%]_ghUXjyp6jy`kU
s^lvV]_^_`b[Xco[XeOf:[Òu ¹pj@ºi X i@}¡ f á¥pq~\p6]bjy[XghpXz©DU@`kj i X ¨xb
¥zop`kUXWdU@`KeSU@gdgoY\cdgo`b[XX`kU@eeåpO^apq`beYchU@^ap6WY\ci X ¨8RU@cWru=UX]s[X~~\[@uzp6Wr8frX`kU6wY\^bU@^aY[XchU@~tnþhpq~W
eSU@gdgoY\cd f º¥pq~\p6]bjy[XghpX88U@cµmU@`bfHi X { v
Z\[Ubpá>^UoUoUqsb,Öu}md_cbpZbrqsb:jzc]_}mb!~jh|[UbÞâmyjkhãjk}d_cb:cyd¨hh|]´~fmcqs^
b!AC~slÞdD`_~sfU`"dDl!]_fU`¸~shzj
¨^Ufmch|~s]_fmjeäZJ[U~jO~j¤då§dDq^dDoUqb"c]_f<h|yz~soU^Uhk~]_fw~sfäd
U
~sh|~s]_fæhk]hk[Ub,]_}Uh|~cedDqA~slpdD`_~fU`çè^mdDq~sh¦_wfU]Da
yzbjkhk]_yzb
hk]!~hzj\}vyz]_}mbeyAcedD}mdDoU~sq~shk~b:je
ÈAqshk[U]_^U`_[p~sh\[md_j·ombebef¸l!bef<h|~s]_fUb
h|[dDh\~f<hkbryjkhkbeqqdDy\jk}md_cbX~jydDhk[Ubey\]_}md_çè^UbXdDh\aAd:§_beqsbrfv`_h|[mj
et_t fUl"wdDfbrÒ}mbeyz~sl!brf<hO]_fhk[UbOºk]_~f<h¤È>}]_qqs]_ kÖu]Du^U¸jk}d_cb:cyd¨h?~sf eÎÏ ±Hjk^v`_`_bjkhkb
hk[mdDh!hk[Ub
~ f<hkbryjkhkbeqqdDy[¬
yz]_`_befp~j}mdDh¢cn[<?befU]_^U`_[phk[mdDh\a\b'lpd:YombdDoUqbh|]!jkbrbjk^21¸c~sbef<hWjkh¢dDyjhk]?}vyz]_omb'hk[Ub
~jkhkyz~soU^Uhk~]_fY]D3[¬
yz]_`_befpcqs]_^
j~fphk[UbXjk]_qdDyWfUbe~s`_[<om]_^Uyz[v]u]
]D3]_^Uy £ dDqdDÒu_WZJ[UbAyzb:jk]_fmdDfcb>q~sfUb:j
±
Á
A g hDB
~jzc]D§_beyzb
5' yd¬j²~f eÙ_Î ±2/[Ub,`_]_hOh|[UbGâmyjkh?ß>]_obrq4iyz~sebp~sfji·[<¬jk~crj¨]_yY[U~jOa\]_yzm
5Ê yd¬j`_]´h|[Uyz]_^v`_[}mbe]_}UqsbpoU^vh¤ ¨]_yzh|^vfmdDh|beqsUfU]_hOh|[vyz]_^U`_[hk[Ub¸dDhkl¤]ujk}v[Ubeyzb
lk]_^æfvbrb
yz]¬cn_brhzj
]_y?jk}md_cbcyd¨hOhk]
m] 5' ydµd_jkhkyz]_fU]_l?_¸ZJ[UbYâmyjkhOc]ujkl!~cn5Ê yd¬jewبyz]_l h|[Ub Öu^Uf3gj.c]_yz]_fmd<wa\beyzb
beh|b:ch|b
¨yz]_l' ¦¥Yyz]¬cn_brhzj\~sf eÎDuÙ
È j·a\bA[d:§_bAjkbebef3w_hk[UbÊx>®`udD}
Ê et_t Ü et fvlò~j·q~shkh|qsbAbeÒu}vq]_yzb
jk]dDy: ]be_]_f
hk[mdDhew5' yd¬j
dDyzb.c]_f¤_bef<hk~]_fmdDqsqs
~s§u~
b
~sf<h|]Z
Ö ]D¨h5' yd¬j
u t Ü be
_ Ù Ü t Ù U f l,
á'dDy
5' yd:<j Ü et_t _ be t Ù Ü t t__t Ù fUl,
T
opCDÁµÂPq À R)Ä_¿A¿\ÄD½Å
P 5' yd?}v[U]_h|]_fjWdDyzb>lÞdD~sfUqsO`_befUbeydDh|b
o<¤dl!~Òh|^vyzbÊ]D3hk[Ub'}Uyz]<cb:jzjkb:j
~jzc^mjzjkb
b:dDyzq~sbey:ÊÃp8ÿ*Ar^KÿC=Kÿ ÿB-s' ¨yz]_l `ud_j'dDh et)t Ü et)uOý ~j>]_fUbãc]_l!l!]_f
brl!~jzjk~]_fYl!bcn[mdDfU~jkl"3Z\[Ub']_hk[UbeyW~jew}mdDyd
]DÒu~crdDqsqs
wDhk[UbXjzdDl!bAl!b:cz[mdDfU~jkl6h|[mdDh`_befUbeydDh|b:j·h|[vbʧ_beyz
q]DaA befUbeyz`_¸yd
~],bel¤~jzjk~s]_ f ZvBØÿw*: P)x4
\ ÿ B<
ÿ C ÿ B-s'>¤ZJ[Ub?yzb:d_jk]_f ¨]_yOh|[U~j
~j'hk[mdDh'hk[Ub?^UqshkydDyzbrqdDhk~§~jkhk~cbeqb:chkyz]_fmj>h|[mdDh`_brfvbrydDhkbOhk[Ub?jkufmcn[Uyz]_hkyz]_f¸bel!~jzjk~s]_f"crdDfªjk}mdDfHd![<^U`_b
ydDfU`_b']D3brfvbryz`_~sb:jewuhk[<^mj·`_~§~fU`?c]_f<hkyz~oU^vh|~s]_fmj~sf¸dDqsqmaAd§_beomdDf
je3È°h|[U~sy
}m]ujzjk~soU~sq~sh¦O~j\d[<uoUyz~
]D
h|[Ub:jkbh¦a\ ] ZDy?ØÿïÇ$*z{B*|}Aá>beyzb_wmd²q]DaA befUbeyz`_!}U[U]_hk]_fLcedDf,ombOjkh|yz^cz!o<¸dDf
^UqhkydDyzbeqdDhk~s§u~jkh|~c>brqsb:ch|yz]_f3wUdDf
u~cz_b
^U}¸hk]?[v~`_[Ubey\befUbeyz`_~sbjewydDhk[Ubey\q~s_b.dOodDqqombe~sfU`¤jkhkyz^mcn¤o<!d
omdDhr
opCDÁµP  q ôP
À O.Àp¿JÇH½Æ,Àp¿ 5' yd¬j`_]Hhk[Uyz]_^U`_[Ël¤]ujkhOlpdDhkbeyz~dDqj.dDf
cedDf3ghOombG¨]¬c^mjkb
o<
]_y
~fdDyz.hkbeqb:jzc]_}mb'l!~syzyz]_yje3xWdDyzqsv5' yd:h|beqsb:jzc]_}b:j^mjkb
`_yz~
jh|]
bâmfvb\h|[Ub>âmbeq
]D¨ §u~sbraw_oU^Uh·]D
c]_^vyjkbʧ_beyz!}]u]_y
bâmfU~shk~]_fpaAd_j]_ovh¢dD~sfUb
~sfph|[v~jaAd_~¶ÿBhBC)
B-0
n5' yd:Yh|beqsbjzc]_}b:j
a\bryzb
be§_beq]_}b
~fOhk[Ub eÎÏt jdDf
a\beyzbW^mjkb
¨]_y`_]u]
~slpdD`_~fU`>~fOhk[Ub\[U~s`_[Uqs.jk^mcecb:jzj¨^UquxW×ßXÖ<ZAxצß
jk}md_cs
b 5' yd!]_omjkbeyz§dDhk]_yz
Z [UbpjkhzdDyzhO]DÊhk[Ub¸¥ jkh!ÌWbef<hk^UyzH[md_jYjkbebrfËh¦a\]Hlpd:ºk]_yOfUbea5Ê ydª]_omjkbeyz§dDh|]_yz~sb:jX~sfjk}md_cb
w
\
aA[U~cz[!dDyzbAjkbrh·h|]h|ydDfj¨]_yzl h|[vb\âmbeq
aA~shk[
dDh¢d']D8]_}Uhk~cedDqs q~s_bAç_^mdDqs~sh¦_Ýß'ÈÖ¬Èg{jJÌW[dDf
yd eÎ_Î_Î
aA~sqq`_~§_bXdDycejkb:c]_f
¦jzcedDqb\~slpdD`_~fU`Yd_j`_]]
d_j]_}Uhk~crdDq zWZ\[vbx·^Uyz]_}b:dDf5s&m& aAd_jWdDqjk]OqdD^vfmcz[Ub
~f eÎ_Î_Î AZJ[U~j\[md_j et hk~sl¤b:j>}]u]_yzbey'yzbjk]_qs^Uhk~]_f,hk[mdDfªÌW[mdDf
yd<wuov^UhA~j>l?^mcn[pl!]_yzbãjkbefmjk~sh|~s§_b_w<aA~sh|[
3
dDom]_^Uh\±hk~l!b:jh|[vbc]_qsqb:chk~fU`?dDyzb:dY¨b 8b:chk~s§_brqs et_t_t cl /h|[v~jlpdOfU]_h«jk]_^vf
l?^mcn[3wDoU^Uh~f!]_y
bey
h|]!`_behAbe§_bef"h|[v~I j 5&& [md_j\hk]!ombXhk[UbOjk~eb]DdYoU^mjz
;E±N>
Ç O.»¿WôÀ!Áæ¿ « ½ ²³~
 OX«  o¶ÄDÀÞ¿ È>fU]_h|[vbry\jk^Uyz}Uyz~jkb]D3bdDyzqsv
5' yd:?d_jkhkyz]_fU]_l?~jhk[mdDh
h|[UbXjk}md_cb'obrha\brbefp`udDqdDÒu~sb:j~fpcqs^mjkh|beyj]D3`udDqdDÒu~sb:j~jfU]_hbrl!}Uh
woU^Uh·âmqsqb
aA~shk[¸d
~´þ^mjkb'[U]_h`ud_j
¶µ et)·Yý Aa>[U~cz[H~j'bel!~hkhk~fU ` 5' yd:<je.×fjk]_l!b?cq^mjkhkbeyj'hk[U b 5' yd:<j'dDyzb¤c]_fmcbef¬hkydDhkb
dDyz]_^Uf
~f
~§~
^dDq`udDqdDÒ~b:E j /~sfæ]_hk[Ubeyjh|[vb
~jkh|yz~soU^Uhk~]_fª~j?jkl!]u]_hk[Ubey:ª'fmcb¸dD`udD~sfh|[vb¸^Uqshk~lpdDhkb¤befUbeyz`_
jk]_^UycbX~jJ`_yd:§~h?Ü!h|[vb.`ud_jJ`_]_h\[U]_hÊd_jJ~h'jkh¢dDyzhkb
hk]pc]_qsqdD}mjkb
^Uyz~sfU`Oh|[vb.[U~jkhk]_yz¤]Dh|[vbã>fU~s§_beyjkb
Öu^mcz["dOqdDyz`_bXlpd_jzj\]D`ud_j\hzdD_b:jÊdY§_beyz!q]_fv`?hk~l!bh|]pc]u]_qÜpqs]_fU`_beyAhk[mdDf¸hk[UbOdD`_bX]Dh|[vbã>fU~s§_beyjkb
w
jk].~sh~jWjkhk~qsqU[U]_h·h|]
d:_ZJ[UbAhkbel¤}beydDh|^vyzbÊdDf
q^Ul!~sfU]ujk~sh¦X]D3hk[U~j[U]_hW`ud_jaA~sqq
br}bef
]_f?hk[Ub'h|]_hzdDq
`_yd:§~hzdDh|~sfU`Xlpd_jzj]D8hk[UbÊcqs^mjkh|beyÜO~sfmcq^
~fU`h|[Ub'jkh¢dDyj~fYh|[Ub>`udDqdDÒ~b:jew_hk[Ub>`ud_j~sh¢jkbeqÍ|wudDf
dDf¬
dDyz
lÞdDhkhkbryOhk[Ubeyzb?lpd"omb_8 &´dD}U}U~sfU`p]_^Uhh|[vb¤cq^jkh|be6 y 5Ê yd"bel¤~jzjk~s]_fµcedDfªh|[Ubef`_~s§_b?^j.cqs^Ub:j.dDo]_^Uh
dDyz!lÞdDhkhkbry:wUdDf
[U]Da®~h>~j
~jkh|yz~soU^Uhkb
WÖubeb¨]_yAbeÒ¬dDl!}Uqsb?ÖupdDf
Z3beqsb:jzc]_}v b &HdDycz[ eÎ_Î ¥¢
;E¸N> Ã ©À
opCDÁµP  q ¾LÂåm Ç ¹H~¶Á´½»æ Å S >~s`_[¬h¨yz]_l h|[Ub,b:dDyzq~sb:jkhOyz]¬cn_brhú~`_[<hzjO~sh?aAd_j
¨]_^Uf
hk[mdDh!h|[Ubeyzb,aAd_j!d"dD~sf<h?oU^vh?^UfU~ͨ]_yzl 5' yd:Ë`_q]DaѨyz]_l dDqq>]§_bey?hk[UbHjk
w>d_j?a\beqsqÊd_jYhk[Ub
~jzcyzbehkbãjk]_^Uycb:jeWסfLjk]D¨h05Ê yd¬jew<a\b[md§_bXfU]Da qb:dDyzfUb
hk[mdDhXdYq]_h>]Dh|[U~j'c]_l¤b:j«¨yz]_l[U]_hA`ud_j>~sf
h|[Ub}UqdDfvbX]DÝhk[Uv b &"~sq¤©d_wuoU^vh\~f,[mdDy
5' yd:<j~hAjkbebelpj\hk]?ob.~jk]_hkyz]_}U~cDwudDf
jk]Y~j}Uyzb:jk^UlpdDoUqs
brÒh|ydD`udDqd_ch|~cD v]_y.lpdDf<"_b:dDyjXhk[Ubeyzb?[d:§_bYombefh¦a\]"yz~s§dDq[<}m]_hk[Ub:jkb:E j Z¤ ~ Ê~shãc]_l!b:j'¨yz]_l d,[U]_h
B8ÿhD
}
$h?B}mbeyz§d
~fU`hk[UbAa>[U]_qbA>fU~s§_beyjkb_wudDf
~s~¨3~sh~jWd_chk^mdDqsqh|[vbÊovq^Uyzyz~sfU`>h|]_`_behk[Ubey
]DþlpdDf<.lpdDf<dD~f<hWjk]_^Uycb:jhk[mdDha\b'jk~l!}Uqs.[d:§_bef3ghyzbjk]_qs§_b
_behܨ]_yWbeÒ<dDl¤}vqb_w
~jkhzdDf¬hçè^md_jzdDyje
× f¸dDf,~f<hkbryzb:jkhk~fv`.beÒ¬dDl!}Uqsb']DÝhk[UbXlY^Uqhk~s aAd§_beqbefU`_hk[!fmdDhk^Uyzb']DÝl!]
beyzfpd_jkhkyz]_}U[<¬jk~crjewèl¤b:dD
jk^Uyzbel¤bef<hzjo<pÌAs]\x¶dDhWl!l0aAd§_beqbefU`_hk[mjjkbebrl1h|]Y[md§_bAyz^Uqsb
]_^vhWhk[Ub'[U]_h`ud_j~
b:d<× 3~shbrÒ~jkhkb
w
~hWjk[v]_^Uq
[d:§_b
~jkhk]_yzh|b
hk[UbÊc]ujkl!~cl!~cyz]aAd§_bomd_cnu`_yz]_^Uf
w:lpdDu~sfU`'~h·fU]_hWç_^v~hkbÊd'}beyk¨b:ch·oUqd_cnu
om]
EEZ\[vb?crdD^jkbO]D«hk[Uv b 5' yd:,omd_cz`_yz]_^Uf
[md_j'hk]¸obdD~sf¬hXd_chk~§_bY`udDqdDÒ~b:j # Öuu$ %ÑZbrqsb:jzc]_}mb_w
ßA]D§_bel?obey eÎ_Î ± (
\Z]T^`_S^a_Sbc]Td egf`hi_cbUjlknmio:bc^pdk8qmBjrdfpd8stbcknmiuvhBwioxdybc^`s=kzhi{|^ahBbc^pmio}~stmiuTkn^`_c^poTwET]TmibcmiodoTdknwi^pd8_M_chBoTox^poTw
b]Td^puudo_cd*kzhBoTwidjlknmiu
: dybm
ic dmikumikndiUhBuuhBkzh
_mBj,Tfbkzh]T^pwi]doTdknwi
hBknddstdd8{:^oTwifp^o:jlknd8iTdo:by\H]Tdknd^`_QoTmm_n_c^T^fp^bjlmikuvhBbcbcdykbcmd]TmibdoTmiTwi]bcmkzhi{:^`hBbd
Tfahis¡ ¢m{:x]Tmibmio_?mBjZ_csn] doTdknwiiU\Z]Td*u=_cbZd=widoTdkzhBbcd8{*^po{|^p£^a{:hBfpf*mibc]Tdkrxknmstd8_n_cd¤_}
^po:£imif£^oTwMb]xd,kndfd¤hi_cd¥mBjo:stfd8hBkZmik?wikzh
£^b¦hBbc^pmiohBfdoTdknwi§
¨ bkzhBo_c^bc^pmio_db©Zddoo:stfd¤hBkUª«oTmib<dyfd8stbknmio6¬doTdknwifd£idfa_
¨ hBknb^`stfpdB®¯hBo:bc^phBknbc^`szfdrhBoxoT^p]x^pf`hBb^mio
¨ {:d8syh
mBjdyfdudyo:bnhBkn=hBknbc^`szfd8_°:kzhi{:^mhiszbc^£id,{:d8sh8
¨ hisystdfdykzhBbc^miomBjsn]hBknwid8{EhBknb^`stfpd8_%ª«:Tf`hisn ]Tmifd¤_t±B¬
m _cb=mBj%b]xd¤_cd*Tknm:szd8_n_cd8_Mxknm{:stdh²©?^`{:d _cTknd8hi{mBjUT]Tmibcmio³doTdknwi^d¤_µ´miud]TmB©Zd£idykMxknm{:std
{:dt¶oT^bdUwhBuuhB¢kzh8QdoTdknwi^pd8_%ª·e¸f^poxd¤_q¹¬[sn]hBkzhiszbcdkn^a_cbc^`s%mBj_cd8sz^¹¶ºs¥d£idyo:bn_}:di»wT
¨ _cd8st^̶6s%kzhi{:^mhiszbc^£idUd£idyo:bn_}diÍwTbc]TdM{:d8syh
mBjnÎyÏ<f¢} gÐi
iÐ dyM
ÑÒÓ'ÓÔÒÕBÖ×ÒØ Ù³Ú=ÛÚÜJÛÝÖÞ UhBuuhBkzh
_6hBkndbcmmUdoTdknwidb^`sßbm%djlmst_cd8{6}i_cm?bc]Tdknd
^a_<{:^ÌàstTfb^po {:dbcdyknu^oT^oTwMb]xdy^kr{|^pknd8stb^mio1
´st^po:bc^pffahBbc^mioástmiTo:bcdykz_}istkn_cbnhBf`_1mBjuvhBbcdkn^ahBf:_csn]=hi_ß_cm{:^Tu4^pm:{:^a{|d}Bdu^pb{:dbcd8szbnhBTfdâºhi_c]Td8_
mBj1£^`_c^pTfd%fp^wi]:b©?]xdyowhBuuhB¢kzh8:_ãmBj1fpmB©doTdknwiªcä
dS¬hBkndåhB_cmiknd8{M^obc]Tdu æ%^pknd8stbc^pmiohBf
{:dbdknu^pohBb^mio^`_Zmmik<çimi*©Zmio1qÍb?_cdydMhBo:whBuuhB¢kzh
MT^astbcTknd8_
´hBkn ,s¡]hBu=dk~bcdfpd8_nstmid¤_ßwi^£idZjhB^knfpSwimm:{{|^pknd8stb^mio_Zª«^¹jºimishBoshBffh<jld©È{:dwikndd8_ãwimm:{¬
jlmikr]x^pwi]Tdkn¢doTdknwi=whBuuhB¢kzh8:_[è<^pwi] £imifbnhBwid¤_<hBkndáhBxTfp^d8{bcm^o:bdknfd¤h
£id8{udb¦hBfTfahBbcd8_<^poFhQm¯
¶fpfd8{*©?^b]hBknwimioFhBo{*oTdmioFwhi_%é¢ostmiu^oTwwhBuuhB¢kzh
_Tknm:{:std,dfd8szbcknmio6®8m_c^pbcknmio hB^kz_Z©<]T^as¡]
bkzh
£idf6b]Tknmixwi]b]xd,whi_<hBo{shB_cdS_chBkn _}©?]T^`sn] hBkndUknd8stmikz{:d8{6}db©Zddobc]TdSTf`hBbd8_
hBbum_cT]Tdkndi}Bo:stfpd8hBkãknd8hiszbc^mio_ãknd8hi{:^fpSwidoTdkzhBbd<uvhBo:SoTd©hBknbc^astfd¤_Hª_cs¡]hi_ãuQTmio_}B_cu=mif^püd8{
K ç]Td8h
£*£idykz_c^mio_HmBj[dfpd8stbknmio_t¬¦}©?]x^as¡]FhBknd=hBfa_cmkndfahBbc^£^`_cb^`sBU\Z]Td8_cdS^pobTknoFstknd8hBbcdSjlTknb]xdykU_cdbn_
mBj[hBknbc^astfd¤_Z\H]Td,knd8_cTfbrhBb<wiknmiTo{Efpd£idf1^`_%h_c]Tm¯©Zdk<mBjbc]Tmi_nhBo{T_ZmBjhBknbc^astfd8_}ThBfpf jlknmiuÔbc]TdSmioTd
^poT^bc^ahBf1stm_cu^asåkzh8i
Ï?knkzh
_QmBjMhBknbc^astfdF{:dbd8stbmikz_EhBkndF_cd8{bcm {:dbcd8szb*bc]Td8_cdFhB^pk_c]xm¯©Zdkz_}duxfpmB^oTw_c^pu^fahBk
bd8sn]xoTmifpmiwibmb]hBbM_cd8{^poÁðr¢kzh
FhBo{whBuuhB¢kzh8F_nhBbcdfpf^pbcd8_ \Z]TdmioTdhBb=è%h
£idkzhB]ïßhBkn FoTd8hBk
6dd8{T_%szmB£idkz_ Â_nihBkndU ^pfmiudbknd8_Z´dyd´ , L/\dyfd8_nstmidi}6´dyxbdu=dk i Ê:5*b]xd,o:TuQdkz_ßmBj
hBknb^`stfpd8_{:dbd8stbd8{6}ûb]TddoTdknwi%mBj|bc]TdZ^oT^b^`hBfid£idyo:bãsyhBoSd[d8_cbc^puhBbcd¤{ ~\H]Td hB^kn_c]TmB©ZdykT]TdoTmiudoTmio
^a_ãhBf`_cmrxknm{:std8{S:UwhBuuhBkzh
%T]Tmibcmio_°8bc]Td8_cd?shBo=]h
£iddoTdknwi^d¤_ã_c^u^pf`hBk1bcmUbc]Tm_cdZmBjszm_cu^`skzh
_}
hBo{Qb]Td^kZMd 0ºd8stbZmiobc]Td%TTdykZhBbcum_cx]Tdknd?^`_[_c^u^pf`hBk8~è<mB©Zd£idyk8}ûb]TdS{:db¦hB^fpd8{Qu^pbxkndrmBj1hBknb^`stfpd8_
{:dbd8stbd8{*^o b]xd,_c]TmB©Zdyk?shBo {:^`_nszkn^u^pohBbd%dyb·©Zdydo*whBuuvhB¢kzh
=d£ido:b¦_<hBo{stm_cu^`sz¢kzh
Md£idyo:bn_
rknmiTo{:¢hi_cd8{SwhBuuhB¢kzh
N L stm_cu^`sz¢kzh
Ubcdyfd8_nstmid8_ZshBohBf`_cmá{:dbcd8szb<Ü
J õOJQPSRQGUT/õûò4VD¢ò6Õ
úDWGP?§bc]T^a_ß^`_bc]TdrhBohBfpmiwiTd<mBjºbc]TdSeÌ_cmioT^asmmiu qBjlknmiuhBknb^`stfpd8_ßbkzh
£idfpf^poxw?jhi_cbcdkZb]hBobc]Td%_cdd8{QmBj
fp^wi]:b~^ohB^k8}BTknm{|st^poTwU{:dbcd8szbnhBTfdZâºhi_c]xd¤_mBjf^wi]bã\Z]T^`_ã^o:jlmiknuhBb^mioM^a__cdtjlTfd8syhB_cdZ^pbwi^£id8_ãbc]Td
{:^pknd8stbc^pmio*mBj~bc]TdU^ostmiu^poTwMhBknb^`stfpd=ª«:bkn^`hBoTwiTf`hBb^mioç_c^poxw=hMo:Tu=dkZmBj~bcdfpd8_nstmid8_<_cTknd8hi{miTb
m¯£idkUhQ©?^`{:d=hBknd8h¬¦?\H]TdMum_cbrhBuQT^pbc^mi_Z_csn] bcdfpd8_nstmidM^a_% Ï 6¥U5 þ S}xoTm¯©ËTo{:dkrstmio_cbcknstbc^pmio ^o
]T^fpdi§ãbc]T^`_[hB^u_ãbm=_cbc{: c_cbb]xdrum_cbdoTdknwidbc^as¥ª
ic%dS¬dy£ido:bn_}i]TmiT^oTwUbcm=_cmifp£id<b]xdruQ_cbcdkn
mBjbc]Td^k?mikn^wi^po1
]zy5_
Ú {ÛÖ×ÒÑ| Ò jåÒÜJÛêBÜ whBuuvhB¢kzh
_jlknmiuXhistb^£idUwhBf`hiszbc^`s%o:stfpd^hBo{
},´ _þH%þ\
]hi_Qknd¤stmikz{:d8{Á]T^wi]T¢dyoxdyknwiwhBuuhB¢kzh8:_¥jlmiknu ho:Tu=dk=mBjá{|^a_cbnhBo:bMwhBfahB^pd8_ oTdb}S´ du^pbn_
um_cbZmBj^b¦_<doTdknwi^obc]TdSð%kzh
T®8whBuuhB¢kzh8Mkndwi^mio1[ÏËwi^`hBobZdffp^Tbc^ashBf6whBf`hBdu^bbc^poxw
t¢d
whBuuhSkzh
_H]hi_ZddyoF{|dybcd8stbd8{6°xumiknd%d£^`{:dostd¥jlmik?^o:jhBfpf6mio:bcmhQhis¡ ]xmifpdB±
ÂiÂ
é¢oFmikz{:dkUbcm*uvhB idMbc]Tdstkn^pbc^`syhBfãfd8hB×mBjZ©Zmikn ^oTwmiTbU©<]hBb,_cmiknb%mBj?mit cd8stbn_%b]xd=whBuuhBkzh
TTkz_cbcdykz_MhBkndi}hBomiTbc^ashBf?stmiTo:bdknhBknbSoTdd8{T_MbcmFdEjlmiTo{6}hBo{Ábc]T^a_^a_Q{:^¹àEstTfbSbmhisn]x^pd£id©?^b]
b]Td fmB©?hisszxkzhiszÁwhBuuhB¢kzh
×m_c^b^mio_ô\Z]xdknd8hBf%Tknd8hB bc]TknmiTwi]©?hi_bc]TdFé¢bnhBfp^`hB o dTm:´Ï<ð
_nhBbdffp^bdi\Z]T^`_Ustmio:b¦hB^o_Uh ç Âi
i
E id.ð%¢kzh8FshBudkzhbc]hBbU^`_,hBTfd=bcm*dQmi^o:bd8{kzhBT^`{:fpE^pobc]Td
widyoxdykzhBfZ{:^knd8szbc^mio×mBjrh²whBuuvhB¢kzh
xTkz_cby*é«jrh*bckzhBo_c^pdo:bUð%¢kzh8_cmiTkzstd=^`_¥jlmiTo{6}bc]T^a_S^`_MhBfum_cb
szdknbnhB^poxfphi_n_cm:st^ahBbcd8{©<^pbc] b]xd,Txkz_cby°:bc]TdMð%¢kzh8m_c^pbc^pmio*^`_?wimm{*bcmdbcbdkUbc]hBo hBkzstu^oxbdi}T_cm
b]hBb=hBoÁmiTb^`shBf[^uhBwidsyhBoÁdbnhB ido1 é¢obc]T^`_S©?h8i}ãbc]TdmiTb^`shBf stmiTo:bcdknhBknb¦_¥mBj?uhBo:FwhBuuhB
kzh8*TTkz_cbn_<]h8£idMddoFjlmixo{6}TknmB£^oTwb]hBb¥b]TdQTTkz_cbn_<b¦hB idQTfahistdM^o jhB^o:bUwhBfahB^pd8_UhBb¥£idykn*]T^wi]
knd¤{x_c]T^Ìjlbn_í\H]Td TTkz_cbn_=hBknd*b]xdykndtjlmiknd*b]xd um_c b
z.doTdknwidb^`s*d£idobn_Q^pobc]Td Tox^p£idkz_cdi§*b]Td
uvhBohBwidMbm kndfd¤hi_cdQknmiTwi]Tfbc]Td_nhBudQdoTdknwiFhi_,h*_cTdyknoxm¯£¯h²ª
)z ¬n}1TxbUb]xdy xTbU^báhBff^obcm
kzhi{:^ahBbc^mio1~Ï knd¤stdo:bTTkz_cbjlknmiu hSwhBf`hBhBb[knd8{T_c]x^Ìjlb d8hB id8{hBb bc]*uvhBwiox^pbc{:dçhBfum_cb[Tkn^wi]:b
dyoxmiTwi]jlmik=bc]Td*ohB id8{didi}jlknmiuPbc]Td*mib]Tdk_c^a{:dmBjUbc]TdToT^£idkz_cdi 'd_c]Tmixfa{ iddmiTkM¶oxwidykz_
szknm_n_cd8{bc]hBb<mioTdá{|md8_co1qÍbZwim=*m 0×^o*b]Td ^f g 'h8ÉyÉÉ
\Z]Td<jh
£imiTknd8{=bc]Tdmikndb^`shBf^`{:d8h¥jlmikZ]TmB©íwhBuuhB¢kzh8STTkz_cbn_[shBof^dkzhBbcdå_cmMuQsn]doTdknwi=^`_
£^`hb]xdMudknwidk%mBj[b©ZmoxdyxbknmioF_cbnhBkz_rÏ%_%^b%]hBTdo_}_cs¡]FhBoFd£ido:br©Zmixfa{*hBfa_cmEd=hBo dM06d¤stbc^p£id
_cmiTkzstdMmBj wikzh
£^b*©?h8£id8_Mª¢_cdd=dyfmB©U¬n}~_cmbc]Tdknd=^`_U_cmiud=sn]hBostd=bc]hBbUbc]T^a_åb]Tdmikn uh
widbSbd8_cbcd¤{
mioTd,{Th
i
oTmik,h*knd£^pd©.mBj?bc]Td miuTbcmioåhBuuhrh
_cdkn£¯hBbmiknq¸_¥©Zmikn }ã_cdd*´ :L \dyfd8_nstmidi}
æ d8szduQdk i Â:\Z]Td^uhistb6mBjdTm:´Ï<ðmioM_cmifp£^poTwZbc]TdwhBuuvhB¢kzh
[xTkz_cbuQ_cbcdknr^`_6{:d8_nstkn^pd8{
%
^po´ L\1dfd8_nszmidi}oxdTknhBkn i Ð:
ZQ[]\=^`_ acbedgfh_ji:ak[lZ
monqpYrtsquFvBuxw y{z|vB}~yovBu:zvBuz{nqp
}p
uqp
}|szqsz?:u:sxp8vB}Hsxy{uFvBz<znp}Qpu:z{}|p8y
vE~q
vBz{p8w|nxvBu Bu:sxp8vB}%}|p8vzuxy
q<z{nFz{nqpYpuw*}|py{sz%z{nxvBzQ./../ <p¡£¢q}|z{uxy
¤¥vB}|z¦pB§|¨Qmonq¦y<}|pvz{ue}|p
p8vy{p8y%vuxy{©wªp}~vBp,vBsqu:z<Bp
uqp
}|«¡£u:sxp8vB}%}|p8v~z{uxy%vBu
xp
}~vBz{p=vp
kz{nqsxy|vBuxwªznxyHBz{nqpY}|py{z{¬vy|yop
uqp
}| Bz{nqp=uqz¦vBxvB}|z¦p8yH®F¯°²±*³
´
§6sqzQv
tµwJ}~vzu Bz{nq©y<sqz{qsqz%¦y<u*¥vzQ
vB}|p8w v
?v8:upsz}|uqty
:p
p
p
u:zvB}|vB}|z¦p8y<?zneuq
|nxvB}|povBuxw¶q}|xvBq%uqQvy|y
·8z{nqp
,os©w¶p8y|vBp}|p
p
¶}|¸z{nqp%rtsqu1By{u~pvBzz{p
}¦y1}~v~z{¦vB
z}~vBuxy{vB}|pu:zzuqp
sqz{}|uqty
¨UmHnq¦y?}|p8vBzp8y?vYq}|p
/}?z{nqp
}%w:p
zp8z{u1¨
rtp
¹p}~vBow:p
z{pz{}~y
oy{pzMsºz{»}=z{nqp8y{p¼t½¾¬¿ÀkÁÂÃÄ:ÀqÅ{Á½¼FvB}|puÆp}~vBz{uqqu
Ç r:È,ǶrqrµÉÊqvBuxwE˵vBxvBu1¨mHnqpÊuqp,vBz?z{nqpÊxz{z{ÌBz{nqp,<py{z|vB»pÊÍuqpQuertsqz{nJÎQvB»z|vYnxvy
Q
xp
p
u}|squququq=B¹p
} Ï p8vB}~y
<uvFÐ
ÑÑ:ÒÑÑÑB¬tvBu*z|vBuq» BÓ ÓÔ¡¬w:}|tÕp8vBuquqYÖxsq¦w§|6vBxy{}|xp8w
´
uqp
sqz}|uqty¹p
}|Yµ
vy{uxvB=~u:¹p
}|zvM×nq}|uqp?vBz{Ì¡ØÓ£§Uz{=vB}|u ¡¥È<}
§|:?nq¦|n
vBuxpQw:p
z{p8~z{p8w6¨
Ù,¨ÛÚ%¨Î%p
z{pz|vBqpuqp
sqz{}|uq*ÖsqÜtp8yMvB}|p¹p
}|«y{vB¬Mz{nqp rµÙ%ÇA¡¬y{©vB}Muqp
sqz{}|uq squqz~§%¦yMw:pÝxuqp8wÞvy
Ð
Ñ:ßxà|áuqp
sqz{}|uâ
vBqz{sq}|p8yxp
}z|vB}|pzvBz{kxp
}y{p8uxw6vBuxwÊz{nqpoxy{p
}|¹p8w%ÖxsÜSvBzz{nqp%rtsqz{n=ÎQvB»z|v
y{z{p,¦y1ã{sy{zozÕo=Bz{nqp8y{p,squz|y
¨
rtu~pÊz{nq©y<quqp
p
}|u=o}|»xznp}¶pÜxp
}|pu:z|y<Bz{nq¦yQy{}|z?nv8¹pMy{zsw:p8w*uqp
sqz{}|uqty}|
znqprsqu1¨¸monpäâÈ%å1åæçèuqp
sqz{}|uqÞw:p
z{p8~z{}<z{nqulvBujé¬z|vB¦vBujsqu:z|vBuhsxy{p8yYznpÞäQv] ä%p
~u:¹p
}~y{u1¨r:ÇêæÉëÊÈ%ìFé{íQëÊÈÙ%ÎQæh¦y?vBu*squxwªp}|}|suxw=
v
¹p}|u*uË:vBxvBuFîïÐ
ÑÑðÌv}|ty|yñÝxp8w
?zn?vBz{p}ovBuxwYuqp8wð?znqnqz{=sz{p
}Uz{sqp8y?¡Ûã{y{psq}|zÕÊ}}|p8y{pvB}~×n=y{zsw:p
uµz|y§|·?zny{sx×n
v ~ty|y|vBuµs=p}¶B??vBzp
}Mp8sqp8y¡Ð
Ñà|ò
§|vEp
}~vB}|prt¦vB}Muqp
sqz{}|uqty¶w: u:z{p
}~v~z¡£zp
uxyðB
p¹p
u:z|yMp}wqv
q§|¨Þmonqp}|p8vz{uÞ¦yMu:¹p
}~y{pÍp
zvw:p8
v
xMóE¸ôº õö¸÷øo¨ÞmHnqpty{z{}|uÞz|vB»p8y
sq]znpFuqp
sqz{}|uFpu:z{sq ovBuxw]y{s:ù6p}~yY}|p
¦vBz¹t¦y{z¦*}|p8¬·?z{nq¦yvBsy{pyúzz{Þp
Íz*Óp
}|p
uq»B¹
}~vw:©vBz{u1t<nq¦|n vBu*pSw:p
z{p8~z{p8w u*znp,?vB©yBz{nqpMv
¹p
}|u1qvBuxwEsxy{p8wz{Í}|p8uxy{z{}|sxzoxz{n z{nqp
w:}|p8z{ueBznpMuqz{©vBuqp
sqz{}|uúvBuw zy%p
uqp
}|¡p8
vBsxy{pMznpÓp
}|p
uq»B¹uqp=vBuqp=w:p
xp
uxwy%u
znqp%¹p
µzØÍB1z{nqpQty{z}|u6§|¨Î%p8y{uqp8wÍvyvBup
Ütp
}|p
u:zz{=w:p
z{pzoq}|z{uw:p8
v8z|yn:sqp?vy|y
úvB»p8y%zÊznqpYty{z,y{p
uxy{z{¹pûÒuqp
sqz{}|uqz{pp8y|xpü6uznqpYo}|©wý¨=È%ìFÈ%Ù%ÎÈ/<xpvEuqp
k»tuxw
Bw:p
zp8z{}?}%nqnq¬p
uqp
}|uqp
sqz}|uqty%¡þ}|Ì©v×»nqpyHp
zB¨ÿ§<uqp
sqz{}|uqtyp
uqp
z}~vBz{uqÈ?u:z|vB}~z¦
©pQ<xp8v8¹p%z{}~vB¦yñBÓp
}|p
uq»B¹M}~vw:¦vBzu1z{=p,w:p
z{pz{p8w:=nqzuw:p
z{p8~z{}~yosq}|p8wðuz{nqpQ¦p
rt» m1p
p8y|~p6ËtsÞÐ B¢ ¥¨MÈ%Ù%mÈÉ%æHr¦y,vEy{©vB}Qy{}|z%B¦w:p8v:sxy{uqy{p
¹p
}~vBUsqq¦M»'B
Èoz{©vBu:z{©%?vBz{p
}%BùFznqpMtvy{z?B q}~vBuxp¨
monqpQu:sqYxp
}oBy{©vB}<uqp
sqz}|uqtyñsquw¦youq Ñ .Bznp,u:sqYxp
}oq}|p8w:¦zp8w:y{z|vBuxwqvB}~w
{y ©vB}µw:p
¦y
¨UéØyzn©yp8
vBsxy{poznp?vy{z}|uq¦
vBtµw:p
¦yvB}|po?}|uqqB}xp8
vBsxy{posq}ñsquxw:p
}~y{z|vBuxw:uq
BâvB}|z¦p*qn:µy{¦
yÊ©yY?}|u Ìmonq¦y=w:p
xvBz{p*}|pvBuxyÊsquq}|p8y{¹p8w6qszMux}|p8vy{uqe¥v
¹sq}~yÊz{nqp
©vBz{zp
}Êxty|y{qzÕ¨MÙ<psz}|uqtyÊ~pYueznq}|p
pEû Ö v
¹sq}~y
üqp
pz{}|u1=squ16vBuxwözvBsuqp
sqz{}|uqty
¨MÈ?u
u:z{}|squqty|y{qzÕ ©yÊz{nxvBzMuqp
sqz{}|uqty%v8 û Ö1üýxp
zÕop
p
uÞz{nqpw:ÿù p
}|p
u:zMû Ö v8¹s}~y
ü1y{ z{nxvBzÊop
vB}|pQ»tuqÊ}?z{nqpQ<}|uqÊ»tuxw:Yznqp¶zpQznpvB}|}|¹pQvBz?ævB}|z{n1¨<Æop
¹p
}oxvB}|z{©p%znqp
}|y|v
µy
znxvBz¶znq¦y<?uq*nxvBqpu ÿuqp
sqz{}|uqty<u*¥vzQnxv
¹pMv¹p
}|Jy{vB1qsqz%uqu¥¹BvBuq©y{nuqYvy|y
·qz{nqp
nqnp}Êz{nqpúvy|yÊw:ÿù p
}|p
uxpxp
zÕop
p
uºw:ÿù6p}|p
u:zQuqp
sqz{}|uqJy{p8p8y
znqpEÍ}|pY}~vBq¦wFz{nqpty|¦vBzu
xp
zÕop
p
u zÕtp8y
¨oé¬u }~w:p
}?z{pÜq¦vBu*znpSy{¦vB}?uqp
sqz{}|uq=}|p8y{sqzy
:vy|y{p8yBvy{úvB}~vzu*BvBu
p 4¡p
uqp
}|Msquz|yvB}|posxy{p8wÊnqp
}|p1zp8vBuxy±*³
´<°²Ðp ʧvB}|pouqp
p8w:p8w6¨é¬z¦yty|y{qpznxvBzñznqpÊ¡¥y{â¥vB}
¢
Z%dFbe_ga²[ ^ acbedgfh_ji:ak[lZ
p
q}|sxvB}|ÍÐ
QvQy{sqp
}|uqƹBvM¡Ð
Æȶ§1p
Üq:w:p8wYuYznqp%å1vB}|p%ìFvBp
¦vBuq¦<Ósxw6vÊ~xvBu
íu
u*Bsq}QäQvB©vBܨ?rYsqz|vBuqp
sxy{tvBuxwExp}|pQz{nqpMy{sqxp
}|uqB¹Ævð?vy?y{p
p
uqz{¦vB:z{nqpMw:p
zp8z{}~y
uºÇ¶rµÈOvBuxw˵vBvBu«y{p}|¹p8weuqp
sqz{}|uJy{nqBop
}~yQB¹p
}Svöxp
}|µwBSÐ
Ñey{p8~uwqy
¨<p
}|p?vyÊw:}|p8z
~uªÝx}|vBzuBUu:sxp8vB}oq}|µp8y|y{p8yu v=
vBz|vy{z}|qn©
vBYpÜqµw:uqÍy{z|vB}8¨monp,y{zvB}8üÛyop
uµz{}~vB6z{p
xp
}~vBz{sq}|pJvBzYznvBzÍp
u:zM©yYp8y{zvBzp8w z nxv8¹pExp
p
u
Ð
Ñ ëY¨monp}|p
pvy{p8w p
uqp
}|Þ=sxy{z
nxv8¹pÊpp
u
Ô
Ð
Ñ áÊËsqp8y
?nq¦|n ¦y%vBxsqz=Ð
Ñ Êz{p8y<znprtsqu1üÛyosqz{qsqz%uºÐ
Ñy{p8uxwqy
¨?mHnqp
Ô
uqp
sqz}|uqtyo
vB}|}|Ív
?v
ÍvBÍty{zovBxBznqp¶pup}|:p8v8¹uqMup}|nvBxyÐâxvB}|zuÞÐ
Ñ z{=Æop
}oz{nqp
Ýx}|p
o}|»µy
¨1æñ¹p
uÍy{qBznpy{sqp
}|uqƹBv?¦v
SvBzvÊw:©y{z|vBuxp¹Ævy{z?z{n=}|p8y{xp8zzÊz{nqp%rtsqu1vBuxwðuqSv<p
w:
pu upsz}|uqty<op
}|pSw:p
z{p8~z{p8w6¨QÙ<y{sqp
}|uqƹBv=nvy?p
p
u«y{p
p
u ?znqu z{nqp=ì » Þv
}QvBsqz
¢tÑÑ p8vB}~y
1<nqp
}|p8vyÊxy{p
}|¹ÆvBzuxy¶BvBu:FtvB¦vBÜtp8yÊ»psq}ÊÆ<uºy{sqp8y{zÊznxvBzMz{nqp}~vBz{py{nqs©w
xp=vBsqz%uqpMp}Ê~p
u:z{sq}|¨é¬z%v
pMznvBzop=vB}|pÊB¹p}~wªsqpQ}Qvµ
vBy{sqp
}|uqƹBv:·qz,sq©w vB¦y{
!
xpYznxvBzÊuµz{p
}|tvB©vz{©Mwªsxy{z%uz{nqpÍìö»t Þv
Íp8vBuxyâznxvBz¶Íty{zÊy{sqxp
}|uqB¹ÆvBpÊsuqxy{p
}|¹p8w6¨,é¬u
pz{nqp
}?vy{pznpupsz}|uqMy{uxvB}| v¶Í}|p<uqp8vB}|:=y{sqxp
}|uqB¹Æv¶osq¦w=p%¹p
}|Íy{z{}|uqÍ¡{î¸Ð
Ñ |ÑÑÑ "
w:p
zp8zp8w*uqp
sqz{}|uqty§:¹tuqYsxy?vðopvBz{nBu:}|vBzuvBsqzoznpSw:uÍy{p8uxwqyoBvy{zvB}8¨
q}| z{nqpexu:zÍBM¹tp
'B,xvB}|z{©p qn:µy{¦
y
ozE©yy{uqÿÝ
vBu:zMznxvBzEznqpeuqp
sqz{}|uqtyð}|
# $
r ÙMРȸz{}~v8¹p
p8wEznpúÐ Ñ:ÛÑÑÑMn:z{¬pvB}~yñ}|OznqpMå1ì ÓlvBuxw*vB}|}|¹p8wvBz<znpSy|vBpÊzp:uxw:p
µ
xp
uxw:p
u:zoBUp
uqp
}|¨mon©yHz|yznp}?vy|yz{=p
B¸vBxsqz,Ð
Ñðp
¡£znp}|<©y{p%z{nqpÊÍty{zp
uqp
}|p
z¦
uqp
sqz}|uqtyosq¦wnxv
¹pQz{}~v8¹p
p8wú¥vy{zp
}%vBuxwvB}|}|¹p8wYÝx}~y{z§¨
v|nqp
¹pw*:*µw:p
}|u*puuqp
p
}|uqq¨mHnqpÊÍty{zQvBYqz{sxyq¦vBu Bzn©y%»uw*©y<}Qvy{v~pÊq}|Æã{p8z
9$
vBpw*å1érµÈÊq?n©×nosq¦wu:¹¹pQvYuqp
zÕo}|»=B y|vBz{pz{p8you}|qz?vB}|squxwúznqpMy{squ1¨
Bop
¹p
}8Æz©yuzp}|z|vBuÍ<nqp
znp}z{nqpsq}|}|p
u:z}|squxw:¬xvy{p8wÊw:p
zp8z}~yñ?ªy{p
p%vBuµznuqq
?
xp8
vBsxy{pznp}*pÜxp8z{pw`y{puy{z{¹zØj¦yty{pFz{]znqpp
Ütp8zp8w`p~ù6p8z}Fvy{z{}|uq©
vBQy{sq}~p8y
¨
ä}~v
¹tzÕF?v
¹p8y,vB}|pp
z{z{pw Íty{zÊy{z}|uq }|èvy|y{¹pYx:w:p8y¶squxw:p
}|uq*nqnÞv
p
p
}~vBz{u1¨
monqpQp8y{z?opQ
vBuvBupos©wpÊznqpÊ~tvBp8y|pu~p¶BvYxvB}oBxvzoúvy|y{¹pQy{zvB}~yñ®vMxvB}
Buqp
sqz{}|uFy{z|vB}~y?os©w*p=¦w:p8vB¥¨Êrsx|nFvBuep¹p
u:z%osq¦w*tp¦wv|nvB}~v~z{p
}|©y{z{¦Mû |nq}|1ü:Bo}~v8¹zØ
?v8¹p8yvyñznp?quxvB}|Êty{p8yñ}|qz|vB:p
uqp
}|¡z{nq}|sqnYz|y}~v
¹tz|vBzuxvB:}~vw:¦vBzu6§1vBuxwYp}|p8yu:zÊv
q©v×»,np¨monpoxp8vB»=y{uxvB}z{nq¦yy{}|zB p
¹p
u:zosq©wÊ}|xvBqQp
ܵp
pwYz{nqpvBB¹p?y{puy{z{¹z{py
p¹p
uö},vBu p
¹pu:zQvBz,ty{¦
vB1w:¦y{zvBuxp8y
¨¶rtu~pMz{nqpMq}|pp
}|}|p8w*µw:p
6}¶tvBÍvB¬}~v
Ísq}~y{zy
©y%vB¦y{=z{nqpMp}|p
}oBvYuqp
sqz{}|uqÕy{z|vB}oquvB}|:z?¦y<xty|y{qp%zvBup,vYouxw:p
}{sqsqzs}|p,¹©y{u
u*?nq¦|n*z{nqp,tvBvB¬}~v8=vBuw}~v
¹tzÕt¬?v
¹py{uxvB¦yñ}|Ìvy{uqp%p
¹pu:z%~s©wExpÊy{zsxwªp8w6¨
Èy{z{}|uqp
}~yw:u1ü zU»p?z{M?vBz
tvBuxwðÍty{zvB}|p?vB}|p8vw:=u:¹tuxp8wMz{nxvBz}~v
¹tzÕM?v8¹p8ynv8¹p
xp
p
uºw:p
z{pz{p8w6¨FmonpÍ}|p8vy{ue}Mz{nq©yʦyMv y{p¦vB:ã{p8~zM
vBp8wz{nqp Å{Á¿À !: ;=< %Ã?¾¬¼t¿Àq¨é¬u ¥v~z
>
znq¦y¶©yÊvy{uqp=qs©y|vB}%u«}|z,vBsqz,vE?nz{pw:?vB}{6u«v*}~vBz{nqp
}Qzn:zQ}|qz=¡ *nqsq}~y<xp
}|µw§|¨
monqpQp
Ü©y{z{p
uxpQBUz{nqpÊqsq¦y|vB}vyovðuxvBzs}~vB µ|»EÍp8vBuxyñz{nxvBz?op,vB}|pQvBqpQz{=sxy{p%z{nqpMÎ%qqp
}y{nqÿz
z*pvy{s}|pMznqpY}|qz¹p
}|Fv
sq}~vBz{px·?nvBzQ©yâsquxwö©yâznxvBz¶znqp}|qz%©y¶y{xp
p8w:uqsq1y{z{nxvBz
? Ô
znqp B¬nqsq}%xp
}|µwe¦y%}|p8w:sxp8w :FvBxszÍÐ
Ñ:ß y{puxwy%xp
}Qp8vB}¡sq©y|vB}Qvy{z{}|uq=EBzp
uºy{nqBy
znq¦yÊouxw:p
}{sqñq}|p8~¦y{u6§¨*monq¦yÊw:tp8y{u1ü z=y{squxwe=s|n16sqz,z{nqp}|zQ¦yÊty{uq p
uqp
}|vBuxwz{nqp
xy{p
}|¹pwE}~vBz{pM©y<p
ܵvz{ <nxvBz%osq©wEpMq}|p8w:©z{p8wE}|Ì}~v
¹tzvBz{uxvB6}~vw:¦vBzu1¨?mHnqp
}|p~}|pÊz{nqp
u:z{p}{p
}|p
zp
}<pÜxp
}|pu:z|y?
vBu pÊ¥vB}|u:Ý w:p
u:zoz{nxvBz?z{nqpM?v
¹pyHznp vB}|pÊ»tuqY}p
Ü©y{zH®
vBuxweznxvBzMw:p
z{pz{uz{nqp
<pusqz{nqpy{zsw: Bz{nqpty{zQp
Ütz{©vy{z{}|uq©
vBµã{p8zyÊp
z
~u:z{p
Í©vBz{p8w6¨
Z?b fÞi
Èy{z{}|uqY=q}|B¹t¦w:p8yu=~p8vB}?p8y{z{vBz{p%}oznpQq}|xvBqzÕMBznp}%ævB}|z{nq¬»p<q©vBupz|y
p
zvBuqp
!)))
}?z{nqp,p
p
}|p
uxpQBU pu:z{pp
uxptz{p8|nqu @
~vBuxwp
Ütz{uxzu1¨ ñsFvBu*q¦v8tvBp8y?zn
ÎQÉ%È%ë,æ<üÿrúæBAâÇ%Èm<é{íQÙ,DC'°FE-GHõJIKGMLNGMOPGMQK,
R
m nq¦y?y|v
µyHznxvBz<znp,u:sqYxp
}oB¹ 8vBz{uyoop,y{nqs©wEp
Ütpz<zy{p
p,vBz%vBu:uqpQz{pʩyHznp,}~vBz{p
o
B}|p8vBzuBy{zvB}~y
z{p8yznpQ}~vz{uz{nxvBz<nxv8¹p%q¦vBuqp
z|y
tz{p8yoz{nqpQu:sqYxp
}BUq¦vBuqp
z|yxp
}y{z|vB}
znxvBzvB}|p?»p?ævB}|znzp8yznpo}~vzu=B1ævB}|znq¬»pq¦vBuqp
zyuÍ<nq¦|nM p<pz|yy{zvB}|z{p8w6z{p8yz{nqp
R
}~v~z{u Bñ
vy{p8yo<nqp
}|p,ÿp,p8vwqyoz{uµz{p
pu~p:z{p8yoz{nqpÊ}~vz{u*B~¹t 8vBz{uxy<z{nxvBz
vBu p
*
xz{nqp
}|p8wMz{,ÍYsquq¦
vBzpBzp8y1z{nqp<p
uqz{n=Bxz{Ípop8v×n͹ vBz{u=¦vy{zy
¨monqpouÊop
¬»tuqB?u
4E
uµs=p}¦y MµvBz{nqsqnq¦vBuqp
zoy{pvB}~×nqp8yvB}|p%uqBjvB»uqM}|}|py|y?zn !G H ·:znqp
}|püÛyo~p8vB}|vMuq
SGML
?v8ezeq¨Jêp}|nvBxy%z{nqpqp8y{z,squ~p
}|z|vBu:zÕ ¦y y{uxpz{nqp}|uB? p¦yMv*}|p8vBz,Yµy{z{p
}|¨
T
rtpu:Öxspu:z{©vBpp:uqz|vBq }|pwE?Btp:nxv
¹p,vB}|sqp8wYznxvBz<ÿpQ}|uvBzp8yuy{xvp y{ppðrt»t
mp
py|p ìevB}~|nÞÐ B¢vBuxw*ËsqÐ
U ¥6vBuxw*¦yQy{q}|p8vwEznq}|snqsqz?z{nqpMtvB¦vBÜt*:Jp
zy
6y{
?GL
znxvBz %°ÐB¨ÞmHnq©yÊv
Þy{pp
vö¥vB}|£pz||npw ©w:p8v:ñqsqzMopy|v
.z{nxvBz=Íp
z{}|pÍvy{z{}|uq=enxvy
w:p
zp8zp8w]qpÜ ûÒ}|tvBuq¦Bü1p8~sp8y
·znpz{}|sqqp¦y
?npz{nqp
}=u]y{xvp}Yu ævB}|znuq¬uqp
squxw:p
}~y{z|vBuxwqyñnqÆgz{pzH}|ÌvBu=v~¦wqyHz p¨
UV
É?p
z{sq}|uquq?z{MÎ%}~vB»püÒyp sxvBz{u1op?w:u1ü z1p
¹p
u»tuqBÞz{nxvBz 9CXW YV
ïÐBy{u~poz{nqp<p sxvBz{uY¹p8y
znqp Z[]\^Z_Y`Z.a *
u:sq=p
}B?¹ vBz{uxy
¨=mon©y%u:sqYxp
}%n:zQpÍy{ y{vBz{nxvBzQz{nqpq}|xvBqzÕEB
$
* S
p¹p
uupâ¹t 8vBzuuz{nqp%pu:z{}|p%squ¹p
}~y{poosq¦wYxpQ¹p
}|~ty{p%z
p
}|q¨oí<Usq}~y{popQvB}|p%nqp
}|p·
bV
qsqzÊ %opnxvw:u1ü zQp
puuq¬uqpYosq©w xpvy{»tuqFvBu: sqp8y{zuxy
¨Fé züÛyÊxty|y{qp=z{nxvBzMopvB}|pz{nqp
UV
~ty{¦op s¹ÆvBp
u:zB6znp<uuqp
}B6znpâÙ%vBz{uvB åýzz{p
}|:vBuxwYz{nxvBzz{nqp
}|pQvB}|p?uqÊz{nqp
}u:zp
p
uµz
xp
uqtyvBu:?nqp
}|p¨È<z{p
}|uxvBz¹p
znpsuq¹p
}~y{ponµzñxp<sqxB1z{nqp
·tz{=w:p8©wªpop%nxv
¹p?z{M»x¨
V dc 2Ye
é¬ujvBujvBzz{p
qzzºy{p
z{zp z{nqp sqp8y{z{u |éÕyvBu:uqpz{nqp
}|p :oÙ%ȶrµÈ'Ysxy{z{p}|p8why{s p
u:z f
qz©y{/z=zñÍ:w:p8y{zsuxwqyñzYz|y<ûrpvB}~×nM}?æñÜtz{}~vBzp
}|}|p8y{z{}|©vBxé uµz{p
pu~pü¨monqpQÙ%ȶrµÈ<
squxw:p8wM}|}~vBÍpz{p
}|uxvBz{p8wMuFÐ qsqz¦yuµz{uµsuqQvyv¶q}|¹ÆvBz{pt£squw:p8wMp
Ütp
}|p
u:z¡¬y{p
p
znqpMop
vwqw:}|p8y|y?vBz<znp,p
uxw*Bz{nqpÊuqz{py§|¨?mHnqp=r:æm<éq}|}~vBÍp<©y<xvy{p8wEu }~vw:=z{pp8y|xp8y
UV g
vBzy{p
¹p
}~vB y{z{py
uz{nqp<}|p sqp
uxY}~vBuqpSÐ<z{JÐ
Ñðäâ :z{nxvBz
vBuy|vBqp?sÍz{JТÊu|nvBuquqp
©y
y{Ysqz|vBuqp
sxy{¨monpQ¥v
¹sq}|p8w?v
¹p
p
uqz{n }|p
u*¦y?z{nqpûÛ?vBzp
}%nqpüq?np}|p¶znqp=ty{¦,vBuqvB©y
~p%z{MzvB»xz{nqp%}~vBuqpp
u:z}|p8wYu ih+j
²Ð
ÑÊJ?np}|p<tvB©vz{©%vBuxwYz{p
}|}|py{z{}|¦vBq}~vw:Qxv|»}|squxw
©yuqvB¬tty{©
vBû vw:¹p
}|z¦y{p8w6ü:znp}|py{p
uxp%BUy{z{}|uqM}~vw:,upyBÊqíQÊq íS¨:monqp
}|p
´
©y<vMv8ã{}?qsqz{uqM}|qp
unxvBuw:uqMy{sx|n*vMn:sqp,wqvBz|vM}~vBz{p,vBuxw*w:p8©wªuqM<nqp
uvEû y{uxvB¬ü
úv
nxv
¹pxp
p
uÞ}|p8p¹p8w6¨È<uqz{nqp
}M?v8eBy{pvB}~×nquqn:zQpz t»e}Ê?vy{zpéÕÉ/}~vw:¦vBz{u
}|ÌvBp
uxy
ü:~ty|y|vB1vy{z{}|¥p
uquqp
p
}|u,q}|Bã{pz|y
¨rtp
p=r»t kmp
p8y|p,Î%p8pYxp
}MÐ :¨
!V 1k
È?uFv
?»?vB}~w sqp8y{z{u ¦y<uÍz{*xp?nqp
urµæm?énvyy{p
u:z µÐ
ÑÑ:ÛÑÑÑ:ÛÑÑÑ:qvBuxw*z
üÛy%uqB
l
ÈÎ ÑÑ qvBuxwz{nqp
}|püÒyuMxty{z¹pQ}|p8y{sqz®?nxvBzuqp
Ütz Jêp}|nvBxyåe¦y¹p
}|úy{vB m)))
| zp8×nquq
vB?v
µyHp8vwqyz{y{pÿÕw:p8y{z}|s~z{u )))
×?nqY»tuqBy l
n*d¸_Jfpocbe_ i:an[ F
_ q ^FfÞi[ a
é¬uvwqw:z{uz{MznqpQvB}|z{©p8yÍp
u:z{uqp8wðuz{nqp%z{pÜz:z{nqp<B?uqÊx»µyv8Yxp%nqp
:sqqÿ1svB}|p
u:z{p}|p8y{z{p8wuÆ<uqMsJvBu:BznqpÊz{q©
yHu*}|p,w:p
zvB¬
r
ÚoÇÉ%ëÊæ /äQÉ%È%È%ì {rµìeéÕm?ÊqÈ<u uµz{}|µwªsxzuz{Í}~vwª=vy{z{}|uq=*¡ÕÓ%¨ÒÇʨÛê¨ §
Ó%È%Éåæor /r:æ FÈ%É%ÎMxæñÜt}|uqMz{nqp,ç%¬}~v
Ísuq¹p
}~y{pM¡ØÓ¨ÛÇ,¨Ûê¨ÿ§
s
åíQÙQäQÈ%éÕÉÊxmHnqp,Ù<p
ïÈ%y{z}|qn::y{©
y%¡£uFÎ%È QéÕæHrF¡£pwý¨ §MmonqpMÙ<p4ên:µy{¦
y
qÓ%¨ÒÇʨÛê¨ §
s
rµìFéÕm<ÊíâÚ<rµæÉ Èm?éíâÙ%È%åºÈ¶rµm<É%íQê Êrµé{Ó<rJ¡ØÓ¨ÛÇʨÒê¨ÿ§ g@
4t . !
m?éÕìeæ Æåé æ<qmonqp,Ù<p
ïÈ%y{z{}|uq=*¡u BtvBpQzn}|sqnz{nqpSÇ<uq¹p
}~y{p,y{p
}|p8y§
monqp
}|pÊ©y%vB¦y{úvúz<Bu:z{p
}|py{z{uq=u:}|vBzu¡¥vBuxw y{p8zvsq¦vB}%q©z{sq}|p8y§v
¹ÆvB©vBqpÊu znqpMop
1¨
@s v8Y»p,z{p
Ütq}|p,y{pQBUz{nqp¶B?uq=vwqw:}|p8y|y{p8y
nµz{z{1 utt8<?ʨ xvB¨Ò
vBzp8×n1¨ p8w:s
nµz{z{1 utt8<?ʨÒy|wqB¨Ûvy{¥¨ z
nµz{z{1 uttÆy{BÝ6v:¨ÒvB}~B¨ uxvy|v:¨
¹
nµz{z{1 ut8t ?v
¹p¨ Ü}~v8¨ xp¨ ͵¨Ûw:p2t8}ty|vBz
nµz{z{1 utÆt ×nxvBuxw:}~v:¨ nxvB}|¹ÆvB}|w6¨ p8w:s
nµz{z{1 utÆt vy{z{}|q¨ py{z{p8B¨ p8y|v:¨ uq
nµz{z{1 ut8t <?ʨ ã{1¨ vBu1¨Ûvƨ sq»v8t Íp
}|u
nµz{z{1 ut8t <?ʨÒvy{z
¨Ò
vB ¨Ûvƨ sq»vt äp
u
nµz{z{1 ut8t u:q¨ÛvBµB¨ uq}~vBq¨ p8w:s
nµz{z{1 ut8t <?< y{»¨ ¦}|}8¨ s¥z{»tq¨ÛvB¨ ã{6Æt wt:wtÆy{»
nµz{z{1 ut8t <?ʨ q¨Û
vBz{p8|n1¨ p8w:s
nµz{z{1 ut8t <?ʨ p
x$ ÑÑ:¨ squq¥nvBuqu:ƹp
}8¨ w:p
nµz{z{1ut8t <?ʨÒy{p
z{¬uxy{z¨ p8w:ssÆt y|pu~p28t qnqq¨ ntz{
IVER
UN S
IT
TH
Y
H
O F
G
E
R
DI
N B U
Problem set 1
where Sk (r) = sin r (k = +1), r (k = 0), or sinh r (k = −1) and dψ is the angular
separation between the two events under consideration.
(a) Derive the radial equation of motion for a photon. Explain what is meant by
the terms ‘particle horizon’ (= ‘horizon’, when used without qualification) and ‘event
horizon’, and use the Friedmann equation to write these as integrals over redshift.
(b) Calculate the horizon size as a function of redshift for a flat universe
containing matter and radiation; the integral is easier if you change variable from
z to a(z) = 1/(1 + z). Show that the horizon size at matter-radiation equality is
(16.0/Ωm h2 ) Mpc.
(c) Given that the current horizon radius in a flat vacuum-dominated model is
approximately (2c/H0 )Ωm−0.4
, calculate the angle subtended today by the horizon at last
scattering (z = 1100).
(d) Write down the integral for the event horizon in a flat universe containing
matter and vacuum energy; argue that in general the integral converges. For pure de
Sitter space, show that the event horizon with respect to a = 0 diverges in comoving
length units, but that the event horizon with respect to events at a time t always takes
the same value in proper length units.
dzobs
= (1 + z)H0 − H(z).
dtobs
For flat vacuum-dominated models, discuss the sign and magnitude of the effect. Apart
from sheer instrumental precision, why might this be hard to detect?
(4) Consider a neutrino with rest mass m 100 eV, which decouples when it is
relativistic. Such a particle can be reconciled with observation if it decays to relativistic
products (assume that these are massless neutrinos). If we approximate the decay as
occurring instantaneously at temperature Td , find the relation between m and Td such
that the universe is still matter-dominated today. Show that the massive neutrino would
have dominated the energy density of the universe at the time of decay (at which point
it was nonrelativistic), so that there would be two epochs of matter-radiation equality
in such a model. Use order-of-magnitude arguments.
IVER
UN S
IT
TH
Y
H
O F
G
E
R
DI
N B U
Problem set 2
(1)
(a) Consider inflation driven by a single real scalar field, φ. Write down the exact
equations for the time dependence of φ, and for the time dependence of the scale factor,
a, in a universe containing only the inflaton.
(b) Explain how this equation of motion in the case of a homogeneous scalar field can
be solved in the slow-roll approximation.
(c) For a potential V (φ) ∝ φα , show that the solution to the slow-roll equation is
1/β
φ/φinitial = (1 − t/tfinal ) ,
where β = 2 − α/2. Explain why this equation does not hold near φ = 0.
(d) Assuming curvature to be negligible, verify that a homogeneous scalar field with
potential s !
16π
V (φ) ∝ exp φ
pm2P
leads to inflation with a scale-factor dependence a(t) ∝ tp (it helps to realise that this
requires V and φ̇2 to be in a fixed ratio, so that neither dominates).
(e) Evaluate the slow-roll parameters ǫ and η for this model, and show that the slow-roll
approximation should be valid provided p ≫ 1.
V3
2 128π
δH = ,
3 m6P V ′2
where V ′ stands for dV /dφ.
d
ln δH2 ≡ n − 1 = 2η − 6ǫ.
d ln k
IVER
UN S
IT
TH
Y
H
O F
G
E
R
DI
N B U
Problem set 3
ȧ
δ̈ + 2 δ̇ = 4πGρm δ,
a
where all other matter contributions are assumed to be uniform, and appear only in
H = ȧ/a on the lhs.
(a) Consider a flat matter-only universe. Use the Friedmann equation to show that
4πGρm = (3/2)a−3 H02 = 2/(3t2 ).
(b) Hence, by looking for power-law solutions, show that δ ∝ t2/3 or t−1 .
(c) At a time t0 , the universe is homogeneous in density, but is given a non-zero comoving
peculiar velocity field, u(x). Calculate the linear density perturbation at an arbitrary
later time, and show that it tends to δ = −(3/5)∇ ∇ · u(t/t0 )2/3 .
(d) Change variables from time to scale factor, a. Denoting δ ′ ≡ dδ/da etc., show that
δ̇ = ȧδ ′ and δ̈ = ȧ2 δ ′′ + äδ ′ , and hence that
(e) Use the Friedmann equation in the form ä = −4πGa(ρ + 3p/c2 )/3, and define
y ≡ ρm /ρr for a matter + radiation universe. Hence show that in this case the growth
equation becomes
2 + 3y ′ 3
δ ′′ + δ − δ = 0;
2y(1 + y) 2y(1 + y)
verify the growing solution δ ∝ y + 2/3 and show that δ ∝ − ln y is a decaying solution
in the y ≪ 1 radiation-dominated regime.
so that the total growth between horizon entry and matter-radiation equality is
Hence argue that the small-scale CDM transfer function should grow as ln k.
(2) The Zeldovich approximation states that the proper coordinate of a given particle
can be written as x(t) = a(t) (q + g(t)f (q)), where q is a Lagrangian coordinate, f is a
comoving displacement field, and g(t) is the linear growth law for density perturbations.
(a) Argue that the density contrast, 1 + δ, is given by the Jacobian determinant
of the transformation between q and x/a.
(d) Now consider collapse of a uniform spherical overdensity: show that collapse
to infinite density is predicted at δlin = 3 and compare with the exact answer.
IVER
UN S
IT
TH
Y
H
O F
G
E
R
DI
N B U
Problem set 4
(1) (a) The cosmological proper number density of a given particle species, n, is given
by the Boltzmann equation:
ṅ + 3Hn = −hσvin2 + S.
Explain the meaning of the terms in this equation, and justify its form.
where Γ = nhσvi and NT is the comoving number density that would apply in thermal
equilibrium.
(2) (a) When the universe cools to T << 1000 K, the material becomes primarily
neutral. At late times, the remaining free electrons are removed at a rate determined
simply by recombination. The rate at which electrons with proper number density ne
undergo recombination is approximately Rn, where R ' 3 × 10−17 (T /K)−1/2 m3 s−1 . If
ne = xntot
e , where x is the fractional ionization and
3H02
ntot
e = Ωb (1 + z)3 ,
8πGµmp
where the parameter µ is approximately 1.143 for a gas of 25% helium by mass, discuss
the freezeout of the ionization of the cosmic plasma. Show that the ionization does not
fall below x ' 10−4 . At what redshift does this freezeout occur?
(b) Ionized material at low redshifts causes CMB photons to undergo Thomson
scattering, which affects the observed CMB anisotropies. Show that the optical depth
due to scattering between a redshift z and z = 0 is
c dz
Z Z
τ= σT ne d`prop = σT ne p
H0 (1 + z) 1 − Ωm + Ωm (1 + z)3
Ωb hp i Ωb
τ = 0.04h 1 + Ωm z(3 + 3z + z 2 ) − 1 ' 0.04h 1/2 z 3/2 .
Ωm Ωm
(d) Describe the effect of this scattering on the pattern of CMB anisotropies.
IVER
UN S
IT
TH
Y
H
O F
G
E
R
DI
N B U
Examinable material
The following equations should be considered examinable in the sense that you may be expected
to remember them and be able to use them. Long derivations are less likely to be required – and
certainly will not be requested unless the purpose of a question is to lead you with hints through
the steps of such a derivation.
RW metric:
c2 dτ 2 = c2 dt2 − R2 (t) dr2 + Sk2 (r) dψ 2 ,
Hubble parameter:
H(t) = Ṙ(t)/R(t).
Dimensionless (current) Hubble parameter:
H0
h≡ ,
100 km s−1 Mpc−1
Density parameter:
ρ 8πGρ
Ω≡ = .
ρc 3H 2
Equation of state:
w ≡ p/ρc2 ⇒ ρ ∝ a−3(w+1) .
Time-dependent Hubble parameter:
2c −0.4
R0 rH ≃ Ω .
H0 m
n ∝ s ∝ T 3.
Planck mass: r
h̄c
mP ≡ ≃ 1019 GeV.
G
Lagrangian density for real scalar field with mass term:
L = 21 (∂ µ φ∂µ φ − µ2 φ2 ).
ρ = 12 φ̇2 + V (φ)
p = 21 φ̇2 − V (φ).
φ̈ + 3H φ̇ + dV /dφ = 0.
Power spectrum: X
hδ 2 i = |δk |2 .
Dimensionless Power spectrum:
L3
∆2 (k) ≡ 4πk 3 |δk |2 .
(2π)3
Newtonian-gauge metric:
u̇ + 2(ȧ/a)u = g/a.
Continuity equation:
d
∇ · u.
δ = −(1 + δ)∇
dt
Linearised continuity equation:
∇ · u.
δ̇ = −∇
Linear equation for matter-dominated density fluctuations, including pressure:
ȧ
δ̈ + 2 δ̇ = δ 4πGρ0 − c2s k 2 /a2 .
a
δ(t) ∝ t2/3 .
δ(t) ∝ t.
Transfer function:
∆2 (k) ∝ k 3+n Tk2 .
Horizon size at matter-radiation equality (break scale in Tk ):
Displacement field:
x(t) = a(t) (q + f (q, t)) .
Density field:
∂q
ρ / ρ̄ =
≃ 1−∇ ·f (f → 0).
∂ x/a
Spherical model criteria for virialization:
δT X
(q̂) = am
ℓ Yℓm (q̂).
T
CMB dimensionless power per log ℓ:
δT 1
= (Φ/c2 ).
T 3
h2 = ω m + ω v + ω k .
IT
TH
Y
H
O F
G
E
R
DI
N B U
J.A. Peacock
Room C20, Royal Observatory; jap@roe.ac.uk
http://www.roe.ac.uk/japwww/teaching/cos5.html
1
Syllabus (subject to alteration)
(2) The hot big bang Thermal history; freezeout; relics; recombination and
last scattering
(3) Inflation – I Initial condition problems; Planck era; Physics beyond the SM;
Scalar fields; Noether’s theorem
(10) Galaxy formation DM haloes; density profiles; mass function; gas cooling;
feedback; observational challenges to hierarchical merging
(15) The physical vacuum Evidence for Λ; the cutoff problem; quintessence
2
1 Review of Friedmann models
Topics to be covered:
• Cosmological spacetime and RW metric
• Expansion dynamics and Friedmann equation
• Calculating distances and times
Here, dashed coordinates are local to the object, undashed are the global coordinates
we use. As usual, the Greek indices run from 0 to 3. Note the ambiguity in defining the
sign of the squared interval. The matrix gµν is the metric tensor, which is found
in principle by solving Einstein’s gravitational field equations. A simpler alternative,
which fortunately matches the observed universe pretty well, is to consider the most
symmetric possibilities for the metric.
x2 + y 2 + z 2 + w2 − v 2 = A2 (2)
Effectively, we have made one coordinate imaginary because we know we want to end
up with the 4D spacetime signature.
This maximally symmetric spacetime is known as de Sitter space. It looks
like a static spacetime, but relativity can be deceptive, as the interpretation depends on
the coordinates you choose. Suppose we re-express things using the analogues of polar
coordinates:
v = A sinh α
w = A cosh α cos β
z = A cosh α sin β cos γ (4)
y = A cosh α sin β sin γ cos δ
x = A cosh α sin β sin γ sin δ.
3
This has the advantage that it is an orthogonal coordinate system: a vector such
as eα = ∂(x, y, z, w, v)/∂α is orthogonal to all the other ei (most simply seen by
considering eδ and imagining continuing the process to still more dimensions). The
squared length of the vector is just the sum of |eαi |2 dαi2 , which makes the metric into
where R(t) is the scale factor. Differentiating this with respect to t gives
The characteristic time of the expansion is called the Hubble time, and takes the
value
4
By analogy with the de Sitter result, we now guess that the spatial metric will
factorize into the scale factor times a comoving part that includes curvature. If the
spatial part of the metric is to have a homogeneous curvature, we must be able to write
it as a 3-sphere embedded in four-dimensional Euclidean space: x2 +y 2 +z 2 +w2 = R2 .
Now define the equivalent of spherical polars and write w = R cos α, z = R sin α cos β,
y = R sin α sin β cos γ, x = R sin α sin β sin γ, where α, β and γ are three arbitrary
angles. As before, the vectors eα etc. generated by increments in the angles are clearly
orthogonal:
so the squared length of the vector is just |eα |2 + |eβ |2 + |eγ |2 , which gives
This is the metric for the case of positive spatial curvature, if we relabel α → r and
(β, γ) → (θ, φ) – the usual polar angles. We could write the angular part just in terms
of the angle dψ that separates two points on the sky, dψ 2 = dθ2 + sin2 θ dφ2 , in which
case the metric is the same form as for the surface of a sphere. This was inevitable:
the hypersurface x2 + y 2 + z 2 + w2 = R2 always allows two points to be chosen to have
w = 0 (the first by choice of origin; the second via rotation), so that their separation is
that of two points on the surface of a sphere.
This k = +1 metric describes a closed universe, in which a traveller who
sets off along a trajectory of fixed β and γ will eventually return to their starting point
(when α = 2π). The model is finite, but unbounded. By contrast, if we define a space
of negative curvature via R → iR and α → iα, then sin α → i sinh α and cos α → cosh α
(so that x, y, z stay real, as they must). The new angle α can increase without limit,
and (x, y, z) never return to their starting values. The k = −1 metric thus describes an
open universe of infinite extent.
In summary, this overall Robertson–Walker metric (RW metric), can be
written as:
The comoving radius r is dimensionless, and the scale factor R really is the radius
of curvature of the universe. Both are required in order to give a comoving distance
dimensions of length – e.g. the combination R0 Sk (r). Nevertheless, it is often convenient
to make the scale factor dimensionless, via
R(t)
a(t) ≡ , (16)
R0
so that a = 1 at the present. This is not the only form: we can define a new radius
which is equal to Sk (r), giving
dr′2
2 2 2 2 2 ′2 2
c dτ = c dt − R (t) + r dψ . (17)
1 − kr′2
5
light propagation and redshift Light follows trajectories with zero proper
time (null geodesics). The radial equation of motion therefore integrates to
Z
r= c dt/R(t). (18)
The comoving distance is constant, whereas the domain of integration in time extends
from temit to tobs ; these are the times of emission and reception of a photon. Thus
dtemit /dtobs = R(temit )/R(tobs ), which means that events on distant galaxies time-
dilate. This dilation also applies to frequency, so
νemit R(tobs )
≡1+z = . (19)
νobs R(temit )
In terms of the normalized scale factor a(t) we have simply a(t) = (1 + z)−1 . So just
by observing shifts in spectral lines, we can learn how big the universe was at the time
the light was emitted. This is the key to performing observational cosmology.
the friedmann equation The equation of motion for the scale factor resembles
Newtonian conservation of energy for a particle at the edge of a uniform sphere of radius
R:
8πG 2
Ṙ2 − ρR = −kc2 . (20)
3
This is almost obviously true, since the Newtonian result that the gravitational field
inside a uniform shell is zero does still hold in general relativity, and is known as
Birkhoff ’s theorem. But there are some surprises hidden here. This energy-like
equation can be turned into a force-like equation by differentiating with respect to time:
To deduce this, we need to know ρ̇, which comes from conservation of energy:
The surprising factor here is the occurrence of the active mass density ρ + 3p/c2 .
This is here because the weak-field form of Einstein’s gravitational field equations is
The extra term from the pressure is important. As an example, consider a radiation-
dominated fluid – i.e. one whose equation of state is the same as that of pure
radiation: p = u/3, where u is the energy density. For such a fluid, ρ + 3p/c2 = 2ρ, so
its gravity is twice as strong as we might have expected.
But the greatest astonishment in the Friedmann equation is the term on the
rhs. This is related to the curvature of spacetime, and k = 0, ±1 is the same integer
that is found in the RW metric. This cannot be completely justified without the Field
Equations, but the flat k = 0 case is readily understood. Write the energy-conservation
equation with an arbitrary rhs, but divide through by R2 :
8πG const
H2 − ρ= . (24)
3 R2
6
Now imagine holding the observables H and ρ constant, but let R → ∞; this has the
effect of making the rhs of the Friedmann equation indistinguishable from zero. Looking
at the metric with k 6= 0, R → ∞ with Rr fixed implies r → 0, so the difference between
Sk (r) and r becomes negligible and we have in effect the k = 0 case.
There is thus a critical density that will yield a flat universe,
3H 2
ρc = . (25)
8πG
It is common to define a dimensionless density parameter as the ratio of density
to critical density:
ρ 8πGρ
Ω≡ = . (26)
ρc 3H 2
which diverges as the universe approaches the flat state with Ω = 1. In practice, Ω0
is such a common symbol in cosmological formulae, that it is normal to omit the zero
subscript. We can also define a dimensionless (current) Hubble parameter as
H0
h≡ , (28)
100 km s−1 Mpc−1
w ≡ p/ρc2 , (30)
so
ρ ∝ a−3(w+1) (32)
7
silly game to play, but cosmology can be the only way to learn if something unexpected
exists. As we will see in more detail later, modern data force us to accept a contribution
that is approximately independent of time with w ≃ −1: a vacuum energy that
is simply an invariant property of empty space. A general name for this contribution
is dark energy, reflecting our ignorance of its nature (although the name is not
very good, since it is too similar to dark matter: ‘dark tension’ would better reflect its
unusual equation of state with negative pressure).
In terms of observables, this means that the density is written as
8πGρ
= H02 (Ωv a−3(w+1) + Ωm a−3 + Ωr a−4 ) (33)
3
(using the normalized scale factor a = R/R0 ). We will generally set w = −1 without
comment, except where we want to focus explicitly on this parameter. This expression
allows us to write the Friedmann equation in a manner useful for practical solution.
Start with the Friedmann equation in the form H 2 = 8πGρ/3 − kc2 /R2 . Inserting the
expression for ρ(a) gives
This equation is in a form that can be integrated immediately to get t(a). This is not
possible analytically in all cases, nor can we always invert to get a(t), but there are
some useful special cases worth knowing. Mostly these refer to the flat universe
with total Ω = 1. Curvature can always be neglected at sufficiently early times, as
can vacuum density (except that the theory of inflation postulates that the vacuum
density was very much higher in the very distant past). The solutions look simplest if
we appreciate that normalization to the current era is arbitrary, so we can choose a = 1
to be at a convenient point where the densities of two main components cross over.
Also, the Hubble parameter at that point (H∗ ) sets a characteristic time, from which
we can make a dimensionless version τ ≡ tH∗ .
2
matter and radiation Using dashes to denote d/d(t/τ ), we have a′ = (a−2 +
−1
a )/2, which is simply integrated to yield
√
2 2 √
τ= 2 + (a − 2) 1 + a . (35)
3
This can be inverted to yield a(τ ), but the full expression is too ugly to be much use.
It will suffice to note the limits:
√
τ ≪ 1 : a = ( 2τ )1/2 .
√ (36)
τ ≫ 1 : a = (3τ /2 2)2/3 ,
so the universe expands as t1/2 in the radiation era, which becomes t2/3 once matter
dominates. Both these powers are shallower than t, reflecting the decelerating nature
of the expansion.
2
radiation and√vacuum
p Now we have a′ = (a−2 + a2 )/2, which is easily solved
2 ′
in the form (a ) / 2 = 1 + (a2 )2 , and simply inverted:
√ 1/2
a = sinh( 2τ ) . (37)
8
negative vacuum density? It is easy to repeat the above exercise defining the critical
era as one where ρr equals |ρv |, in which case the solution is the same, except with
sinh → sin. A negative vacuum density always leads to eventual recollapse into a big
crunch.
2
matter and vacuum Here, a′ = (a−1 + a2 )/2, which can be tackled via the
3/2
substitution y = a , to yield
√ 2/3
a = sinh(3τ /2 2) . (38)
This transition from the flat matter-dominated a ∝ t2/3 to de Sitter space seems to be
the one that describes our actual universe (apart from the radiation era at z > 4
∼ 10 ). It
is therefore worth being explicit about how to translate units to the usual normalization
at a = 1 today. We have to multiply the above expression by a∗ , which is the usual
scale factor at matter-vacuum equality, and we have to relate the Hubble parameter H∗
to the usual H0 . This is easily done using the Friedmann equation:
curved models We will not be very strongly concerned with highly curved models
in this course, but it is worth knowing some basic facts, as shown in figure 1 (neglecting
radiation). On a plot of the Ωm − Ωv plane, the diagonal line Ωm + Ωv = 1 always
separates open and closed models. If Ωv < 0, recollapse always occurs – whereas a
positive vacuum density does not always guarantee expansion to infinity, especially
when the matter density is high. For closed models with sufficiently high vacuum
density, there was no big bang in the past, and the universe must have emerged from
a ‘bounce’ at some finite minimum radius. All these statements can be deduced quite
simply from the Friedmann equation.
dz R0 dR
=− 2 = −(1 + z)H(z), (40)
dt R dt
R
so t = H(z) dz/(1 + z), where
This can’t be done analytically in general, but the following simple approximate formula
is accurate to a few % for cases of practical interest:
2
H0 t 0 ≃ (0.7Ωm − 0.3Ωv + 0.3)−0.3 . (42)
3
9
Figure 1. This plot shows the different possibilities for the
cosmological expansion as a function of matter density and vacuum energy.
Models with total Ω > 1 are always spatially closed (open for Ω < 1),
although closed models can still expand to infinity if Ωv 6= 0. If the
cosmological constant is negative, recollapse always occurs; recollapse is
also possible with a positive Ωv if Ωm ≫ Ωv . If Ωv > 1 and Ωm is small,
there is the possibility of a ‘loitering’ solution with some maximum redshift
and infinite age (top left); for even larger values of vacuum energy, there
is no big bang singularity.
c
Z
R0 r = dz. (43)
H(z)
Remember that non-flat models need the combination R0 Sk (r), so one has to divide the
above integral by R0 = (c/H0 )|Ω − 1|−1/2 , apply the Sk function, and then multiply by
R0 again. Once more, this process is not analytic in general.
2c −0.4
R0 rH ≃ Ω . (44)
H0 m
At high redshift, where H increases, this tends to zero. The onset of radiation
domination does not change this: even though the presently visible universe was once
very small, it expanded so quickly that causal contact was not easy. The observed
large-scale near-homogeneity is therefore something of a puzzle.
10
angular diameters Recall the RW metric:
c2 dτ 2 = c2 dt2 − R2 (t) dr2 + Sk2 (r) dψ 2 .
(45)
The spatial parts of the metric give the proper transverse size of an object seen by us
as its comoving size dψ Sk (r) times the scale factor at the time of emission:
dℓ⊥ = dψ R(z)Sk (r) = dψ R0 Sk (r)/(1 + z). (46)
If we know r, we can therefore convert the angle subtended by an object into its physical
extent perpendicular to the line of sight.
surface brightness The flux density is the product of the specific intensity
Iν and the solid angle subtended by the source: Sν = Iν dΩ. Combining the angular
size and flux-density relations gives a relation that is independent of cosmology:
Bν ([1 + z]ν0 )
Iν (ν0 ) = , (48)
(1 + z)3
where Bν is surface brightness (luminosity emitted into unit solid angle per unit
area of source). This (1 + z)3 dimming makes it hard to detect extended objects at
very high redshift. The factor becomes (1 + z)4 if we integrate over frequency to get a
bolometric quantity.
effective distances The angle and flux relations can be made to look Euclidean:
Some example distance-redshift relations are shown in figure 2. Notice how a high matter
density tends to make high-redshift objects brighter: stronger deceleration means they
are closer for a given redshift.
11
Figure 2. A plot of dimensionless angular-diameter distance versus
redshift for various cosmologies. Solid lines show models with zero vacuum
energy; dashed lines show flat models with Ωm + Ωv = 1. In both
cases, results for Ωm = 1, 0.3, 0 are shown; higher density results in lower
distance at high z, due to gravitational focusing of light rays.
Although the timescale for expansion of the early universe is very short, the density is
also very high, so it is normally sensible to assume that conditions are close to thermal
equilibrium. Also the fluids of interest are simple enough that we can treat them as
perfect gases. The thermodynamics of such a gas is derived staring with a box of
volume V = L3 , and expanding the fields inside into periodic waves with harmonic
boundary conditions. The density of states in k space is
V
dN = g 3
d3 k (50)
(2π)
(where g is a degeneracy factor for spin etc.). The equilibrium occupation number
for a quantum state of energy ǫ is given generally by
h i−1
hf i = e(ǫ−µ)/kT ± 1 (51)
(+ for fermions, − for bosons). Now, for a thermal radiation background, the chemical
potential, µ is always zero. The reason for this is quite simple: µ appears in the first
law of thermodynamics as the change in energy associated with a change in particle
number, dE = T dS − P dV + µdN . So, as N adjusts to its equilibrium value, we expect
that the system will be stationary with respect to small changes in N . The thermal
equilibrium background number density of particles is
∞
1 1 4π p2 dp
Z Z
n= f dN = g , (52)
V (2πh̄)3 0 eǫ(p)/kT ± 1
12
p
where we have changed to momentum space; ǫ = m2 c4 + p2 c2 and g is the degeneracy
factor. There are two interesting limits of this expression.
(1) Ultrarelativistic limit. For kT ≫ mc2 the particles behave as if they were mass-
less, and we get
3 ∞
y 2 dy
kT 4πg
Z
n= . (53)
c (2πh̄)3 0 ey ± 1
(2) Non-relativistic limit. Here we can neglect the ±1 in the occupation number, in
which case the number is suppressed by a dominant exp(−mc2 /kT ) factor:
n
= (mc2 /kT )3/2 exp(−mc2 /kT ) × [0.695 (F) 0.521 (B)]. (54)
nultrarel
This shows us that the background ‘switches on’ at about kT ∼ mc2 ; at this
energy, known as a threshold, photons and other species in equilibrium will
have sufficient energy to create particle-antiparticle pairs.
The above thermodynamics also gives the energy density of the background, since
it is only necessary to multiply the integrand by a factor ǫ(p) for the energy in each
mode:
Z ∞
2 1 4π p2 dp
u = ρc = g ǫ(p). (55)
(2πh̄)3 0 eǫ(p)/kT ± 1
π2
u= g (kT )4 (bosons), (56)
30(h̄c)3
Equating the dV and dT parts gives the familiar ∂E/∂T = T ∂S/∂T and
E + PV
S= (58)
T
These results take an interesting and simple form in the ultrarelativistic limit.
The energy density, u, obeys the usual black-body scaling u ∝ T 4 . In the ultrarelativistic
limit, we also know that the pressure is P = u/3, so that the entropy density is
2π 2 k
s = (4/3)u/T = g (kT )3 (bosons), (59)
45(h̄c)3
and 7/8 of this for fermions. Now, we saw earlier that the number density of an
ultrarelativistic background also scales as T 3 – therefore we have the simple result
13
that entropy just counts the number of particles. This justifies a common piece of
terminology, in which the ratio of the number density of photons in the universe to
the number density of baryons (protons plus neutrons) is called the entropy per
baryon. As we will see later, this ratio is about 109 . The fact that this ratio is so
large justifies the adiabatic assumption: pretty well all the entropy is in the photons.
π2 X 7 X
ρc2 = geff (kT )4 ; geff ≡ gi + gj , (60)
30(h̄c)3 8
bosons fermions
14
time and temperature This temperature-dependent equilibrium density sets
the timescale for expansion in the early universe. Using the relation between time
and density for a flat radiation–dominated universe, t = (32πGρ/3)−1/2 , we can deduce
the time–temperature relation:
−1/2 −2
t/seconds = geff T /1010.26 K . (61)
This temperature was of course higher in the past, owing to the adiabatic expansion of
the universe. Frequently, we will assume
which is justified informally by arguing that photon energies scale as E ∝ 1/a and
saying that the typical energy in black-body radiation is ∼ kT . Being more careful,
we should conserve entropy, so that s ∝ a−3 . Since s ∝ T 3 while heff is constant, this
requires T ∝ 1/a. But clearly this does not apply near a threshold. At these points,
heff changes rapidly and the universe will expand at nearly constant temperature for a
period.
The energy density in photons is supplemented by that of the neutrino
background. Because they have a lower temperature, as shown below, they contribute
an energy density 0.68 times that from the photons (if the neutrinos are massless and
therefore relativistic). If there are no other contributions to the energy density from
relativistic particles, then the total effective radiation density is Ωr h2 ≃ 4.2 × 10−5 and
the redshift of matter–radiation equality is
The time of this change in the global equation of state is one of the key epochs in
determining the appearance of the present-day universe.
The following table shows some of the key events in the history of the universe.
Note that, for very high temperatures, energy units for kT are often quoted instead of
T . The conversion is kT = 1 eV for T = 104.06 K. Some of the numbers are rounded,
rather than exact; also, some of them depend a little on Ω and H0 . Where necessary, a
flat model with Ω = 0.3 and h = 0.7 has been assumed.
15
2.2 Freezeout and relics
So far, we have assumed that thermal equilibrium will be followed in the early universe,
but this is far from obvious. Equilibrium is produced by reactions that involve individual
particles, e.g. e+ e− ↔ 2γ converts between electron-positron pairs and photons. When
the temperature is low, typical photon energies are too low for this reaction to proceed
from right to left, so there is nothing to balance annihilations.
Nevertheless, the annihilations only proceed at a finite rate: each member of the
pair has to find a partner to interact with. We can express this by writing a simple
differential equation for the electron density, called the Boltzmann equation:
where σ is the reaction cross-section, v is the particle velocity, and S is a source term
that represents thermal particle production. The 3Hn term just represents dilution by
the expansion of the universe. Leaving aside the source term for the moment, we see
that the change in n involves two timescales:
Both these times increase as the universe expands, but the interaction time usually
changes fastest. Two-body reaction rates scale proportional to density, times a cross-
section that is often a declining function of energy, so that the interaction time changes
at least as fast as R3 . In contrast, the Hubble time changes no faster than R2 (in the
radiation era), so that there is inevitably a crossover.
The situation therefore changes from one of thermal equilibrium at early times
to a state of freezeout or decoupling at late times. Once the interaction timescale
becomes much longer than the age of the universe, the particle has effectively ceased to
interact. It thus preserves a ‘snapshot’ of the properties of the universe at the time the
particle was last in thermal equilibrium. This phenomenon of freezeout is essential to
the understanding of the present-day nature of the universe. It allows for a whole set of
relics to exist from different stages of the hot big bang. The photons of the microwave
background are one such relic, generated at redshift z ≃ 1100. A more exotic example
is the case of neutrinos.
To complete the Boltzmann equation, we need the source term S. This term can
be fixed by a thermodynamic equilibrium argument: for a non-expanding universe, n
will be constant at the equilibrium value for that temperature, nT , showing that
S = hσvin2T . (67)
16
size, halving the temperature. A good example is the microwave background, whose
photons last interacted with matter at z ≃ 1000. The CMB nevertheless still appears to
be equilibrium black-body radiation because the number density of photons has fallen
by the right amount to compensate for the redshifting of photon energy. This sounds
like an incredible coincidence, but is in fact quite inevitable when looked at from the
quantum-mechanical point of view. This says that the occupation number of a given
mode, = (exp h̄ω/kT − 1)−1 for thermal radiation, is an adiabatic invariant that does
not change as the universe expands – only the frequency alters, and thus the apparent
temperature.
Now consider the opposite case, where the thermal solution would be
nonrelativistic, with NT ∝ T −3/2 exp(−mc2 /kT ). If the background stays at the
equilibrium value, the lhs of the rate equation will therefore be negative and ≫ 1
in magnitude. This is consistent if Γ/H ≫ 1, because then the (NT /N )2 term
on the rhs can still be close to unity. However, if Γ/H ≪ 1, there must be a
deviation from equilibrium. When NT changes sufficiently fast with a, the actual
abundance cannot keep up, so that the (NT /N )2 term on the rhs becomes negligible
and d ln N/d ln a ≃ −Γ/H, which is ≪ 1. There is therefore a critical time at which
the reaction rate drops low enough that particles are simply conserved as the universe
expands – the population has frozen out. This provides a more detailed justification
for the intuitive rule-of-thumb used above to define decoupling,
Exact numerical solutions of the rate equation almost always turn out very close to this
simple rule, as shown in figure 4.
17
the relic density The above freezeout criterion can be used to deduce a simple
and very important expression for the present-day density of a non-relativistic relic:
10−41.5 m2
Ωrelic h2 ∼ , (70)
σ
so that only a small range of annihilation cross-sections will be of observational
interest. The steps needed to get this formula are as follows. (1) From Γ/H = 1,
the number density of relics at freezeout is nf = Hf /hσvi; (2) H = (8πGρ/3)1/2 , where
ρc2 = (π 2 /30h̄3 c3 )geff (kT )4 ; (3) Ωrelic = 8πGmn0 /3H02 . The only missing ingredient
here is how to relate the present number density n0 to the density nf at temperature
Tf . Since the relics are conserved, the number density must have fallen by the same
factor as the entropy density:
Today, h0eff = 43/11, and hfeff = geff at high redshift. This allows us to deduce the relic
density, given the mass, cross-section and temperature of freezeout:
10−33.0 m2 mc2
2 −1/2
Ωrelic h ≃ geff . (72)
hσvi kTf
We see from figure 4 that mc2 /kTf ∼ 10 with only a logarithmic dependence on reaction
rate, which roughly cancels the last factor on the rhs. Finally, since particles are nearly
relativistic at freezeout, we set hσvi = σc to get our final estimate of the typical cross-
section for an interesting relic abundance. The eventual conclusion makes sense: the
higher the cross-section, the longer the particle can stay in equilibrium, and the more
effective annihilations can be in suppressing the number density. Note that, in detail,
we need to worry about whether the particle is a Majorana particle (i.e. its own
antiparticle) or a Dirac particle where particles and antiparticles are distinct.
neutrino decoupling The best case for application of this freezeout apparatus
is to relic neutrinos. At the later stages of the big bang, energies are such that only
light particles survive in equilibrium: photons (γ), neutrinos (ν) and e+ e− pairs. As the
temperature falls below Te = 109.7 K), the pairs will annihilate. Electrons can interact
via either the electromagnetic or the weak interaction, so in principle the annihilations
might yield pairs of photons or neutrinos. However, in practice the weak reactions freeze
out earlier, at T ≃ 1010 K.
The effect of the electron-positron annihilation is therefore to enhance the
numbers of photons relative to neutrinos. Strictly, what is conserved in this process
is the entropy. The entropy of an e± + γ gas is easily found by remembering that it
is proportional to the number density, and that all three particle species have g = 2
(polarization or spin). The total is then
11
s(γ + e+ + e− ) = s(γ). (73)
4
Equating this to photon entropy at a new temperature gives the factor by which the
photon temperature is enhanced with respect to that of the neutrinos. Equivalently,
given the observed photon temperature today, we infer the existence of a neutrino
background with a temperature
1/3
4
Tν = Tγ = 1.945 K, (74)
11
for Tγ = 2.725 K. Although it is hard to see how such low energy neutrinos could ever
be detected directly, their gravitation is certainly not negligible: they contribute an
18
energy density that is a factor (7/8) × (4/11)4/3 times that of the photons. For three
neutrino species, this enhances the energy density in relativistic particles by a factor
1.68 (there are three different kinds of neutrinos, just as there are three leptons: the
µ and τ particles are heavy analogues of the electron).
massive neutrinos Although for many years the conventional wisdom was that
neutrinos were massless, this assumption began to be increasingly challenged around
the end of the 1970s. Theoretical progress in understanding the origin of masses in
particle physics meant that it was no longer natural for the neutrino to be completely
devoid of mass. Also, there is now some experimental evidence that neutrinos have a
small non-zero mass. The consequences of this for cosmology could be quite profound,
as relic neutrinos are expected to be very abundant. The above section showed that
n(ν +ν) = (3/4)n(γ; T = 1.95 K). That yields a total of 113 relic neutrinos in every cm3
for each species. The density contributed by these particles is easily worked out provided
the mass is small enough. If this is the case, then the neutrinos were ultrarelativistic
at decoupling and their statistics were those of massless particles. As the universe
expands to kT < mν c2 , the total number of neutrinos is preserved. We therefore obtain
the present-day mass density in neutrinos just by multiplying the zero-mass number
density by mν , and the consequence for the cosmological density in light neutrinos is
easily worked out to be
P
2 mi
Ων h = . (75)
94.1 eV
The more complicated case of neutrinos that decouple when they are already
nonrelativistic is studied below.
19
The current direct laboratory limits to the neutrino masses are
νe <
∼ 2.2 eV νµ <
∼ 0.17 MeV ντ <
∼ 15 MeV. (76)
Based on this, even the electron neutrino could be of great cosmological significance.
But in practice, we will see later that studies of cosmological large-scale structure limit
the sum of the masses to a maximum of about 0.5 eV. This is becoming interesting, since
it is known that neutrino masses must be non-zero. In brief, this comes from studies
of neutrino mixing, in which each neutrino type is a mixture of energy eigenstates.
The energy differences can be measured, which yields a measure of the difference in the
square of the masses (consider the relativistic relation E 2 = m2 + p2 , and expand to
get E ≃ m + m2 /2p). These mixings are known from wonderfully precise experiments
detecting neutrinos generated in the sun and the Earth’s atmosphere:
where m1 , m2 and m3 are the three mass eigenstates. This information does not give
the absolute mass scale, nor does it tell us whether there is a normal hierarchy
with m3 ≫ m2 ≫ m1 , or an inverted hierarchy in which states 1 & 2 are a close
doublet lying well above state 3. Cosmology can settle both these issues by measuring
the total density in neutrinos. The absolute minimum situation is a normal hierarchy
with m1 negligibly small, in which case the mass is dominated by m3 , which is around
0.05 eV. The cosmological limits are within a power of 10 of this interesting point.
relic particles as dark matter Many other particles exist in the early
universe, so there are a number of possible relics in addition to the massive neutrino. A
common collective term for these particles is WIMP – standing for weakly interacting
massive particle. There are really three generic types to consider, as follows.
(1) Hot Dark Matter (HDM) These are particles that decouple when
relativistic, and which have a number density roughly equal to that of photons;
eV-mass neutrinos are the archetype. The relic density scales linearly with the
particle mass.
(2) Warm Dark Matter (WDM) If the particle decouples sufficiently early,
the relative abundance of photons can then be boosted by annihilations other
than just e± . In modern particle physics theories, there are of order 100 distinct
particle species, so the critical particle mass to make Ω = 1 can be boosted to
around 1 – 10 keV.
(3) Cold Dark Matter (CDM) If the relic particles decouple while they are
nonrelativistic, the number density can be exponentially suppressed. If the
interactions are like those of neutrinos, then the freezeout temperature is about
1 MeV, and the relic mass density then falls with increasing mass (see figure 6).
For weak interactions, cross-sections scale as (energy)2 , so that the relic density
falls as 1/m2 . Interesting masses then lie in the ≃ 10 GeV range, this cannot
correspond to the known neutrinos, since such particles would have been seen in
accelerators. But beyond about 90 GeV (the mass of the Z boson), the strength
of the weak interaction is reduced, with cross-section going as (energy)−2 . The
relic density now rises as m2 , so that the observed dark matter density is
attained at m ≃ 1 TeV. Plausible candidates of this sort are found among so-
called supersymmetric theories, which predict many new weakly-interacting
particles. The favoured particle for a CDM relic is called the neutralino.
Since these particles exist to explain galaxy rotation curves, they must be passing
through us right now. There is therefore a huge effort in the direct laboratory detection
of dark matter, mainly via cryogenic detectors that look for the recoil of a single nucleon
when hit by a DM particle (in deep mines, to shield from cosmic rays). So far, there
20
Figure 6. The contribution to the density parameter produced by
relic neutrinos (or neutrino-like particles) as a function of their rest mass.
The shaded band shows a factor of 2 either side of the observed CDM
density. At low masses, the neutrinos are highly relativistic when they
decouple: their abundance takes the zero-mass value, and the density is
just proportional to the mass. Above about 1 MeV, the neutrinos are non-
relativistic at decoupling, and their relic density is reduced by annihilation.
Above the mass of the Z boson, the cross-section falls, so that annihilation
is less effective and the relic density rises again.
are no detections, but well-constructed experiments with low backgrounds are starting
to set interesting limits, as shown in figure 7. There is no unique target to aim for,
since even the simplest examples of supersymmetric models contain a variety of free
parameters. These allow models that are optimistically close to current limits, but also
some that will be hard to verify. The public-domain package DarkSUSY is available
at www.physto.se/~edsjo/darksusy to make these detailed abundance calculations.
What is particularly exciting is that the properties of these relic particles can
also be observed via new examples manufactured in particle accelerators. The most
wonderful outcome would be for the same particle to be found in these two different
ways. The chances of success in this enterprise are hard to estimate, and some models
exist in which detection would be impossible for many decades. But it would be a
tremendous scientific achievement if dark matter particles were to be detected in this
way, and a good part of the plausible parameter space will be covered over the next
decade.
21
−4
10
−8
10
−10
10
−12
10 1 2 3 4
10 10 10 10
WIMP Mass (GeV)
temperatures there were 1 + O(10−9 ) protons for every antiproton. If baryon number is
conserved, this imbalance cannot be altered once it is set in the initial conditions; but
what generates it? This is clearly one of the big challenges in cosmology, but our ideas
are less well formed here than in many other areas.
1MeV = 1010.065 K
me = 0.511 MeV
(78)
mp = 939 MeV
mn − mp = 1.3 MeV
22
Nucleosynthesis begins when equilibrium abundances of nuclei can no longer be
maintained by the weak interaction:
p + e− ↔ n + ν
(79)
n + e+ ↔ p + ν̄.
Because neutrons are more massive, they are less abundant in equilibrium. Once
equilibrium is lost, these remaining neutrons can be locked up in nuclei (although some
decay before this happens). The first nucleus to form is Deuterium, although this is
rapidly built up to Helium. In practice, the chain progresses little beyond Helium. This
is because Deuterium is not very strongly bound, and so forms late – at a time when the
universe has cooled sufficiently that nuclear reactions are starting to switch off. Helium
is much more strongly bound, and this assists its formation once Deuterium arises, but
thereafter nuclear burning ceases.
Since most He is 4 He, with 2 nucleons out of 4 being neutrons, the He fraction
by mass (denoted by Y ) is
4 × nn /2 2
Y = = (80)
nn + np 1 + np /nn
(neglecting neutrons in other elements). So, Y = 0.25 requires freezeout at nn /np ≃ 1/7.
Apart from Helium, the main nuclear residue of the big bang is therefore those
Deuterium nuclei that escape being mopped up into Helium, plus a trace of 3 He. This
residual Deuterium has a number density fixed at whatever sets the reaction rate low
enough to survive for a Hubble time. Since this density is a fixed quantity, the proportion
of the baryonic density that survives as Deuterium (or 3 He) should thus decline roughly
as 1/(density). This provides a relatively sensitive means of weighing the baryonic
content of the universe. In detail what is actually measured is the reciprocal of the
entropy per baryon:
(although lower values are favoured if a higher weight is given to the Helium abundance).
Baryons therefore cannot close the universe. If Ω = 1, the dark matter must be non-
baryonic.
2.3 Recombination
Moving closer to the present, and passing through matter-radiation equality at z ∼ 104 ,
the next critical epoch in the evolution of the universe is reached when the temperature
drops to the point (T ∼ 1000 K) where it is thermodynamically favourable for the
ionized plasma to form neutral atoms. This process is known as recombination: a
complete misnomer, as the plasma has always been completely ionized up to this time.
the rate equation A natural first thought is that the ionization of the plasma
may be treated by a thermal-equilibrium approach, using the Saha equation, which
applies in stellar atmospheres. In fact, such an approach is almost always invalid.
This is not because electromagnetic interactions are too slow to maintain equilibrium:
rather, thay are too fast. Consider a single recombination; if this were to occur directly
to the ground state, a photon with h̄ω > χ would be produced. Such photons are almost
immediately destroyed by ionizing another neutral atom. Similarly, reaching the ground
state requires the production of photons at least as energetic as the 2P → 1S spacing
23
Figure 8. The predicted primordial abundances of the light elements,
as a function of the baryon-to-photon ratio η (Smith, Kawano & Malaney
1993). For a microwave-background temperature of 2.73 K, this is related
to the baryonic density parameter via Ωb h2 = η/2.74×10−8 . Concordance
with the data is found only for η ≃ 3 × 10−10 , shown by the shaded strip.
(Lyman α, with λ = 1216Å), and these also are re-absorbed very efficiently. This is
a common phenomenon in astrophysics: the Lyman α photons undergo resonant
scattering and are very hard to get rid of (unlike a finite HII region, where the Lyα
photons can escape).
There is a way out, however, using two-photon emission. The 2S → 1S
transition is strictly forbidden at first order and one can only conserve energy and
angular momentum in the transition by emitting a pair of photons. Because of this
slow bottleneck, the ionization at low redshift is far higher than would be suggested by
the Saha prediction.
A highly stripped-down analysis of events simplifies the hydrogen atom to just
two levels (1S and 2S). Any chain of recombinations that reaches the ground state
can be ignored through the above argument: these reactions produce photons that are
immediately re-absorbed elsewhere, so they have no effect on the ionization balance.
The main chance of reaching the ground state comes through the recombinations that
reach the 2S state, since some fraction of the atoms that reach that state will suffer two-
photon decay before being re-excited. The rate equation for the fractional ionization is
thus
d(nx) Λ2γ
= −R (nx)2 , (83)
dt Λ2γ + ΛU (T )
24
where n is the number density of protons, x is the fractional ionization, R is the
recombination coefficient (R ≃ 3×10−17 T −1/2 m3 s−1 ), Λ2γ is the two-photon decay rate,
and ΛU (T ) is the stimulated transition rate upwards from the 2S state. This equation
just says that recombinations are a two-body process, which create excited states that
cascade down to the 2S level, from whence a competition between the upward and
downward transition rates determines the fraction that make the downward transition.
A fuller discussion (see chapter 6 of Peebles 1993 or section 3.6 of Mukhanov 2005)
would include a number of other processes: depopulation of the ground state by inverse
2-photon absorption; redshifting of Ly alpha photons due to the universal expansion,
which can prevent them being re-absorbed. At the redshifts of practical interest (1000
to 10), the simplified equation captures the main effect, although detailed calculations
have to include the recombination of He as well as H.
An important point about the rate equation is that it is only necessary to solve it
once, and the results can then be scaled immediately to some other cosmological model.
Consider the rhs: both R and ΛU (T ) are functions of temperature, and thus of redshift
only, so that any parameter dependence is carried just by n2 , which scales ∝ (Ωb h2 )2 ,
where Ωb is the baryonic density parameter. Similarly, the lhs depends on Ωb h2 through
n; the other parameter dependence comes if we convert time derivatives to derivatives
with respect to redshift:
dt
≃ −3.09 × 1017 (Ωm h2 )−1/2 z −5/2 s, (84)
dz
for a matter-dominated model at large redshift. Putting these together, the fractional
ionization must scale as
(Ωm h2 )1/2
x(z) ∝ . (85)
Ωb h2
This is a very different scaling from the prediction of the Saha equation.
The simplest case for solution of the rate equation is at late times, where the
recombination rate is slow enough for the two-photon process to cope with the influx
of excited atoms, and where the universe is cool enough for excitations from 2S to be
neglected. The evolution equation is then
d ln x Ωb h2
= 60 x z . (86)
d ln z (Ωm h2 )1/2
This expression neglects the expansion, and so it is invalid if the lhs is smaller than unity.
This criterion defines the freeze-out level of ionization at a given redshift; typically, the
full non-expanding solution for x(z) falls below this level at redshifts of a few hundred,
so that x never falls below ∼ 10−4 .
For a fully ionized plasma with 25% He by mass, the electron number density is
25
Figure 9. The ‘visibility function’ governing where photons in the
CMB undergo their final scattering. This is very nearly independent of
cosmological parameters, as illustrated by the effect of a 50% increase in
Ωb (dotted line), Ωm (dot-dashed line) and h (dashed line), relative to the
standard model (solid line).
Also, dℓproper = c dt, which brings in a factor of (Ωm h2 )−1/2 . These two density terms
automatically cancel the principal dependence of x(z), so we predict that the optical
depth should be very largely a function of redshift only. For standard parameters, a
good approximation around τ = 1 is
13
1+z
τ (z) ≃ (89)
1080
k × 2.725 K × (1 + z) = χ, (90)
which gives z ≃ 104.8 . The difference is partly because the ionization falls slower than
Saha, but also because even a tiny ionization easily causes scattering. The fact that
the properties of the last-scattering surface are almost independent of all the unknowns
in cosmology is immensely satisfying, and gives us at least one relatively solid piece of
ground to act as a base in exploring the trackless swamp of cosmology.
26
3 Inflation – I
Topics to be covered:
• Initial condition problems
• Dynamics of scalar fields
• Noether’s theorem
The expanding universe of the big-bang model is surprising in many ways: (1) What
caused the expansion? (2) Why is the expansion so close to flat – i.e. Ω ∼ 1 today?
(3) Why is the universe close to isotropic (the same in all directions)? (4) Why does
it contain structure? Some of these problems may seem larger than others, but when
examined in detail all point to something missing in our description of the early stages of
cosmological history. It is normally assumed that the solution to this will be some piece
of additional physics that is important at high energies. Before discussing possibilities,
there are a few important general ideas that will be needed.
k = h̄ = c = µ0 = ǫ0 = 1. (92)
This convention makes the meaning of equations clearer by reducing the algebraic
clutter, and is also useful in the construction of intuitive arguments for the order of
magnitude of quantities of interest. In the professional world of cosmology, natural
units are frequently adopted without comment, so this is something to watch out for.
Hereafter, natural units will frequently be adopted, although it will occasionally be
convenient to re-insert explicit powers of h̄ etc.
The adoption of natural units corresponds to fixing the units of charge, mass,
length and time relative to each other. This leaves one free unit, usually taken to be
energy. Natural units are thus one step short of the Planck system, in which G = 1
27
also, so that all units are fixed and all physical quantities are dimensionless. In natural
units, the following dimensional equalities hold:
[E] = [T ] = [m]
(93)
[L] = [m]−1
with units often quoted in GeV4 . It is however often of interest to express things in
Planck units: energy as a multiple of mP , energy density as a multiple of m4P etc. The
gravitational constant itself is then
G = m−2
P . (95)
flatness problem Now to quantify the first of the many puzzles concerning initial
conditions. From the Friedmann equation, we can write the density parameter as a
function of era:
(and corresponding expressions for the Ω(a) corresponding to any one component just
by picking the appropriate term on the top line). This tells us that, if the total Ω is
unity today, it was always unity (a geometrical statement: if k = 0, it can’t make a
continuous transition to k = ±1). But if Ω 6= 1, how does Ω(a) evolve? It should
be clear that Ω(a) → 1 at very large and very small a, provided Ωv is nonzero in the
former case, and provided Ωm or Ωr is nonzero in the latter case (without vacuum
energy, Ω = 1 is unstable). In short, the Ω = 1 state is an attractor, looking in either
direction in time. It has long been clear that this presents a puzzle with regard to the
initial conditions. These will be radiation dominated, so we have
(Ω − 1) 2
Ω(ainit ) ≃ 1 + ainit . (97)
Ωr
If we are willing to consider a Planck-scale origin with ainit ∼ 10−32 , then clearly
conditions at that time must be flat to perhaps 60 powers of 10. A more democratic
initial condition might be thought to have Ω − 1 of order unity, so some mechanism
to make it very small (or zero) is clearly required. This ‘how could the universe have
known?’ argument is a general basis for a prejudice that Ω = 1 holds exactly today:
it would seem contrived for the expansion to have proceeded for so many powers of 10
with Ω ≃ 1, only to depart just as we observe it.
horizon problem We have already mentioned the puzzle that it has apparently
been impossible to establish causal contact throughout the present observable universe.
Consider the integral for the horizon length:
c dt
Z
rH = . (98)
R(t)
The standard radiation-dominated R ∝ t1/2 law makes this integral converge near t = 0.
To solve the horizon problem and allow causal contact over the whole of the region
observed at last scattering requires a universe that expands ‘faster than light’ near t = 0:
R ∝ tα , with α > 1. It is tempting to assert that the observed homogeneity proves that
such causal contact must once have occurred, but this means that the equation of state
28
Figure 10. Illustrating the true history of the scale factor in the
simplest possible inflationary model. Here, the universe stays in an
exponential de Sitter phase for an indefinite time until its equation of
state abruptly changes from vacuum dominated to radiation dominated
at time tc . This must occur in such a way as to match R and Ṙ, leading to
the solid curve, where the plotted point indicates the join. For 0 < t < tc ,
the dashed curve indicates the time dependence we would infer if vacuum
energy was ignored. This reaches R = 0 at t = 0: the classical ‘big
bang’. The inflationary solution clearly removes this feature, placing any
singularity at large negative time. The universe is much older than we
would expect from observations at t > tc , which is one way of seeing how
the horizon problem can be evaded.
at early times must have been different. Indeed, if we look at Friedmann’s equation in
its second form,
and realize that R ∝ tα , with α > 1 implies an accelerating expansion, we see that what
is needed is negative pressure:
de sitter space The familiar example of negative pressure is vacuum energy, and
this is therefore a hint that the universe may have been vacuum-dominated at early
times. The Friedmann equation in the k = 0 vacuum-dominated case has the de
Sitter solution:
29
causally-connected patch to a size large enough to cover the whole presently-observable
universe. This is illustrated by in figure 10, where we assume that the universe can
be made to change its equation of state abruptly from vacuum dominated to radiation
dominated at some time tc . Before tc , we have R ∝ exp Ht; after tc , R ∝ t1/2 . We
have to match R and Ṙ at the join (otherwise the acceleration form of Friedmann’s
equation would be singular); it is then easy to show that tc = 1/2H. When we observe
the universe at t > tc , we predict that there was a singularity at t = 0, but the real
universe existed far earlier than this. In principle, the question ‘what happened before
the big bang?’ is now answered: there was no big bang. There might have still been a
singularity at large negative time, but one could imagine the de Sitter phase being of
indefinite duration, so that the true origin of everything can be pushed back to t = −∞.
In a sense, then, an inflationary start to the expansion would in reality be a very slow
one – as compared to the common popular description of ‘an extraordinarily rapid phase
of expansion’.
This idea of a non-singular origin to the universe was first proposed by the Soviet
cosmologist E.B. Gliner, in 1969. He suggested no mechanism my which the vacuum
energy could change its level, however. Before trying to plug this critical gap, we can
note that an early phase of vacuum-dominated expansion can also solve the flatness
problem. Consider the Friedmann equation,
8πGρR2
Ṙ2 = − kc2 . (102)
3
In a vacuum-dominated phase, ρR2 increases as the universe expands. This term can
therefore always be made to dominate over the curvature term, making a universe that
is close to being flat (the curvature scale has increased exponentially). In more detail,
the Friedmann equation in the vacuum-dominated case has three solutions:
sinh Ht (k = −1)
R ∝ cosh Ht (k = +1) (103)
exp Ht (k = 0),
p
where H = 8πGρvac /3. Note that H is not the Hubble parameter at an arbitrary
time (unless k = 0), but it becomes so exponentially fast as the hyperbolic trigonometric
functions tend to the exponential. If we assume that the initial conditions are not fine
tuned (i.e. Ω = O(1) initially), then maintaining the expansion for a factor f produces
Ω = 1 + O(f −2 ). (104)
This can solve the flatness problem, provided f is large enough. To obtain Ω of order
unity today requires |Ω − 1| <
∼ 10
−52
at the GUT epoch, and so
ln f >
∼ 60 (105)
e-foldings of expansion are needed; it will be proved below that this is also exactly
the number needed to solve the horizon problem. It then seems almost inevitable that
the process should go to completion and yield Ω = 1 to measurable accuracy today.
This is one of the most robust predictions of inflation (although, as we have seen, the
expectation of flatness is fairly general).
zP ≃ 1031.6 . (106)
30
This expands the Planck length to 0.4 mm today. This is far short of the present horizon
(∼ 6000 h−1 Mpc), by a factor of nearly 1030 , or e69 . It is more common to assume that
inflation happened at a safer distance from quantum gravity, at about the GUT energy
of 1015 GeV. The GUT-scale horizon needs to be stretched by ‘only’ a factor e60 in
order to be compatible with observed homogeneity. This tells us a minimum duration
for the inflationary era:
−1
∆tinflation > 60 Hinflation . (107)
The GUT energy corresponds to a time of about 10−35 s in the conventional radiation-
dominated model, and we have seen that this switchover time should be of order
−1
Hinflation . Therefore, the whole inflationary episode need last no longer than about
−33
10 s.
Since 1981, these ideas have been set on a more specific foundation using models for a
variable vacuum energy that come from particle physics. There are many variants, but
the simplest concentrate on scalar fields. These are fields like the electromagnetic
field, but differing in a number of respects. First, the field has only one degree of
freedom: just a number that varies with position, not a vector like the EM field. The
wave equation obeyed by such a field in flat space can be of a familiar form:
1
φ̈ − ∇2 φ = 0, (108)
c2
but this applies only for the special case where the quanta associated with the φ field are
massless. If they have mass, the equation becomes the Klein–Gordon equation:
1
2
φ̈ − ∇2 φ + (m2 c2 /h̄2 )φ = 0, (109)
c
This is easy to derive just by substituting the de Broglie relations p = −ih̄∇ ∇ and
E = ih̄∂/∂t into E 2 = p2 c2 + m2 c4 . To apply this to cosmology, we neglect the spatial
derivatives, since we imagine some initial domain in which we have a homogeneous
scalar field. This synchronizes the subsequent dynamics of φ(t) throughout the
observable universe (i.e. the patch that we inflate). The differential equation is now
d
φ̈ = − V (φ); V (φ) = (m2 c4 /h̄2 )φ2 /2. (110)
dφ
This is just a harmonic oscillator equation, and we can see that the field will oscillate
in the potential. In classical dynamics of a ball in a potential, this motion will conserve
energy: φ̇2 /2 + V (φ) = constant. The energy transforms itself from all potential at
the top of the motion, to all kinetic at the bottom. This behaviour is rather different
to the familiar oscillations of the electromagnetic field: if the field is homogeneous,
it does not oscillate. This is because the familiar energy density in electromagnetism
(ǫ0 E 2 /2 + B 2 /2µ0 ) is entirely kinetic energy in this analogy (to see this, write the
fields in terms of the potentials: B = ∇ ∧ A and E = −∇ ∇ φ − Ȧ. We don’t see coherent
oscillations in electromagnetism because the photon has no mass.
The ability of scalar-field oscillations to have a state of pure potential energy is
what makes inflation possible. When the energy is potential-only, we will show below
that this is like adding to the vacuum energy. The gravitational effects of large V (φ)
are repulsive and can start exponential expansion. In this simple model, the universe is
started in a potential-dominated state, and inflates until the field falls enough that the
kinetic energy becomes important. In practical models, this stage will be associated with
reheating: although weakly interacting, the field does couple to other particles, and
31
its oscillations can generate other particles – thus transforming the scalar-field energy
into energy of a normal radiation-dominated universe.
lagrangians and fields To understand what scalar fields can do for cosmology,
it is necessary to use the more powerful Lagrangian description of the dynamics.
Consider the formulation of classical mechanics in terms of an action principle.
We write the variational equation
Z
δ L dt = 0, (111)
where the Lagrangian L is the difference of the kinetic and potential Renergies,
L = T − V , for some set of particles with coordinates qi (t). The integral L dt is
the action; this equation says that the paths qi (t) travelled by each particle between
given starting and finishing positions are such that the action is extremal (not necessarily
a minimum, even though one often speaks of the principle of least action) with respect
to small variations in the paths. If we take the ‘positions’ qi (t) and ‘velocities’ q̇i (t) to
be independent variables, then expansion of L(q, q̇) in terms of small variations in the
paths and integration by parts leads to Euler’s equation for each particle
d ∂L ∂L
= . (112)
dt ∂ q̇i ∂qi
Let’s briefly revise the derivation. We say that one coordinate (say, the y value for
particle 42) follows a trajectory q(t), with velocity q̇(t). If we perturb the trajectory
by δq = ǫ(t), the velocity is perturbed by δ q̇ = ǫ̇(t). The change to the Lagrangian
is δL = (∂L/∂q)δq + (∂L/∂ q̇)δ q̇; the derivatives treat q and q̇ as independent, even
though the perturbations are related. We now integrate the second term by parts
to get a common factor of ǫ in the integral. The usual parts term of the form [uv]
vanishes
R if Rwe choose ǫ = 0 at either end of the trajectory. The result is of the form
δ L dt = X(t) ǫ(t) dt = 0. Since ǫ(t) is arbitrary, we require X(t) = 0, which is what
gives Euler’s equation.
Notice that we get one Euler equation for each degree of freedom in the
system: 3N in the case of N interacting particles. When the coordinates are Cartesian,
this clearly gives Newton’s law of motion (at least, for the simple case of velocity-
independent forces). The power of the Lagrangian approach is that, having deduced
the action principle, one can choose generalized coordinates to obtain less obvious
equations of motion.
A field may be regarded as a dynamical system, but with an infinite number of
degrees of freedom. How do we handle this? A hint is provided by electromagnetism,
where we are familiar with writing the total energy in terms of a density which, as we
are dealing with generalized mechanics, we may formally call the Hamiltonian density:
ǫ0 E 2 B2
Z Z
H = H dV = + dV. (113)
2 2µ0
Actually, this expression applies in special relativity, but not in general. If we make
a coordinate transformation, the volume element in a multidimensional integral picks
up a factor of the Jacobian determinant, and we want the action to be independent of
coordinate transformations. Therefore, the Lagrangian L (which is an invariant) should
32
√
be multiplied by a factor −g, where g is the determinant of the metric (the square
root is because the metric gets two powers of the transformation matrix). In the FRW
case, g = −R6 , so this factor is important in treating fields in cosmology.
We can now apply the variational principle as before, considering L to be a
function of a general coordinate, φ, which is the field, and the ‘4-velocity’ ∂µ φ. The
argument is similar to the simple Lagrangian case: perturb the space-time trajectory
φ(xµ ), which involves five corrections to L: one from δφ and four from its 4-derivatives.
We integrate these latter terms by parts in the corresponding coordinate, and obtain
the Euler–Lagrange equation
∂ ∂L ∂L
µ
= . (115)
∂x ∂(∂µ φ) ∂φ
The Lagrangian L and the field equations are therefore generally equivalent, although
the Lagrangian arguably seems more fundamental: we can obtain the field equations
given the Lagrangian, but inverting the process is less straightforward.
For a simple example, consider waves on a string. If the density per unit length
is σ, the tension is T , and we call the transverse displacement of the string y, then the
(one-dimensional) Lagrangian density is
L = 12 σ ẏ 2 − 21 T y ′2 (116)
(at least for small displacements). The potential term comes from the work done in
stretching the string. Inserting this in the Euler–Lagrange equation yields a familiar
result:
σ ÿ − T y ′′ = 0. (117)
This
p is just the wave equation, and it tells us that the speed of sound in a plucked string
is T /σ.
For quantum mechanics, we want a Lagrangian that will yield the Klein–Gordon
equation. The required form is
where the only subtlety is that we treat φ and φ∗ as independent variables (rather than
the real and imaginary parts of φ). For a real field (which we will shortly see corresponds
to a neutral particle), the Lagrangian becomes
L = 12 (∂ µ φ∂µ φ − µ2 φ2 ). (119)
gauge and higgs fields Scalar fields manifest themselves in particle physics via
the Higgs mechanism, and it’s helpful to summarise briefly what this is. The
simplest Lagrangian that contains the main features of particle physics is
φ → exp[iθ(xµ )] φ. (121)
33
But inserting this expression in the simple Klein–Gordon Lagrangian generates terms
involving the derivatives of θ(xµ ), and this is why the electromagnetic field must be
there: its purpose is to ‘eat’ the phase variation by simultaneously changing Aµ . This is
an amazing argument: requiring quantum mechanics to be gauge invariant means
that electromagnetism must exist. The kinetic term for electromagnetism must be gauge
invariant, so is built out of the field tensor F µν = ∂µ Aν − ∂ν Aµ . But a mass term for
the photon, proportional to Aµ Aµ would not be gauge-invariant, so the photon must
apparently be massless.
The idea of a massless photon is good for electromagnetism, but what about the
weak interaction? This is mediated by the W and Z bosons, which are like photons
but with a mass of around 90 GeV. This is where Higgs comes in. The potential
V (φ) might be expected to be just a mass term V = m2 |φ|2 , but it can be made more
complicated by adding in a term that looks like a self-interaction: V (φ) = m2 |φ|2 +λ|φ|4 .
In the Higgs mechanism, we take this further by doing something that looks crazy at
first sight and letting the number m2 be negative:
Such a choice yields the potential shown in figure 11, which is named for its resemblance
to a Mexican hat. It forces the system to undergo symmetry breaking so that the
ground state has |φ| =
6 0. If we now define a new field φ = φground +φnew , the Lagrangian
can be rewritten as one in which both the φnew field and the Aµ field appear to have
mass. The Higgs mechanism is therefore greatly admired by particle physicists, since it
allows us to have massive bosons and yet retain the elegant principle of gauge invariance.
Detecting the Higgs is a prime goal for the next generation of particle accelerators. For
cosmology, it provides a strong physical motivation for introducing scalar fields with
unusual potential functions.
φ = u V(φ)
φ
2
φ
1
Figure 11. The Higgs potential for a complex scalar field, V (φ) =
−|µ2 | |φ|2 +λ|φ|4 . When the number λ is chosen to be positive, the ground
state energy occurs along a circle of |φ| =
6 0. The phase of the minimum
is arbitrary, however.
34
noether’s theorem The final ingredient we need before applying scalar fields to
cosmology is to understand that they can be treated as a fluid with thermodynamic
properties like pressure. these properties are derived by a profoundly important general
argument that relates the existence of global symmetries to conservation laws in physics.
In classical mechanics, conservation of energy and momentum arise by considering
Euler’s equation
d ∂L ∂L
− = 0, (123)
dt ∂ q̇i ∂qi
P
where L = i Ti − Vi is a sum over the difference in kinetic and potential energies
for the particles in a system. If L is independent of all the position coordinates qi ,
then we obtain conservation of momentum (or angular momentum, if q is an angular
coordinate): pi ≡ ∂L/∂ q̇i = constant for each particle. More realistically, the potential
will depend on the qi , but homogeneity of space says that the Lagrangian as a whole
will be unchanged by a global translation: qi → qi + dq, where dq is some constant.
Using Euler’s equation, this gives conservation of total momentum:
X ∂L d X
dL = dq ⇒ pi = 0. (124)
i
∂q i dt i
Something rather similar happens in the case of quantum (or classical) field
theory: the existence of a global symmetry leads directly to a conservation law. The
difference between discrete dynamics and field dynamics, where the Lagrangian is a
density, is that the result is expressed as a conserved current rather than a
simple constant of the motion. Suppose the Lagrangian has no explicit dependence
on spacetime (i.e. it depends on xµ only implicitly through the fields and their 4-
derivatives). The algebra is similar to that above, and results in
d ∂L ∂φ d µν
− Lg µν ≡ T = 0. (127)
dxν ∂(∂ ν φ) ∂xµ dxν
This is a conservation law, as we can see by analogy with a simple case like the
conservation of charge. There, we would write
∂µ J µ = ρ̇ + ∇ · j = 0, (128)
where ρ is the charge density, j is the current density, and J µ is the 4-current. We have
effectively four such equations (one for each value of ν) so there must be four conserved
quantities: clearly energy and the four components of momentum. Conservation of 4-
momentum is expressed by T µν , which is the 4-current of 4-momentum. For a simple
fluid, it is just
so now we can read off the density and pressure generated by a scalar field. Note
immediately the important consequence for cosmology: a potential term −V (φ) in the
Lagrangian produces T µν = V (φ)g µν . This is the p = −ρ equation of state characteristic
of the cosmological constant. If we now follow the evolution of φ, the cosmological
‘constant’ changes and we have the basis for models of inflationary cosmology.
35
4 Inflation – II
Topics to be covered:
• Models for inflation
• Slow roll dynamics
• Ending inflation
Most of the main features of inflation can be illustrated using the simplest case of a
single real scalar field, with Lagrangian
L = 12 ∂µ φ ∂ µ φ − V (φ). (130)
It turns out that we can get inflation with even the simple mass potential V (φ) =
m2 φ2 /2, but it is easy to keep things general. Noether’s theorem gives the energy–
momentum tensor for the field as
T µν = ∂ µ φ∂ ν φ − g µν L. (131)
From this, we can read off the energy density and pressure:
If the field is constant both spatially and temporally, the equation of state is then
p = −ρ, as required if the scalar field is to act as a cosmological constant; note that
derivatives of the field spoil this identification.
The equation of motion for the scalar field can be R obtained
√ in several ways.
The most direct route is to write down the action S = L −g d4 x and apply the
Euler–Lagrange equation that arises from a stationary action. This gives
φ̈ + 3H φ̇ − ∇2 φ + dV /dφ = 0. (133)
This is a wave equation similar to the one in flat space. The Hubble drag term
√
3H φ̇ is the main new feature, and it arises directly from −g = R3 (t) for an FRW
model. Apart from this, the wave equation is something that looks just like a classical
oscillator driven by a force −dV /dφ. For inflationary applications, we will generally
ignore the spatial derivatives. This is because ∇φ = ∇comoving φ/R. Since R increases
exponentially, these perturbations are damped away: assuming V is large enough for
inflation to start in the first place, inhomogeneities rapidly become negligible.
In the homogeneous case, a simpler and more physical derivation of the equation
of motion is to appeal to energy conservation:
d ln ρ
= −3(1 + w) = −3φ̇2 /(φ̇2 /2 + V ), (134)
d ln a
following which the relations H = d ln a/dt and V̇ = φ̇V ′ can be used to change variables
to t, and the damped oscillator equation for φ follows.
scalar fields as dark matter One interesting limit of the scalar-field equa-
tion is if the ‘acceleration’ from the potential exceeds the Hubble drag (i.e. the universe
expands sufficiently slowly that this term can be neglected). If we further assume that
36
the potential is mass-like (or at least parabolic near its minimum), then we have the
simple oscillator equation φ̈ + m2 φ = 0, with solution φ = A sin mt (for a suitable origin
of time). The density and pressure are
ρ = m2 A2 /2
(135)
p = (m2 A2 /2) cos 2mt.
Therefore, averaged over many cycles, the oscillating scalar field has the equation of
state of pressureless matter.
It is therefore possible that the cosmological dark matter may take the form of a
light scalar field rather than a supersymmetric relic WIMP. This scalar-field dark matter
is normally considered to be a particle called the Axion, which has some motivation
in particle physics. Notice that the mass can in principle be anything, since the density
depends on m and on the field amplitude, A. In practice, other constraints on the axion
model focus attention on
This is very light dark matter indeed, so surely it should be very hot and fail to make
Ω ∼ 1 by a large factor? This is not so: the axion will act as cold dark matter,
and can have a significant relic density. The answer to the apparent paradox is that
these particles should not be thought of as having been in thermal equilibrium. We
are dealing with a classical field that interacts extremely weakly with ordinary matter.
If this interaction was zero, there would be no prospect of detecting the axion other
than via cosmology. In practice, as with WIMPs, there is some level of interaction,
but the strategy for detection is completely different: the axion can interact with
electromagnetic waves, and the low mass means that microwave frequencies are involved.
There is therefore an active experimental programme searching for the axion using tuned
microwave cavities. The problem is that for sensitivity reasons the bandwidth needs
to be very narrow, and it takes a long time to scan an interesting frequency range:
probably the axion model will be fully explored and ruled out within the next decade –
or it could be detected any day now.
The solution of the equation of motion becomes tractable if we both ignore spatial
inhomogeneities in φ and make the slow-rolling approximation that |φ̈| is
negligible in comparison with |3H φ̇| and |dV /dφ|. Both these steps are required in order
for de Sitter-like behaviour to happen; we have shown above that the vacuum equation
of state only holds if in some sense φ changes slowly both spatially and temporally.
Suppose there are characteristic temporal and spatial scales T and X for the scalar
field; the conditions for inflation are that the negative-pressure equation of state from
V (φ) must dominate the normal-pressure effects of time and space derivatives:
V ≫ φ2 /T 2 , V ≫ φ2 /X 2 , (137)
hence |dV /dφ| ∼ V /φ must be ≫ φ/T 2 ∼ φ̈. The φ̈ term can therefore be neglected in
the equation of motion, which then takes the slow-rolling form for homogeneous fields:
8π 8π
H2 = 2
(φ̇2 /2 + V ) ≃ V, (139)
3mP 3m2P
37
This gives a powerful but simple apparatus for deducing the expansion history of any
inflationary model.
The conditions for inflation can be cast into useful dimensionless forms. The
basic condition V ≫ φ̇2 can now be rewritten using the slow-roll relation as
m2P
ǫ≡ (V ′ /V )2 ≪ 1. (140)
16π
Also, we can differentiate this expression to obtain the criterion V ′′ ≪ ′
√ V /mP . Using
slow-roll once more gives 3H φ̇/mP for the rhs, which is in turn ≪ 3H V /mP because
φ̇2 ≪ V , giving finally
m2P
η≡ (V ′′ /V ) ≪ 1 (141)
8π
√
(using H ∼ V /mP in natural units). These two criteria make perfect intuitive sense:
the potential must be flat in the sense of having small derivatives if the field is to roll
slowly enough for inflation to be possible.
The curse and joy of inflationary modelling is that nothing is known about the inflaton
field φ, nor about its potential. We therefore consider simple classes of possible example
models, with varying degrees of physical motivation.
(a) (b)
V V
φ φ
Figure 12. The two main classes of single-field inflation models: (a)
large-field inflation; (b) small-field inflation. The former is motivated by a
mass-like potential, the latter by something more like the Higgs potential.
If we think about a single field, models can be divided into two basic classes, as
illustrated in figure 12. The simplest are large-field inflation models, in which the
field is strongly displaced from the origin. There is nothing to prevent the scalar field
from reaching the minimum of the potential – but it can take a long time to do so, and
the universe meanwhile inflates by a large factor. In this case, inflation is realized by
means of ‘inertial confinement’. The opposite is when the potential is something like
the Higgs potential, where the gradient vanishes at the origin: this is a model of small-
field inflation. In principle, the field can stay at φ = 0 forever if it is placed exactly
there. One would say that the universe then inhabited a state of false vacuum, as
opposed to the true vacuum at V = 0 (but it is important to be clear that there is no
fundamental reason why the minimum should be at zero density exactly; we will return
to this point).
38
The first inflation model (Guth 1981) was of the small-field type, but large-field
models have tended to be considered more plausible, for two reasons. The first is to
do with initial conditions. If inflation starts from thermal equilibrium at a temperature
TGUT , we expect thermal fluctuations in φ; the potential should generally differ from
4
its minimum by an amount V ∼ TGUT . How then is the special case needed to trap
the potential near φ = 0 to arise? We have returned to the sort of fine-tuned initial
conditions from which inflation was designed to save us. The other issue with simple
small-field models relates to the issue of how inflation ends. This can be viewed as a
form of phase transition, which is continuous or second order in the case of large-field
models. For small-field models, however, the transition to the true vacuum can come
about by quantum tunnelling, so that the transition is effectively discontinuous and first
order. As we will discuss further below, this can lead to a universe that is insufficiently
homogeneous to be consistent with observations.
(2) Power-law inflation. On the other hand, a(t) ∝ tp would suffice, provided
p > 1. The potential required to produce this behaviour is
s !
16π
V (φ) ∝ exp φ . (142)
p m2P
(3) Intermediate inflation. Another simple time dependence that suffices for
inflation is a(t) ∝ exp[(t/t0 )f ]. In the slow-roll approximation, the required
potential here is V (φ) ∝ φ−β , where β = 4(f −1 − 1).
39
V
hybrid inflation One way in which the symmetric nature of the initial condition
for small-field inflation can be made more plausible is to go beyond the space of single-
field inflation. The most popular model in this generalized class is hybrid inflation,
in which there are two fields, with potential
1
V (φ, ψ) = (M 2 − λψ 2 )2 + U (φ) + 21 g 2 ψ 2 φ2 . (143)
4λ
We can think of this as being primarily V (ψ), but with the form of V controlled by the
second field, φ. For φ = 0, we have the standard symmetry-breaking potential; but for
large φ, φ > M/g, the dependence on ψ becomes parabolic. Evolution in this parabolic
trough at large φ can thus naturally lower ψ close to ψ = 0. If this happens, we have
inflation driven by φ as the inflaton, with V (φ) = U (φ) + λM 2 /4. This extra constant
in the potential raises H, so the Hubble damping term is particularly high, keeping the
field from rolling away from ψ = 0 until near to φ = 0.
Hybrid inflation therefore has the ability to make some of the features of
the simplest inflation models seem more plausible, while introducing sufficient extra
complexity that one can try to test the robustness of the predictions of the simple
models. The form of the Lagrangian is also claimed to have some fundamental
motivation (although this has been said of many Lagrangians). As a result, hybrid
inflation is rather popular with inflationary theorists.
criteria for inflation Successful inflation in any of these models requires > 60
e-foldings of the expansion. The implications of this are easily calculated using the
slow-roll equation, which gives the number of e-foldings between φ1 and φ2 as
Z φ2
8π V
Z
N = H dt = − 2 dφ (144)
mP φ1 V ′
For any potential that is relatively smooth, V ′ ∼ V /φ, and so we get N ∼ (φstart /mP )2 ,
assuming that inflation terminates at a value of φ rather smaller than at the start. The
criterion for successful inflation is thus that the initial value of the field exceeds the
Planck scale:
φstart ≫ mP . (145)
40
This is the real origin of the term ‘large-field’: it means that φ has to be large in
comparison to the Planck scale. By the same argument, it is easily seen that this is also
the criterion needed to make the slow-roll parameters ǫ and η ≪ 1. To summarize, any
model in which the potential is sufficiently flat that slow-roll inflation can commence
will probably achieve the critical 60 e-foldings.
It is interesting to review this conclusion for some of the specific inflation models
listed above. Consider a mass-like potential V = m2 φ2 . If inflation starts near the
Planck scale, the fluctuations in V are ∼ m4P and these will drive φstart to φstart ≫ mP
provided m ≪ mP ; similarly, for V = λφ4 , the condition is weak coupling: λ ≪ 1.
Any field with a rather flat potential will thus tend to inflate, just because typical
fluctuations leave it a long way from home in the form of the potential minimum.
This requirement for weak coupling and/or small mass scales near the Planck
epoch is suspicious, since quantum corrections will tend to re-introduce the Planck
scale. In this sense, especially with the appearance of the Planck scale as the minimum
required field value, it is not clear that the aim of realizing inflation in a classical way
distinct from quantum gravity has been fulfilled.
Γ/H 4 ∼ 1. (146)
41
density as it started with, but now in the form of normal matter and radiation – which
starts the usual FRW phase, albeit with the desired special ‘initial’ conditions.
As well as being of interest for completing the picture of inflation, it is essential to
realize that these closing stages of inflation are the only ones of observational relevance.
Inflation might well continue for a huge number of e-foldings, all but the last few
satisfying ǫ, η ≪ 1. However, the scales that left the de Sitter horizon at these early
times are now vastly greater than our observable horizon, c/H0 , which exceeds the de
Sitter horizon by only a finite factor – about e60 for GUT-scale inflation, as we saw
earlier. Realizing that the observational regime corresponds only to the terminal phases
of inflation is both depressing and stimulating: depressing, because φ may well not
move very much during the last phases – our observations relate only to a small piece
of the potential, and we cannot hope to recover its form without substantial a priori
knowledge; stimulating, because observations even on very large scales must relate to
a period where the simple concepts of exponential inflation and scale-invariant density
fluctuations were coming close to breaking down. This opens the possibility of testing
inflation theories in a way that would not be possible with data relating to only the
simpler early phases. These tests take the form of tilt and gravitational waves in the
final perturbation spectrum, to be discussed further below.
We now need to consider the greatest achievement of inflation, which was not anticipated
when the theory was first put forward: it provides a concrete mechanism for generating
the seeds of structure in the universe. In essence, the idea is that the inevitable small
quantum fluctuations in the inflaton field φ are transformed into residual classical
fluctuations in density when inflation is over. The details of this process can be technical,
and could easily fill a lecture course. The following treatment is therefore simplified as
far as possible, while still making contact with the full results.
42
Consider applying this to the case of a uniform universe; here ρ only depends on time,
so that
ρ′ = ρ − ǫ0 ρ̇. (148)
dη = dt/a(t), (150)
where four scalar functions φ, ψ, E and B are involved. It turns out that this situation
can be simplified by defining variables that are unchanged by a gauge transformation:
1
Φ≡φ+ [(B − E ′ )a]′
a
(152)
a′
Ψ≡ψ− (B − E ′ ),
a
where primes denote derivatives with respect to conformal time.
These gauge-invariant ‘potentials’ have a fairly direct physical interpretation,
since they are closely related to the Newtonian potential. The easiest way to evaluate
the gauge-invariant fields is to make a specific gauge choice and work with the
longitudinal gauge in which E and B vanish, so that Φ = φ and Ψ = ψ. A
43
second key result is that inserting the longitudinal metric into the Einstein equations
shows that φ and ψ are identical in the case of fluid-like perturbations where off-diagonal
elements of the energy–momentum tensor vanish. In this case, the longitudinal gauge
becomes identical to the Newtonian gauge, in which perturbations are described
by a single scalar field, which is the gravitational potential. The conclusion is thus that
the gravitational potential can for many purposes give an effectively gauge-invariant
measure of cosmological perturbations, and this provides a sounder justification for the
Newtonian approach that we now adopt. The Newtonian-gauge metric therefore looks
like this:
and this should be quite familiar. If we consider small scales, so that the spatial metric
γij becomes that of flat space, then this form matches, for example, the Schwarzschild
metric with Φ = −GM/r, in the limit Φ/c2 ≪ 1.
Informally, the potential Φ is a measure of space-time curvature which solves the
gauge issue and has meaning on super-horizon scales. A key property, which is perhaps
intuitively reasonable, is that Φ is constant in time for perturbations with wavelengths
much larger than the horizon. Conversely, interesting effects can happen inside the
horizon, which imprints characteristic scale-dependent features on the cosmological
inhomogeneities. A full justification of the constancy of Φ using a complete relativistic
treatment would take too much space in the present course, and we will generally discuss
perturbations using a Newtonian approach. This does yield the correct conclusion
regarding the constancy of Φ, but we should be clear that this is at best a consistency
check, since we will use a treatment of gravity that is restricted to static fields.
ρ − ρ̄
δ≡ . (155)
ρ̄
the factor of a2 is there so we can use comoving length units in ∇2 and the factor
(1 + 3w) accounts for the relativistic active mass density ρ + 3p.
We are very often interested in asking how these fluctuations depend on scale,
which amounts to making a Fourier expansion:
X
δ(x) = δk e−ik·x , (156)
where k is the comoving wavevector. What are the allowed modes? If the field were
periodic within some box of side L, we would have the usual harmonic boundary
conditions
2π
kx = n , n = 1, 2 · · · , (157)
L
and the inverse Fourier relation would be
3 Z
1
δ(x) exp ik · x d3 x.
δk (k) = (158)
L
44
Working in Fourier space in this way is powerful because it immediately gives a way
of solving Poisson’s equation and relating fluctuations in density and potential. For a
single mode, ∇2 → −k 2 , and so
The fluctuating density field can be described by its statistical properties. The
mean is zero by construction; the variance is obtained by taking the volume average of
δ2 :
X
hδ 2 i = |δk |2 . (160)
To see this result, write the lhs instead as hδδ ∗ i (makes no difference for a real field),
and appreciate that all cross terms integrate to zero via the boundary conditions. For
obvious reasons, the quantity
L3
X Z Z
2 2 3
hδ i = |δk | → P (k) d k = ∆2 (k) d ln k. (162)
(2π)3
L3
∆2 (k) ≡ 4πk 3 P (k), (163)
(2π)3
which absorbs the box size into the definition of a dimensionless power spectrum, which
gives the contribution to the variance from each logarithmic range of wavenumber (or
wavelength).
overview It was realized very quickly after the invention of inflation that the theory
might also solve the other big puzzle with the initial condition of the universe. When
we study gravitational instability, we will see that the present-day structure requires
that the universe at even the Planck era would have had to possess a finite degree of
inhomogeneity. Inflation suggests an audacious explanation for this structure, which is
that it is an amplified form of the quantum fluctuations that are inevitable when the
universe is sufficiently small. The present standard theory of this process was worked out
by a number of researchers and generally agreed at a historic 1982 Nuffield conference
in Cambridge.
45
r=c/H
Figure 14. The event horizon in de Sitter space. Particles outside the
sphere at r = c/H can never receive light signals from the origin, nor can
an observer at the origin receive information from outside the sphere. The
exponential expansion quickly accelerates any freely falling observers to
the point where their recession from the origin is effectively superluminal.
The wave trains represent the generation of fluctuations in this spacetime.
Waves with λ ≪ c/H effectively occupy flat space, and so undergo the
normal quantum fluctuations for a vacuum state. As these modes (of fixed
comoving wavelength) are expanded to sizes ≫ c/H, causality forces the
quantum fluctuation to become frozen as a classical amplitude that can
seed large-scale structure.
The essence of the idea can be seen in figure 14. This reminds us that de Sitter
space contains an event horizon, in that the comoving distance that particles can
travel between a time t0 and t = ∞ is finite,
Z ∞
c dt
rEH = ; (164)
t0 R(t)
this is not to be confused with the particle horizon, where the integral would be between
0 and t0 . With R ∝ exp(Ht), the proper radius of the horizon is given by R0 rEH = c/H.
The exponential expansion literally makes distant regions of space move faster than
light, so that points separated by > c/H can never communicate with each other. If we
imagine expanding the inflaton, φ, using comoving Fourier modes, then there are two
interesting limits for the mode wavelength:
(1) ‘Inside the horizon’: a/k ≪ c/H. Here the de Sitter expansion is negligible, just
as we neglect the modern vacuum energy in the Solar system. The fluctuations
in φ can be calculated exactly as in flat-space quantum field theory.
(2) ‘Outside the horizon’: a/k ≫ c/H. Now the mode has a wavelength that exceeds
the scale over which causal influences can operate. Therefore, it must now act as
a ‘frozen’ quantity, which has the character of a classical disturbance. This field
fluctuation can act as the seed for subsequent density fluctuations.
46
Before going any further, we can immediately note that a natural prediction will
be a spectrum of perturbations that are nearly scale invariant. This means that the
metric fluctuations of spacetime receive equal levels of distortion from each decade of
perturbation wavelength, and may be quantified in terms of the dimensionless power
spectrum, ∆2Φ , of the Newtonian gravitational potential, Φ (c = 1):
d σ 2 (Φ)
∆2Φ ≡ = constant ≡ δH2 . (165)
d ln k
The origin of the term ‘scale-invariant’ is clear: since potential fluctuations modify
spacetime, this is equivalent to saying that spacetime must be a fractal: it has the same
level of deviation from the exact RW form on each level of resolution. It is common
to denote the level of metric fluctuations by δH – the horizon-scale amplitude
(which we know to be about 10−5 ). The justification for this name is that the potential
perturbation is of the same order as the density fluctuation on the scale of the horizon
at any given time. We can see this from Poisson’s equation in Fourier space:
a2 a2 2
Φk = − 4πG ρ̄ δ k = −(3/2) H δk (166)
k2 k2
(where we have taken Ω = 1 and a w = 0 pressureless equation of state). This says that
Φk /c2 ∼ δk when the reciprocal of the physical wavenumber is c/H, i.e. is of order the
horizon size.
The intuitive argument for scale invariance is that de Sitter space is invariant
under time translation: there is no natural origin of time under exponential expansion.
At a given time, the only length scale in the model is the horizon size c/H, so it is
inevitable that the fluctuations that exist on this scale are the same at all times. By our
causality argument, these metric fluctuations must be copied unchanged to larger scales
as the universe exponentiates, so that the appearance of the universe is independent of
the scale at which it is viewed.
If we accept this rough argument, then the implied density power spectrum is
interesting. because of the relation between potential and density, it must be
∆2 (k) ∝ k 4 , (167)
So the density field is very strongly inhomogeneous on small scales. Another way of
putting this is in terms of a standard power-law notation for the non-dimensionless
spectrum:
P (k) ∝ k n ; n = 1. (168)
To get a feeling for what this means, consider the case of a matter distribution built
up by the random placement of particles. It is not hard to show that this corresponds
to white noise: a power spectrum that is independent of scale – i.e. n = 0. The
scale-invariant spectrum therefore represents a density field that is super-uniform on
large scales, but with enhanced small-scale fluctuations.
This n = 1 spectrum was considered a generic possibility long before inflation,
and is also known as the Zeldovich spectrum. It is possible to alter this prediction
of scale invariance only if the expansion is non-exponential; but we have seen that such
deviations must exist towards the end of inflation. As we will see, it is natural for n to
deviate from unity by a few %, and this is one of the predictions of inflation.
47
To anticipate the final answer, the inflationary prediction is of a horizon-scale
amplitude
H2
δH = (169)
2π φ̇
which can be understood as follows. Imagine that the main effect of fluctuations is to
make different parts of the universe have fields that are perturbed by an amount δφ. In
other words, we are dealing with various copies of the same rolling behaviour φ(t), but
viewed at different times
δφ
δt = . (170)
φ̇
These universes will then finish inflation at different times, leading to a spread in energy
densities (figure 15). The horizon-scale density amplitude is given by the different
amounts that the universes have expanded following the end of inflation:
H H H H2
δH ≃ H δt = δφ = × = . (171)
φ̇ φ̇ 2π 2π φ̇
The δH ≃ H δt argument relies on R(t) ∝ exp(Ht) and that δH is of order the fractional
change in R. We will not attempt here to do better than justify the order of magnitude.
V φ
δφ
δt
δφ
φ t
Figure 15. This plot shows how fluctuations in the scalar field
transform themselves into density fluctuations at the end of inflation.
Different points of the universe inflate from points on the potential
perturbed by a fluctuation δφ, like two balls rolling from different starting
points. Inflation finishes at times separated by δt in time for these two
points, inducing a density fluctuation δ = Hδt.
The last step uses the crucial input of quantum field theory, which says that the
rms δφ is given by H/2π, and we now sketch the derivation of this result, What we need
to do is consider the equation of motion obeyed by perturbations in the inflaton field.
The basic equation of motion is
φ̈ + 3H φ̇ − ∇2 φ + V ′ (φ) = 0, (172)
and we seek the corresponding equation for the perturbation δφ obtained by starting
inflation with slightly different values of φ in different places. Suppose this perturbation
takes the form of a comoving plane-wave perturbation of comoving wavenumber k and
amplitude A: δφ = A exp(ik · x−ikt/a). If the slow-roll conditions are also assumed, so
that V ′ may be treated as a constant, then the perturbed field δφ obeys the first-order
perturbation of the equation of motion for the main field:
¨ + 3H [δφ]
[δφ] ˙ + (k/a)2 [δφ] = 0, (173)
48
which is a standard wave equation for a massless field evolving in an expanding universe.
Having seen that the inflaton perturbation behaves in this way, it is not much
work to obtain the quantum fluctuations that result in the field at late times (i.e.
on scales much larger than the de Sitter horizon). First consider the fluctuations in
flat space on scales well inside the horizon. In principle, this requires quantum field
theory, but the vacuum fluctuations in φ can be derived by a simple argument using
the uncertainty principle. First of all, note that the sub-horizon equation of motion is
¨ + ω 2 [δφ] = 0, where ω = k/a. For an
just that for a simple harmonic oscillator: [δφ]
oscillator of mass m and position coordinate q, the rms uncertainty in q in the ground
state is
1/2
h̄
qrms = . (174)
2mω
This can be derived immediately from the uncertainty principle, which says that the
minimum uncertainty is
For a classical oscillation with q(t) ∝ eiωt , the momentum is p(t) = mq̇ = iωmq(t).
Quantum uncertainty can be thought of as saying that we lack a knowledge of the
amplitude of the oscillator, but in any case the amplitudes in momentum and coordinate
must be related by prms = mωqrms . The uncertainty principle therefore says
m2 ω 2 qrms
4
= h̄2 /4, (176)
This is the correct expression that results from a full treatment in quantum field theory.
With this boundary condition, it straightforward to check by substitution that
the following expression satisfies the evolution equation:
49
being exp(ik/aH). However, since (d/dt)[k/aH] = −k/a, we see that the oscillatory
term has a leading dependence on t of the desired −kt/a form.
At the opposite extreme, aH/k ≫ 1, the squared fluctuation amplitude becomes
frozen out at the value
H2
h0| |φk |2 |0i = , (181)
2k 3 V
where we have emphasised that this is the vacuum expectation value. The fluctuations
in φ depend on k in such a way that the fluctuations per decade are constant:
2
d (δφ)2 4πk 3 V
H
= h0| |φk |2 |0i = (182)
d ln k (2π)3 2π
(the factor V /(2π)3 comes from the density of states in the Fourier transform, and
cancels the 1/V in the field variance; 4πk 2 dk = 4πk 3 d ln k comes from the k-space
volume element).
This completes the argument. The initial quantum zero-point fluctuations in
the field have been transcribed to a constant classical fluctuation that can eventually
manifest itself as large-scale structure. The rms value of fluctuations in φ can be used
as above to deduce the power spectrum of mass fluctuations well after inflation is over.
In terms of the variance per ln k in potential perturbations, the answer is
H4
δH2 ≡ ∆2Φ (k) =
(2π φ̇)2
8π V (183)
H2 =
3 m2P
3H φ̇ = −V ′ ,
where we have also written once again the slow-roll condition and the corresponding
relation between H and V , since manipulation of these three equations is often required
in derivations.
A few other comments are in order. These perturbations should be Gaussian
and adiabatic in nature. A Gaussian density field is one for which the joint probability
distribution of the density at any given number of points is a multivariate Gaussian.
The easiest way for this to arise in practice is for the density field to be constructed as a
superposition of Fourier modes with independent random phases; the Gaussian property
then follows from the central limit theorem. This holds in the case of inflation, since
modes of different wavelength behave independently and have independent quantum
fluctuations. Adiabatic perturbations are ones where the matter and photon number
densities are perturbed equally, which is the natural outcome of post-inflation reheating
(although not inevitable in multi-field models).
Suppose V (φ) takes the form V = λφ4 , so that N = H 2 /(2λφ2 ). The density
perturbations can then be expressed as
H2 3H 3
δH ∼ = ∼ λ1/2 N 3/2 . (185)
φ̇ V′
50
Since N >
∼ 60, the observed δH ∼ 10
−5
requires
λ<
∼ 10
−15
. (186)
√
Alternatively, in the case of V = m2 φ2 , δH = 3H 3 /(2m2 φ). Since H ∼ V /mP , this
gives δH ∼ mφ2 /m3P ∼ 10−5 . Since we have already seen that φ > ∼ mP is needed for
inflation, this gives
m<
∼ 10
−5
mP . (187)
V 1/2
δH ∼ . (188)
m2P ǫ1/2
We have argued that inflation will end with ǫ of order unity; if the potential were to
4
have the characteristic value V ∼ EGUT then this would give the simple result
2
mGUT
δH ∼ . (189)
mP
tilt Finally, deviations from exact exponential expansion must exist at the end of
inflation, and the corresponding change in the fluctuation power spectrum is a potential
test of inflation. Define the tilt of the fluctuation spectrum as follows:
d ln δH2
tilt ≡ 1 − n ≡ − . (190)
d ln k
We then want to express the tilt in terms of parameters of the inflationary potential, ǫ
and η. These are of order unity when inflation terminates; ǫ and η must therefore be
evaluated when the observed universe left the horizon, recalling that we only observe
the last 60-odd e-foldings of inflation. The way to introduce scale dependence is to write
the condition for a mode of given comoving wavenumber to cross the de Sitter horizon,
a/k = H −1 . (191)
Since H is nearly constant during the inflationary evolution, we can replace d/d ln k by
d ln a, and use the slow-roll condition to obtain
d d φ̇ d m2 V ′ d
=a = =− P . (192)
d ln k da H dφ 8π V dφ
We can now work out the tilt, since the horizon-scale amplitude is
H4 V3
2 128π
δH = = , (193)
(2π φ̇)2 3 m6P V ′2
1 − n = 6ǫ − 2η (194)
51
For most models in which the potential is a smooth polynomial-like function,
|η| ≃ |ǫ|. Since ǫ has the larger coefficient and is positive by definition, the simplest
inflation models tend to predict that the spectrum of scalar perturbations should be
slightly tilted, in the sense that n is slightly less than unity.
It is interesting to put flesh on the bones of this general expression and evaluate
the tilt for some specific inflationary models. This is easy in the case of power-law
inflation with a ∝ tp because the inflation parameters are constant: ǫ = η/2 = 1/p, so
that the tilt here is always
1 − n = 2/p (195)
8π φ φ 4π
Z
60 = 2 dφ = 2 (φ2 − φ2end ). (197)
mP φend α mP α
1 − n = (2 + α)/120. (198)
The predictions of simple chaotic inflation are thus very close to scale invariance in
practice: n = 0.97 for α = 2 and n = 0.95 for α = 4. However, such a tilt has a
significant effect over the several decades in k from CMB anisotropy measurements to
small-scale galaxy clustering. These results are in some sense the default inflationary
predictions: exact scale invariance would be surprising, as would large amounts of tilt.
Either observation would indicate that the potential must have a more complicated
structure, or that the inflationary framework is not correct.
6 Structure formation – I
6.1 Newtonian equations of motion
We have decided that perturbations will in most cases effectively be described by the
Newtonian potential, Φ. Now we need to develop an equation of motion for Φ, or
equivalently for the density fluctuation ρ ≡ (1 + δ)ρ̄. In the Newtonian approach, we
treat dynamics of cosmological matter exactly as we would in the laboratory, by finding
the equations of motion induced by either pressure or gravity. We begin by casting the
problem in comoving units:
x(t) = a(t)r(t)
(199)
δv(t) = a(t)u(t),
where the rhs must be equal to the Hubble flow Hx, plus the peculiar velocity δv = au.
52
The equation of motion follows from writing the Eulerian equation of motion as
ẍ = g0 + g, where g = −∇ ∇ Φ/a is the peculiar acceleration, and g0 is the acceleration
that acts on a particle in a homogeneous universe (neglecting pressure forces to start
with, for simplicity). Differentiating x = ar twice gives
ä
ẍ = au̇ + 2ȧu + x = g0 + g. (201)
a
The unperturbed equation corresponds to zero peculiar velocity and zero peculiar
acceleration: (ä/a) x = g0 ; subtracting this gives the perturbed equation of motion
This equation of motion for the peculiar velocity shows that u is affected by gravitational
acceleration and by the Hubble drag term, 2(ȧ/a)u. This arises because the peculiar
velocity falls with time as a particle attempts to catch up with successively more distant
(and therefore more rapidly receding) neighbours. In the absence of gravity, we get
δv ∝ 1/a: momentum redshifts away, just as with photon energy.
The peculiar velocity is directly related to the evolution of the density field,
through conservation of mass. This is expressed via the continuity equation, which
takes the form
d
ρ0 (1 + δ) = −ρ0 (1 + δ) ∇ · u. (203)
dt
As usual, spatial derivatives are with respect to comoving coordinates:
∇ ≡ a ∇ proper , (204)
d
∇ · u.
δ = −(1 + δ)∇ (206)
dt
∇ Φ/a is denoted by g.
where the peculiar gravitational acceleration −∇
The solutions of these equations can be decomposed into modes either parallel to
g or independent of g (these are the homogeneous and inhomogeneous solutions to the
equation of motion). The homogeneous case corresponds to no peculiar gravity – i.e.
zero density perturbation. This is consistent with the linearized continuity equation,
∇ · u = −δ̇, which says that it is possible to have vorticity modes with ∇ · u = 0 for
53
which δ̇ vanishes, so there is no growth of structure in this case. The proper velocities
of these vorticity modes decay as v = au ∝ a−1 , as with the kinematic analysis for a
single particle.
The first solution, with δ(t) ∝ a(t) is the growing mode, corresponding to the
gravitational instability of density perturbations. Given some small initial seed
fluctuations, this is the simplest way of creating a universe with any desired degree
of inhomogeneity.
jeans scale So far, we have mainly considered the collisionless component. For
the photon-baryon gas, all that changes is that the peculiar acceleration gains a term
from the pressure gradients:
∇ Φ/a − ∇ p/(aρ).
g = −∇ (210)
The pressure fluctuations are related to the density perturbations via the sound speed
∂p
c2s ≡ . (211)
∂ρ
Now think of a plane-wave disturbance δ ∝ e−ik·r , where k and r are in comoving units.
All time dependence is carried by the amplitude of the wave. The linearized equation
of motion for δ then gains an extra term on the rhs from the pressure gradient:
ȧ
δ̈ + 2 δ̇ = δ 4πGρ0 − c2s k 2 /a2 .
(212)
a
This shows that there is a critical proper wavelength, the Jeans length, at which
we switch from the possibility of gravity-driven growth for long-wavelength modes to
standing sound waves at short wavelengths. This critical length is
2π π
r
proper
λJ = proper = cs . (213)
kJ Gρ
It is interesting to work out the value of this critical length. Consider a universe
with a coupled photon-baryon fluid and ignore dark matter (which we can do at high
redshifts, near matter-radiation equality). The sound speed, c2S = ∂p/∂ρ, may be found
by thinking about the response of matter and radiation to small adiabatic compressions:
implying
−1 −1
2 2 9 ρm 2 9 1 + zrad
cS = c 3+ = c 3+ . (215)
4 ρr 4 1+z
54
Here, zrad is the redshift of equality between matter and photons; 1+zrad =
√1.68(1+zeq )
because of the neutrino contribution. At z ≪ zrad , we therefore have cS ∝ 1 + z. Since
ρ = (1 + z)3 3ΩB H02 /(8πG), the comoving Jeans length is constant at
1/2
32π 2
c
λJ = = 50 (ΩB h2 )−1 Mpc. (216)
H0 27ΩB (1 + zrad )
the general case We have solved the growth equation for the matter-dominated
Ω = 1 case. It is possible to cope with other special cases (e.g. matter + curvature)
with some effort. In the general case (especially with a general vacuum having w 6= −1),
it is necessary to integrate the differential equation numerically. At high z, we always
have the matter-dominated δ ∝ a, and this serves as an initial condition. In general, we
can write
where the factor f expresses a deviation from the simple growth law. The case of
matter + cosmological constant is of the most common practical interest, and a very
good approximation to the answer is given by Carroll et al. (1992):
h i−1
f (Ω) ≃ 25 Ωm Ω4/7 1
m − Ωv + (1 + 2 Ωm )(1 +
1
70 Ωv ) . (218)
This is accurate, but still hard to remember. For flat models with Ωm + Ωv = 1, a
simpler approximation is f ≃ Ω0.23
m , which is less marked than f ≃ Ωm
0.65
in the Λ = 0
case. This reflects the more rapid variation of Ωv with redshift; if the cosmological
constant is important dynamically, this only became so very recently, and the universe
spent more of its history in a nearly Einstein–de Sitter state by comparison with an
open universe of the same Ωm .
in Eulerian units.
The special-relativity fluid mechanics is not hard in principle, but we lack the time
to go through it here. The logic is fairly straightforward, since the conservation equation
we need comes from the 4-divergence of the energy-momentum tensor: T.µν /dxµ = 0.
In the rest frame, T µν = diag(ρc2 , p, p, p), so we just need to apply a shift to the lab
frame: T → M̃ · T · M, where M is the matrix of Lorentz transformation coefficients.
An easier way of achieving this is to use manifest covariance:
T µν = (ρ + p/c2 )U µ U ν − pg µν , (220)
where U µ is the 4-velocity. This is clearly a tensor, so we immediately have all the
components when the velocity is non-zero, and it is straightforward to generate the
55
equations of relativistic fluid mechanics. These equations look quite similar to the
nonrelativistic equations, but with extra terms where the pressure is non-negligible. The
main change from the previous treatment come from a factor of 2 from the (ρ + 3p/c2 )
term in Poisson’s equation, and other contributions of the pressure to the relativistic
equation of motion, such as (ρ + p/c2 ) = (4/3)ρ. The resulting evolution equation for δ
has a driving term on the rhs that is a factor 8/3 higher than in the matter-dominated
case
ȧ 32π
δ̈ + 2 δ̇ = Gρ0 δ, (221)
a 3
2
matter domination (a ∝ t2/3 ) : 4πGρ0 =
3t2 (222)
1/2 1
radiation domination (a ∝ t ) : 32πGρ0 /3 = 2 .
t
Every term in the equation for δ is thus the product of derivatives of δ and powers of
t, and a power-law solution is obviously possible. If we try δ ∝ tn , then the result is
n = 2/3 or −1 for matter domination; for radiation domination, this becomes n = ±1.
For the growing
R mode, these can be combined rather conveniently using the conformal
time η ≡ dt/a:
δ ∝ η2 . (223)
The quantity η is proportional to the comoving size of the cosmological particle horizon.
One further way of stating this result is that gravitational potential perturbations
are independent of time (at least while Ω = 1). Poisson’s equation tells us that
−k 2 Φ/a2 ∝ ρ δ; since ρ ∝ a−3 for matter domination or a−4 for radiation, that gives
Φ ∝ δ/a or δ/a2 respectively, so that Φ is independent of a in either case. In other
words, the metric fluctuations resulting from potential perturbations are frozen, at least
for perturbations with wavelengths greater than the horizon size.
but now H 2 = 8πG(ρm + ρr )/3. If we change variable to y ≡ ρm /ρr = a/aeq , and use
the Friedmann equation, then the growth equation becomes
2 + 3y ′ 3
δ ′′ + δ − δ=0 (225)
2y(1 + y) 2y(1 + y)
(for zero curvature, as appropriate for early times). It may be seen by inspection that
a growing solution exists with δ ′′ = 0:
δ ∝ y + 2/3. (226)
56
It is also possible to derive the decaying mode. This is simple in the radiation-dominated
case (y ≪ 1): δ ∝ − ln y is easily seen to be an approximate solution in this limit.
What this says is that, at early times, the dominant energy of radiation drives
the universe to expand so fast that the matter has no time to respond, and δ is frozen
at a constant value. At late times, the radiation becomes negligible, and the growth
increases smoothly to the Einstein–de Sitter δ ∝ a behaviour (Mészáros 1974). The
overall behaviour is therefore reminiscent to the effects of pressure on a coupled fluid,
where growth is suppressed below the Jeans scale. However, the two phenomena are
really quite different. In the fluid case, the radiation pressure prevents the perturbations
from collapsing further; in the collisionless case, the photons have free-streamed away,
and the matter perturbation fails to collapse only because radiation domination ensures
that the universe expands too quickly for the matter to have time to self-gravitate.
This effect is critical in shaping the late-time power spectrum. For scales
greater than the horizon, perturbations in matter and radiation can grow together,
so fluctuations at early times grow at the same rate, independent of wavenumber. But
this growth ceases once the perturbations ‘enter the horizon’ – i.e. when the horizon
grows sufficiently to exceed the perturbation wavelength. At this point, growth ceases,
so the universe preserves a ‘snapshot’ of the amplitude of the mode at horizon crossing.
For a scale-invariant spectrum, this implies a dimensionless power δ 2 (k) ≃ δH2 on small
scales, breaking to the initial δ 2 (k) ∝ k 4 on large scales. Observing this break and using
it to measure the density of the universe has been one of the great success stories in
recent cosmological research.
∂2 2ȧ ∂
L δ = driving term, L≡ 2
+ . (227)
∂t a ∂t
The rhs represents the effects of gravity, and particles will respond to gravity whatever
its source. The coupled equations for several species are thus given by summing the
driving terms for all species.
matter plus radiation The only subtlety is that we must take into account the
peculiarity that radiation and pressureless matter respond to gravity in different ways,
as seen in the equations of fluid mechanics. The coupled equations for perturbation
growth are thus
δm ρm 2ρr δm
L = 4πG . (228)
δr 4ρm /3 8ρr /3 δr
Solutions to this will be simple if the matrix has time-independent eigenvectors. Only
one of these is in fact time independent: (1, 4/3). This is the adiabatic mode
in which δr = 4δm /3 at all times. This corresponds to some initial disturbance in
which matter particles and photons are compressed together. The entropy per baryon is
unchanged, δ(T 3 )/(T 3 ) = δm , hence the name ‘adiabatic’. In this case, the perturbation
amplitude for both species obeys Lδ = 4πG(ρm + 8ρr /3)δ. We also expect the baryons
and photons to obey this adiabatic relation very closely even on small scales: the
tight coupling approximation says that Thomson scattering is very effective
at suppressing motion of the photon and baryon fluids relative to each other.
57
isocurvature modes Since there are two degrees of freedom in the matter-
radiation perturbation, there must be a second independent perturbation mode to
complement the adiabatic solution. This clearly must correspond to a perturbation
in the entropy. If we define S = T 3 /ρm , then
δ ln S = 34 δr − δm . (229)
In the case of adiabatic initial conditions, S is not perturbed, but the overall energy
density fluctuates. The opposite extreme, therefore, is to keep the energy density
constant but allow the entropy to fluctuate. We therefore consider constant density
at some initial time:
ρim δm
i
= −ρir δri . (230)
Because spacetime curvature depends on the overall perturbation to the matter density,
these initial conditions are known as the isocurvature mode. Normally, we will
imagine that whatever may generate such a fluctuation in the equation of state happens
at early times where things are heavily radiation dominated. In this case, the initial
conditions are effectively isothermal:
δri → 0. (231)
4
δr = (δm − δi ), (232)
3
where δi is the initial value of δm .
Subsequent evolution attempts to preserve the initial constant density by making
the matter perturbations decrease while the amplitude of δr increases (this δρ = 0
condition is not satisfied in detail, but we will skip the full solution here). At late
times, δm → 0, while δr → −4δi /3. Hence, as the universe becomes strongly matter
dominated, the entropy perturbation becomes carried entirely by the photons. This
leads to an increased amplitude of microwave-background anisotropies in isocurvature
models, which is one reason why such models are not popular.
It should now be clear why the isothermal perturbation is not a proper mode. At
a general time, it corresponds to a mixture of adiabatic and isocurvature perturbations
and so cannot stay isothermal. Similarly, in the curvaton model, a scalar field with
inflationary fluctuations decays at late time to produce a fluctuating radiation field with
δm = 0. This case yields a mixture of adiabatic and isocurvature modes, such that the
temperature fluctuations are comparable at matter-radiation equality. As with pure
isocurvature fluctuations, this can be rejected with current data. But more complex
models with a small isocurvature fraction are always permitted; one of the tasks of
modern cosmology is to improve the limits on such mixtures to the point where the
initial conditions are proved to be in effect perfectly adiabatic.
58
6.4 Transfer functions and characteristic scales
The above discussion can be summed up in the from of the linear transfer function
for density perturbations, where we factor out the long-wavelength growth law from a
term that expresses how growth is modulated as a function of wavenumber:
While curvature is negligible, we have seen that g(a) is proportional to the square of
conformal time for adiabatic perturbations. In principle, there is a transfer function
for each constituent of the universe, and these evolve with time. As we have discussed,
however, the different matter ingredients tend to come together at late times, and the
overall transfer function tends to something that is the same for all matter components
and which does not change with time for low redshifts. This late-time transfer function
is therefore an important tool for cosmologists who want to predict observed properties
of density fields in the current universe.
We have discussed the main effects that contribute to the form of the transfer
function, but a full calculation is a technical challenge. In detail, we have a mixture of
matter (both collisionless dark particles and baryonic plasma) and relativistic particles
(collisionless neutrinos and collisional photons), which does not behave as a simple fluid.
Particular problems are caused by the change in the photon component from being a
fluid tightly coupled to the baryons by Thomson scattering, to being collisionless after
recombination. Accurate results require a solution of the Boltzmann equation to follow
the evolution of the full phase-space distribution. This was first computed accurately
by Bond & Szalay (1983), and is today routinely available via public-domain codes such
as cmbfast.
Some illustrative results are shown in figure 16. Leaving aside the isocurvature
models, all adiabatic cases have T → 1 on large scales – i.e. there is growth at the
universal rate (which is such that the amplitude of potential perturbations is constant
until the vacuum starts to be important at z <∼ 1). The different shapes of the functions
can be understood intuitively in terms of a few special length scales, as follows:
This process continues until the universe becomes matter dominated. We therefore
expect a characteristic ‘break’ in the fluctuation spectrum around the comoving horizon
length at this time, which we have seen is DH (zeq ) = 16 (Ωm h2 )−1 Mpc. Since distances
in cosmology always scale as h−1 , this means that Ωm h should be observable.
(2) Free-streaming length. This relatively gentle filtering away of the
initial fluctuations is all that applies to a universe dominated by Cold Dark Matter, in
which random velocities are negligible. A CDM universe thus contains fluctuations in
the dark matter on all scales, and structure formation proceeds via hierarchical process
in which nonlinear structures grow via mergers. Examples of CDM would be thermal
relic WIMPs with masses of order 100 GeV, but a more interesting case arises when
thermal relics have lower masses. For collisionless dark matter, perturbations can be
erased simply by free streaming: random particle velocities cause blobs to disperse. At
early times (kT > mc2 ), the particles will travel at c, and so any perturbation that
59
Figure 16. A plot of transfer functions for various adiabatic models,
in which Tk → 1 at small k. A number of possible matter contents are
illustrated: pure baryons; pure CDM; pure HDM. For dark-matter models,
the characteristic wavenumber scales proportional to Ωm h2 , marking
the break scale corresponding to the horizon length at matter-radiation
equality. The scaling for baryonic models does not obey this exactly; the
plotted case corresponds to Ωm = 1, h = 0.5.
has entered the horizon will be damped. This process switches off when the particles
become non-relativistic, so that perturbations are erased up to proper lengthscales of
≃ ct(kT = mc2 ). This translates to a comoving horizon scale (2ct/a during the radiation
era) at kT = mc2 of
(in detail, the appropriate figure for neutrinos will be smaller by (4/11)1/3 since they
have a smaller temperature than the photons). A light neutrino-like relic that decouples
while it is relativistic satisfies
Ων h2 = m/94.1 eV (236)
Thus, the damping scale for HDM (Hot Dark Matter) is of order the bend scale. The
existence of galaxies at z ≃ 6 tells us that the coherence scale must have been below
about 100 kpc, so the DM mass must exceed about 1 keV.
A more interesting (and probably practically relevant) case is when the dark
matter is a mixture of hot and cold components. The free-streaming length for the hot
component can therefore be very large, but within range of observations. The dispersal
of HDM fluctuations reduces the CDM growth rate on all scales below Lfree−stream – or,
relative to small scales, there is an enhancement in large-scale power.
(3) Acoustic horizon length. The horizon at matter-radiation equality
also enters in the properties of the baryon component. Since the sound speed is of order
c, the largest scales that can undergo a single acoustic oscillation are of order the horizon.
The transfer function for a pure baryon universe shows large modulations, reflecting the
number of oscillations that have been completed before the universe becomes matter
60
dominated and the pressure support drops. The lack of such large modulations in
real data is one of the most generic reasons for believing in collisionless dark matter.
Acoustic oscillations persist even when baryons are subdominant, however, and can be
detectable as lower-level modulations in the transfer function. We will say more about
this later.
(4) Silk damping length. Acoustic oscillations are also damped on small
scales, where the process is called Silk damping: the mean free path of photons due to
scattering by the plasma is non-zero, and so radiation can diffuse out of a perturbation,
convecting the plasma with it.√ The typical distance of a random walk in terms of the
diffusion coefficient D is x ≃ Dt, which gives a damping length of
p
λS ≃ λDH , (237)
the geometric mean of the horizon size and the mean free path. Since λ = 1/(nσT ) =
44.3(1 + z)−3 (Ωb h2 )−1 proper Gpc, we obtain a comoving damping length of
This becomes close to the horizon length by the time of last scattering, 1 + z ≃ 1100.
The resulting damping effect can be seen in figure 16 at k ∼ 10kH .
spectrum normalization We now have a full recipe for specifying the matter
power spectrum:
3
Z
σ (R) = ∆2 (k) |Wk |2 d ln k; Wk =
2
(sin kR − kR cos kR). (240)
(kR)3
Unlike the power spectrum, σ(R) is monotonic, and the value at any scale is sufficient
to fix the normalization. The traditional choice is to specify σ8 , corresponding to
R = 8 h−1 Mpc. As a final complication, this measure is normally taken to apply to the
rms in the filtered linear-theory density field. The best current estimate is σ8 ≃ 0.8,
so clearly nonlinear corrections matter in interpreting this number. The virtue of this
convention is that it is then easy to calculate the spectrum normalization at any early
time.
7 Structure formation – II
The equations of motion are nonlinear, and we have only solved them in the limit
of linear perturbations. We now discuss evolution beyond the linear regime, first
considering the full numerical solution of the equations of motion, and then a key
analytic approximation by which the ‘exact’ results can be understood.
n-body models The exact evolution of the density field is usually performed by
means of an N-body simulation, in which the density field is represented by the
sum of a set of fictitious discrete particles. We need to solve the equations of motion for
each particle, as it moves in the gravitational field due to all the other particles. Using
61
comoving units for length and velocity (v = au), we have previously seen the equation
of motion
d ȧ 1
u = −2 u − 2 ∇ Φ, (241)
dt a a
where Φ is the Newtonian gravitational potential due to density perturbations. The time
derivative is already in the required form of the convective time derivative observed by
a particle, rather than the partial ∂/∂t.
In outline, this is straightforward to solve, given some initial positions and
velocities. Defining some timestep dt, particles are moved according to dx = u dt,
and their velocities updated according to du = u̇ dt, with u̇ given by the equation of
motion (in practice, more sophisticated time integration schemes are used). The hard
part is finding the gravitational force, since this involves summation over (N − 1) other
particles each time we need a force for one particle. All the craft in the field involves
finding clever ways in which all the forces can be evaluated in less than the raw O(N 2 )
computations per timestep. We will have to omit the details of this, unfortunately, but
one obvious way of proceeding is to solve Poisson’s equation on a mesh using a Fast
Fourier Transform. This can convert the O(N 2 ) time scaling to O(N ln N ), which is a
qualitative difference given that N can be as large as 1010 .
Computing lives by the ‘garbage in, garbage out’ rule, so how are the
initial conditions in the simulation set? This can be understood by thinking
of density fluctuations in Lagrangian terms (also known as the Zeldovich
approximation). The proper coordinate of a given particle can be written as
where q is the usual comoving position, and the displacement field f (q, t) tends
to zero at t = 0. The comoving peculiar velocity is just the time derivative of this
displacement:
∂f
u= (243)
∂t
so the linear density perturbation δ is just (minus) the divergence of the displacement
field. All this can be handled quite simply if we define a displacement potential:
∇ ψ(q),
f = −∇ (246)
ψk = −δk /k 2 , (247)
62
and hence the displacement and velocity
fk = ik ψk
(248)
uk = ik ψ̇k .
Thus, given the density power spectrum to specify |δk | and the assumption of random
phases, we can set up a field of consistent small displacements and consistent velocities.
These are applied to a uniform particle ‘load’, and then integrated forward into the
nonlinear regime.
the spherical model N -body models can yield evolved density fields that are
nearly exact solutions to the equations of motion, but working out what the results
mean is then more a question of data analysis than of deep insight. Where possible,
it is important to have analytic models that guide the interpretation of the numerical
results. The most important model of this sort is the spherical density perturbation,
which can be analysed immediately using the tools developed for the Friedmann models,
since Birkhoff’s theorem tells us that such a perturbation behaves in exactly the same
way as part of a closed universe. The equations of motion are the same as for the scale
factor, and we can therefore write down the cycloid solution immediately. For a
matter-dominated universe, the relation between the proper radius of the sphere and
time is
r = A(1 − cos θ)
(249)
t = B(θ − sin θ).
This all agrees with what we knew already: at early times the sphere expands with the
a ∝ t2/3 Hubble flow and density perturbations grow proportional to a.
We can now see how linear theory breaks down as the perturbation evolves.
There are three interesting epochs in the final stages of its development, which we can
read directly from the above solutions. Here, to keep things simple, we compare only
with linear theory for an Ω = 1 background.
(1) Turnround. The sphere breaks away from the general expansion and reaches a
maximum radius at θ = π, t = πB. At this point, the true density enhancement
with respect to the background is just [A(6t/B)2/3 /2]3 /r3 = 9π 2 /16 ≃ 5.55.
(2) Collapse. If only gravity operates, then the sphere will collapse to a singularity
at θ = 2π.
63
theorem. Conventionally, it is assumed that this stable virialized radius is
eventually achieved only at the collapse time, at which point the density contrast
is ρ/ρ̄ = (6π)2 /2 ≃ 178 and δlin ≃ 1.686.
These calculations are the basis for a common ‘rule of thumb’, whereby one assumes
that linear theory applies until δlin is equal to some δc a little greater than unity, at
which point virialization is deemed to have occurred. Although the above only applies
for Ω = 1, analogous results can be worked out from the full δlin (z, Ω) and t(z, Ω)
relations. These indicate that δlin ≃ 1 is a good criterion for collapse for any value
of Ω likely to be of practical relevance. The density contrast at virialization tends
to be higher in low-density universes, where the faster expansion means that, by the
time a perturbation has turned round and collapsed to its final radius, a larger density
contrast has been produced. For real non-spherical systems, it is not clear that this
effect is meaningful, and in practice a fixed density contrast of around 200 is used to
define the virial radius that marks the boundary of an object.
press–schechter and the halo mass function What relevance does the
spherical model have to the real world? Despite the lack of spherical symmetry, we
can still use the model to argue that nonlinear collapse will occur whenever we have a
region within which the mean linear-theory density contrast is of order unity. This has
an interesting consequence in the context of the CDM model, where there is power on all
scales: the sequence of structure formation must be hierarchical. This means that we
expect the universe to fragment into low-mass clumps at high redshift, following which
a number of clumps merge into larger units at later times. This process is controlled
by the density variance as a function of smoothing scale, σ 2 (R). In a hierarchical model,
this increases without limit as R → 0, so there is always a critical scale at which σ ≃ 1.
As the density fluctuations grow, this critical scale grows also. These collapsed systems
are known as dark-matter haloes; a name that dates back to the 1970s, when the
existence of extended dark matter around galaxies was first firmly established. The
largest such haloes, forming today, are the rich clusters of galaxies. Galaxy-scale haloes
formed earlier, and this process effectively dictates the era of galaxy formation.
We can improve on this outline, and calculate the distribution of halo masses
that exist at any one time, using a method pioneered by Press & Schechter (1974). If
the density field is Gaussian, the probability that a given point lies in a region with
δ > δc (the critical overdensity for collapse) is
∞
1
Z
exp −δ 2 /2σ 2 (R) dδ,
p(δ > δc | R) = √ (252)
2π δc
where σ(R) is the linear rms in the filtered version of δ. The PS argument now takes
this probability to be proportional to the probability that a given point has ever been
part of a collapsed object of scale > R. This is really assuming that the only objects
that exist at a given epoch are those that have only just reached the δ = δc collapse
threshold; if a point has δ > δc for a given R, then it will have δ = δc when filtered
on some larger scale and will be counted as an object of the larger scale. The problem
with this argument is that half the mass remains unaccounted for: PS therefore simply
multiplying the probability by a factor 2. This fudge can be given some justification,
but we just accept it for now. The fraction of the universe condensed into objects with
mass > M can then be written in the universal form
r Z ∞
2
F (> M ) = exp(−ν 2 /2) dν, (253)
π ν
4π
M= ρ̄ R3 . (254)
3
64
In other words, M is the mass contained in a sphere of comoving radius R in a
homogeneous universe. This is the linear-theory view, before the object has collapsed;
its final virialized radius will be R/2001/3 . The integral collapse probability is related to
the mass function f (M ) (defined such that f (M ) dM is the comoving number density
of objects in the range dM ) via
where (a, b, c) = (1.529, 0.704, 0.412). Empirically, one can use δc = 1.686 independent
of the density parameter. A plot of the mass function according to this prescription is
given in figure 17, assuming what we believe to be the best values for the cosmological
parameters. This shows that the Press-Schechter formula captures the main features of
the evolution, even though it is inaccurate in detail. We see that the richest clusters of
galaxies, with M ≃ 1015 h−1 M⊙ , are just coming into existence now, whereas at z = 5
even a halo with the mass of the Milky Way, M ≃ 1012 h−1 M⊙ was similarly rare. It
can be seen that the abundance of low-mass haloes declines with redshift, reflecting
their destruction in the merging processes that build up the large haloes.
Since the appreciation in the 1970s that galaxies seemed to be embedded in haloes of
dark matter, it has been clear that one should be able to construct an approximate
theory for the assembly of galaxies based on the assumption that everything is
dominated by the dark matter. Therefore, once we understand the history of the
haloes, we should be able to make plausible guesses about how the baryonic material
will behave. Over the years, this route has been followed to the point where there now
exists an elaborate apparatus known as semianalytic galaxy formation. This
is not yet a fully satisfactory theory, in that it is not able to make robust predictions
of the properties that the galaxy population should have. However, it has succeeded
in illuminating the main issues that need to be understood in a complete theory. In
essence, semianalytic models include the following elements:
(1) Merger trees. A halo that exists at a given time will have been constructed
by the merging of smaller fragments over time. We need to be able to predict this history.
(2) Fate of subhaloes. When haloes merge, they do not instantly lose their
identity. Their cores survive as distinct subhaloes for some time. In group/cluster
scale haloes, these will mark the locations of the galaxies. In general, subhaloes will
eventually merge within the parent halo, and sink to the centre. Thus there is always a
tendency to have a dominant central galaxy (e.g. the Milky Way is surrounded by the
much smaller Magellanic Cloud dwarfs).
65
Figure 17. The mass function in the form of the multiplicity
function: fraction of mass in the universe found in virialized haloes per
unit range in ln M . The solid lines show a fitting formula to N -body data
and the dashed lines contrast the original Press-Schechter formula.
66
(3) Accounting of gas and stars. The first generation of haloes is assumed
to start life with gas distributed along with the dark matter in the universal ratio
Ωb /Ωdm . From the density of the gas, the cooling rate can be calculated. Whatever gas
reaches a temperature below 104 K is deemed to be a reservoir of cold gas suitable for
star formation. Some empirical relation based on the amount and density of this gas is
then used to predict a star-formation rate. When haloes merge, their contents of stars,
cold gas, and hot gas are added.
(4) Feedback. As we will show below, the above recipe fails to match
observation, as it predicts that stars should form most efficiently in the smallest galaxies
– so that a system of the size of the Milky Way should be just a collection of globular
clusters, rather than predominately a giant gaseous disk. Therefore, the critical (and
so far unsolved) problem in galaxy formation is to make gas cool less efficiently. The
idea here is that energy is put back into the hot gas as a result of the nonlinear events
that happen inside galaxies. These are principally of two kinds: supernova explosions
and nuclear activity around a central black hole.
virial temperature There is no time here to dig very deeply into the component
parts of this recipe, but a few points are worth making. First consider the characteristic
density of a virialized halo. We have argued that this is some multiple fc ≃ 200 of the
background density at virialization (or ‘collapse’):
ρc = fc ρ0 (1 + zc )3 . (258)
The virialized potential energy for constant density is 3GM 2 /(5r), where the radius
satisfies 4πρc r3 /3 = M . This energy must equal 3M kT /(µmp ), where µ = 0.59 for a
plasma with 75% hydrogen by mass. Hence, using ρ0 = 2.78 × 1011 Ωm h2 M⊙ Mpc−3 ,
we obtain the virial temperature:
Tvirial /K = 105.1 (M/1012 M⊙ )2/3 (fc Ωm h2 )1/3 (1 + zc ). (259)
This is an illuminating expression. It tells us that the most massive systems forming
today, with M ≃ 1015 M⊙ , will have temperatures of 107 – 108 K. The intergalactic
medium in clusters is thus very hot, and emits in X-rays. It also cools very inefficiently,
since such hot plasmas emit only bremsstrahlung. Conversely, pregalactic systems with
M < 9
∼ 10 M⊙ at z ≃ 10 have a virial temperature that is barely at the level of 10 K
4
required for ionization. Their gas is thus predominately neutral, and should form stars
with maximum efficiency. This is the cooling paradox referred to above.
But the same formula allows us to see how to escape from the paradox. The
virial temperature is equivalent to a velocity dispersion, which is essentially the velocity
at which particles orbit in the dark-matter potential well. This velocity therefore also
gives the order of magnitude of the escape velocity for the system. Haloes with a virial
temperature of only ∼ 104 K thus constitute very shallow potential wells and will lose
any of their gas that becomes heated to > 5
∼ 10 K. This is liable to happen as soon as any
supernovae from the first generation of star formation explode. For type II supernovae
associated with massive stars, this can be virtually instantaneous (< 7
∼ 10 years). Star
formation in these early dwarf galaxies might well be expected to be self-quenching.
Indications that this process did happen can be found when measuring HI rotation
curves of dwarfs: the typical baryon fraction is only about 1% (as opposed to something
close to the global 20% in clusters).
8 CMB anisotropies – I
So far, we have concentrated on describing perturbations in the matter density, and
will go on to discuss ways in which these may be observed. But first, we should put in
place the corresponding machinery for the fluctuations in the radiation density. These
can be observed directly in terms of fluctuations in the temperature of the CMB, which
relate to the density fluctuation field at z ≃ 1100. We therefore have the chance to
observe both current cosmic structure and its early seeds. By putting the two together
and requiring consistency, the cosmological model can be pinned down with amazing
precision.
67
8.1 Anisotropy mechanisms
The spherical harmonics are familiar as the eigenfunctions of the angular part of ∇2 , and
there are 2ℓ + 1 modes of given ℓ, hence the notation for the angular power spectrum,
Cℓ . For ℓ ≫ 1, the spherical harmonics become equivalent to Fourier modes, in which
the angular wavenumber is ℓ; therefore one can associate a ‘wavelength’ 2π/ℓ with each
mode.
Once again, it is common to define a ‘power per octave’ measure for the
temperature fluctuations:
T 2 (ℓ) = ℓ(ℓ + 1)Cℓ /2π (262)
2
(although shouldn’t this be ℓ(ℓ + 1/2)? – se later). Note that T (ℓ) is a power per ln ℓ;
the modern trend is often to plot CMB fluctuations with a linear scale for ℓ – in which
case one should really use T 2 (ℓ)/ℓ.
We now list the mechanisms that cause primary anisotropies in the CMB
(as opposed to secondary anisotropies, which are generated by scattering along
the line of sight). There are three basic primary effects, illustrated in figure 19, which
are important on respectively large, intermediate and small angular scales:
v
Doppler 1111
0000 adiabatic
0000
1111
0000
1111
0000
1111
Φ 0000
1111
kT ne 0000
1111
0000
1111
SW
δρ ,
δΦ
θ ISW last
SZ scattering
observer shell
density perturbation σ
68
(1) Gravitational (Sachs–Wolfe) perturbations. Photons from high-density regions at
last scattering have to climb out of potential wells, and are thus redshifted:
δT 1
= (Φ/c2 ). (263)
T 3
The factor 1/3 is a surprise, which arises because Φ has two effects: (i) it redshifts the
photons we see, so that an overdensity cools the background as the photons climb out,
δT /T = Φ/c2 ; (ii) it causes time dilation at the last-scattering surface, so that we seem
to be looking at a younger (and hence hotter) universe where there is an overdensity.
The time dilation is δt/t = Φ/c2 ; since the time dependence of the scale factor is a ∝ t2/3
and T ∝ 1/a, this produces the counterterm δT /T = −(2/3)Φ/c2 .
(2) Intrinsic (adiabatic) perturbations. In high-density regions, the coupling of matter
and radiation can compress the radiation also, giving a higher temperature:
δT δ(zLS )
= , (264)
T 3
δT δv · r̂
= . (265)
T c
δT Φ̇
Z
=2 dt. (266)
T c2
This is twice as large as one might have expected from Newtonian intuition. It would
be nice to be able to give a simple argument for this along the lines of the factor 2
for relativistic light deflection (where the one-line argument is that the gravitational
potential modifies both the time and space parts of the metric, and each contribute
equally to the effective change in the coordinate speed of light). But it is not quite
as easy to make the ISW factor obvious, so we will just accept it here. As we have
seen, the potential Φ stays constant in the linear regime during the matter-dominated
era, as long as Ωm ≃ 1, so the source term for the ISW effect vanishes for much of the
universe’s history. The ISW effect then becomes only important quite near to the last
scattering redshift (because radiation is still important) and al low z (because of Λ).
Other foreground effects are to do with the development of nonlinear structure,
and are mainly on small scales (principally the Sunyaev–Zeldovich effect from IGM
Comptonization). The exception is the effect of reionization; to a good approximation,
this merely damps the fluctuations on all scales:
δT δT
→ exp −τ, (267)
T T
where the optical depth must exceed τ ≃ 0.04, based on the highest-redshift quasars
and the BBN baryon density. As we will see later, CMB polarization data have detected
a signature consistent with τ = 0.1 ± 0.03, implying reionization at z ≃ 10.
69
8.2 Power spectrum
We now need to see how the angular power spectrum of the CMB arises from the
implementation of these effects. The physical separation we have made is useful for
insight, although it is not exactly how things are calculated in practice. We have not
been able to spend time going into the detailed formalism used on CMB anisotropies,
but the exact equations to be integrated are not so terrifying, and it is illuminating to
see them written down.
L ∞
X Z
P2D (kx , ky ) = P3D (kx , ky , kz ) = P3D (k) dkz . (269)
2π −∞
kz
70
In terms of dimensionless power, this is
2 ∞
L
Z
∆22D (K) = 2
2πK P2D (K) = K 2
∆23D (k) dkz /k 3 , (270)
2π 0
where k 2 = K 2 + kz2 . The 2D spectrum is thus a smeared version of the 3D one, but
the relation is pleasingly simple for a scale-invariant spectrum in which ∆23D (k) is a
constant:
δH ≃ 3 × 10−5 . (272)
This immediately determines the large-scale matter power spectrum in the universe
today. We know from Poisson’s equation that the relation between potential and density
power spectra at scale factor a is
This is modified on small scales by the transfer function, but it shows how mass
fluctuations today and CMB anisotropies can be related. As an aside, a more informal
argument in the opposite direction is to say that we can estimate the depth of potential
wells today:
GM Φ v2
v2 ∼ ⇒ ∼ , (275)
r c2 c2
so the potential well of the richest clusters with velocity dispersion ∼ 1000 km s−1 is of
order 10−5 deep. It is therefore no surprise to see this level of fluctuation on the CMB
sky.
Finally, it is also possible with some effort to calculate the full spherical-harmonic
spectrum from the 3D spatial spectrum. For a scale-invariant spectrum, the result is
6
Cℓ = C2 , (276)
ℓ(ℓ + 1)
which is why the broad-band measure of the ‘power per log ℓ’ is defined as
ℓ(ℓ + 1)
T 2 (ℓ) = Cℓ . (277)
2π
71
scale-invariant spectrum
acoustic peak
Silk damping
Ω=0.3
flat Ω=0.3
open
Figure 20. Angular power spectra T 2 (ℓ) = ℓ(ℓ + 1)Cℓ /2π for the
CMB, plotted against angular wavenumber ℓ in radians−1 . For references
to the experimental data, see Spergel et al. (2006). The two lines
show model predictions for adiabatic scale-invariant CDM fluctuations,
calculated using the CMBFAST package (Seljak & Zaldarriaga 1996). These
have (n, Ωm , Ωb , h) = (1, 0.3, 0.05, 0.65) and have respectively Ωv = 1−Ωm
(‘flat’) and Ωv = 0 (‘open’). The main effect is that open models shift the
peaks to the right, as discussed in the text.
δT 1Φ δT 1 δρ
≃ (gravity); ≃ (adiabatic). (278)
T 3 c2 T 3 ρ
In other words, perturbations with wavelengths above the horizon size at last
scattering generate δT /T via gravitational redshift, but on smaller scales it is adiabatic
perturbations that matter.
72
The significance of the main acoustic peak is therefore that it picks out the
(sound) horizon at last scattering. The redshift of last scattering is almost independent
of cosmological parameters at zLS ≃ 1100, as we have seen. If we assume that the
universe is matter dominated at last scattering, the horizon size is
DHLS = 184 (Ωm h2 )−1/2 Mpc. (280)
The angle this subtends is given by dividing by the current size of the horizon (strictly,
the comoving angular-diameter distance to zLS ). Again, for a matter-dominated model,
this is
DH = 6000 Ω−1
m h
−1
Mpc ⇒ θH = DHLS /DH = 1.8 Ω0.5
m degrees. (281)
Figure 20 shows that heavily open universes thus yield a main CMB peak at scales
much smaller than the observed ℓ ≃ 220, and these can be ruled out. Indeed, open
models were disfavoured for this reason long before any useful data existed near the
peak, simply because of strict upper limits at ℓ ≃ 1500 (Bond & Efstathiou 1984). In
contrast, a flat vacuum-dominated universe has DH ≃ 6000 Ω−0.4
m h−1 Mpc, so the peak
is predicted at ℓ ≃ 2π/(184/6000) ≃ 200 almost independent of parameters. These
expression lie behind the common statement that the CMB data require a flat universe
– although it turns out that large degrees of spatial curvature and Λ can also match
the CMB well.
The second dominant scale is imposed by the fact that the last-scattering surface
is fuzzy – with a width in redshift of about δz = 80. This imposes a radial smearing
over scales σr = 7(Ωm h2 )−1/2 Mpc. This subtends an angle
θr ≃ 4 arcmin, (282)
for flat models. This is partly responsible for the fall in power at high ℓ (Silk damping
also contributes). Finally, a characteristic scale in many density power spectra is set by
the horizon at zeq . This is 16(Ωh2 )−1 Mpc and subtends a similar angle to θr .
73
9 CMB anisotropies – II
Having given an outline of the physical mechanisms that contribute to the CMB
anisotropies, we now examine how the CMB is used in conjunction with other probes
to pin down the cosmological model.
The first point to appreciate is that the appearance of the CMB only depends on
a restricted set of combinations of cosmological parameters, and we cannot measure
all parameters at once from the CMB alone. For a given primordial spectrum (i.e.
given ns ), the CMB temperature power spectrum depends on the physical densities
ωm ≡ Ωm h2 , ωb ≡ Ωb h2 . Now, it is possible to vary both Ωv and the curvature to keep
a fixed value of the angular size distance to last scattering, so that the resulting angular
CMB pattern will be invariant. The usual expression for the comoving angular-diameter
distance is
c
R0 Sk (r) = |1 − Ω|−1/2 ×
H0
(286)
"Z #
z
|1 − Ω|1/2 dz ′
Sk p ,
0 (1 − Ω)(1 + z ′ )2 + Ωv + Ωm (1 + z ′ )3
h2 = ωm + ωv + ωk , (288)
horizon-angle degeneracy For flat models, we therefore see that perfect data
on the CMB can in principle determine all the main cosmological parameters. At
present, the CMB is not measured perfectly, and there is in any case a fundamental
limit to how well this can be done. This comes from cosmic variance: observers at
widely spaced locations will see different realizations of the density fluctuation process
74
Figure 21. The CMB geometrical degeneracy, for the fiducial flat
vacuum-dominated model. This says that zero vacuum density can be
tolerated, provided the ‘curvature density’ is negative – i.e. a closed
universe. Since the fiducial value of ωm is 0.133, ωk = −.031 implies that
a Hubble parameter h = 0.32 would be required in this case. Provided we
are convinced that h must be higher than this, vacuum energy is required.
around them. On the finite sky, there is only a finite number of independent modes,
and this limits how well the power spectrum can be measured to a fractional accuracy
1/2
that is just 1/Nmodes . We therefore expect
1/2
2
δCℓ = Cℓ2 2
+ σnoise (289)
2ℓ + 1
(the number of independent modes is half of 2ℓ+1 because the temperature field is real).
In practice, the instrumental σnoise rises at high ℓ, so we expect the large-scale CMB
measurements to be cosmic-variance limited. In the 2006 WMAP data, the crossover
occurs at about ℓ ≃ 400; with luck, ESA’s Planck mission will raise this to beyond 1000.
Given these limitations, the information we gain from the CMB is dominated by
the main acoustic peak at ℓ = 220, and it is interesting to ask what this tells us. We
have argued that the location of this feature marks the angle subtended by the acoustic
horizon at last scattering, which has been given as DHLS = 184 (Ωm h2 )−1/2 Mpc. Using
the current size of the horizon, the angle subtended in a flat model is
DH = 6000 Ω−0.4
m h−1 Mpc ⇒ θH = DHLS /DH ∝ Ω−0.1
m , (290)
75
and this is not perfectly true. The comoving sound horizon size at last scattering is
defined by
Z aLS
1 cS
DS (zLS ) ≡ 1/2
da (291)
H0 Ωm 0 (a + aeq )1/2
where vacuum energy is neglected at these high redshifts; the expansion factor a ≡
(1 + z)−1 and aLS , aeq are the values at last scattering and matter-radiation equality
respectively. In practice, zLS ≃ 1100 independent of the matter and baryon densities,
and cS is fixed by Ωb . Thus the main effect is that aeq depends on Ωm . Dividing by
DH (z = 0) therefore gives the angle subtended today by the light horizon as
Ω−0.1
r r
m aeq aeq
θH ≃ √ 1+ − , (292)
1 + zLS aLS aLS
where zLS = 1100 and aeq = (23900 ωm )−1 . This remarkably simple result captures
most of the parameter dependence of CMB peak locations within flat ΛCDM models.
Differentiating this equation near a fiducial ωm = 0.13 gives
−1/2
∂ ln θH ∂ ln θH 1 aLS
= −0.1; = 2 1+ = +0.25, (293)
∂ ln Ωm ωm ∂ ln ωm Ωm aeq
Thus for moderate variations from a ‘fiducial’ model, the CMB peak multipole
number scales approximately as ℓpeak ∝ Ω−0.15
m h−0.5 , i.e. the condition for constant
CMB peak location is well approximated as
Ωm h3.3 = constant. (294)
It is now clear how LSS data combines with the CMB: Ωm h is the main combination
probed by the matter power spectrum so this approximate degeneracy is strongly broken
using the combined data.
Although the main horizon-scale peak in the power spectrum dominates the appearance
of the CMB, giving degenerate information about cosmological parameters, the fine
detail of the pattern is also important. As the quality of the CMB measurements
improve, more information can be extracted, and the parameter degeneracies are
increasingly broken by the CMB alone. Regarding the structure around the peak, two
physical effects are important in giving this extra information:
(1) Early ISW. We have seen that the transition from radiation domination
to matter domination occurs only just before last scattering. Although we have proved
that potential fluctuations Φ stay constant during the radiation and matter eras (while
vacuum and curvature are negligible), this is not true at the junction, and there is a
small change in Φ during the radiation–matter transition (by a factor 9/10: see chapter
7 of Muhkanov’s book). This introduces an additional ISW effect, which boosts the
amplitude of the peak, especially for models with low Ωm h2 , which brings zeq right
down to zLS (see the first panel of figure 22).
(2) Baryon loading. If we keep the overall matter density fixed but alter
the baryon fraction, the sound speed at last scattering changes. This has the effect
of making a change in the amplitude of the acoustic oscillations beyond the fist peak:
the drop to the second peak is more pronounced if the baryon fraction is high (see the
second panel of figure 22).
Overall, the kind of precision data now delivered by WMAP allows these effects
to be measured, and the degeneracy between Ωm , Ωb and h broken without external
data.
76
Figure 22. The location of the principal peak in the CMB power
spectrum is largely determined by the combination Ωm h3.3 , representing
the scaling of the angular size of the horizon at last scattering. The
two other main characteristics are the rise to the peak, and the fall to
the second and subsequent maxima. The former (the height of the peak
above the Sachs-Wolfe plateau) is influenced by the early-time ISW effect:
the change in gravitational potential associated with the transition from
radiation domination to matter domination This is illustrated in the first
panel, where we fix Ωb h2 and hence the sound speed. For fixed peak
location, higher h gives lower matter density, and hence a higher peak
from the early-time ISW effect (all models are normalized at ℓ = 20). The
second panel shows the influence of varying the baryon density at constant
matter density, where we see that a higher baryon fraction increases the
amplitude of the acoustic oscillations. Thus, if we assume flatness, the
three observables of the peak location and drops to either side suffice to
determine Ωm , Ωb and h. Given data over a broader range of ℓ, the CMB
power spectrum is also sensitive to the tilt, ns .
All of our discussion to date applies to models in which scalar modes dominate. But
we know that gravity-wave metric perturbations in the form of a traceless symmetric
77
tensor hµν are also possible. One of the fascinating features of inflation is that it
predicts that that these modes must be excited to some extent, for the same reason that
fluctuations in scalar fields exist. Recall our discussion of these: small-scale quantum
fluctuations lead to an amplitude δφ = H/2π on horizon exit, which transforms to a
metric fluctuation δH = Hδφ/φ̇. Tensor modes behave similarly, but with some critical
differences. The fluctuations in hµν obey a formula resembling the one for the inflaton
fluctuations – except that h must be dimensionless, whereas φ has dimensions of mass.
On dimensional grounds, then, the formula for the tensor fluctuations is plausible:
But unlike fluctuations in the inflaton, the tensor fluctuation do not affect the progress
of inflation: they already constitute a curvature of spacetime. Once generated, they
play no further part in events and survive to the present day.
These tensor metric distortions are observable via the large-scale CMB
anisotropies, where the tensor modes produce a spectrum with the same scale
dependence as the Sachs–Wolfe gravitational redshift from scalar metric perturbations.
In the scalar case, we have δT /T ∼ φ/3c2 , i.e. of order the Newtonian metric
perturbation; similarly, the tensor effect is
δT
∼ hrms . (296)
T GW
Could the large-scale CMB anisotropies actually be tensor modes? This would be
tremendously exciting, since it would be a direct window into the inflationary era. The
Hubble parameter in inflation is H 2 = 8πGρ/3 ∼ V (φ)/m2P , so that
δT
∼ hrms ∼ H/mP ∼ V 1/2 /m2P . (297)
T GW
A measurement of the tensor modes in the CMB would therefore tell us directly the
energy scale of inflation: Einflation ∼ V 1/4 . This is more direct than the scalar signature,
which was
V 1/2
δH ∼ , (298)
m2P ǫ1/2
(putting in the factor of 16 from an exact analysis). We have argued that ǫ cannot be
too small if inflation is to end, so significant tensor contributions to the CMB anisotropy
are a clear prediction. As a concrete example, consider power-law inflation with a ∝ tp ,
where we showed that ǫ = η/2 = 1/p. In this case,
r = 8(1 − ns ), (300)
so the larger the tilt, the more important the tensors. We will see below that there is
fairly strong evidence for a non-zero tilt with ns ≃ 0.96, so the simplest expectation
would be a tensor contribution of r ≃ 0.3. Of course, this only applies for a large-
field model like power-law inflation; it is quite possible to have a small-field model with
|η| ≫ |ǫ|, in which case there can be tilt without tensors.
An order unity tensor contribution would imply metric distortions at the level
of 10−5 , which might sound easy to detect directly. The reason this is not so is that
78
the small-scale tensor fluctuations are reduced today: their energy density (which is
∝ h2 ) redshifts away as a−4 once they enter the horizon. This redshifting produces a
break in the spectrum of waves, reminiscent of the matter transfer spectrum, so that
the tensor contribution to the CMB declines for ℓ > ∼ 100. This redshifting means that
the present-day metric distortions are more like 10−27 on relevant scales (kHz gravity
waves) than the canonical 10−5 . Even so, direct detection of these relic gravity waves
can be contemplated, but this will be challenging in the extreme. At the current rate of
progress in technology, the necessary sensitivity may be achieved around 2050; but the
signal may be higher than in the simple models, so one should be open to the possibility
of detecting this ultimate probe of the early universe.
the galaxy power spectrum A key aim in observational cosmology has long
been to use the expected feature at the zeq horizon scale to measure the density of the
universe. Data on galaxy clustering is now sufficiently good that this can be done quite
accurately. The measured power spectrum from the 2dF Galaxy Redshift Survey is
contrasted with CDM models (for which |δk |2 ∝ k n Tk2 ) in figure 23. The curvature of
the spectrum is clearly measured, leading to the constraint
79
Figure 23. The galaxy power spectrum from the 2dF Galaxy Redshift
Survey, shown as the contribution to the fractional density variance
per ln k against wavenumber (spatial wavelength is λ = 2π/k). The
data are contrasted with CDM models having scale-invariant primordial
fluctuations (ns = 1) and Ωm h = 0.1, 0.15, 0.2, 0.25, 0.3. The dotted
lines show pure CDM models, whereas the solid lines show the effect of
baryons at the nucleosynthesis level (assuming Ωb = 0.04 and h = 0.7).
Parameter Meaning
ωdm Physical density of dark matter
ωb Physical density of baryons
ωv Physical density of vacuum
w Equation of state of vacuum
ωk Curvature ‘density’
ns Scalar spectral index
r Tensor-to-scalar ratio
nt Tensor spectral index
σ8 Spectrum normalization
τ Optical depth from reionization
fν Neutrino mass fraction
This is frequently reduced to 7 free parameters (ignoring tensors and the neutrino
mass fraction, and assuming w = −1): a scalar CDM universe. In this case, the
interesting parameter to focus on is the curvature. The likelihood of the data given
the model parameters is regarded as a probability density for the parameters, and we
marginalize by integrating this distribution over the uninteresting parameters. This
leaves a probability distribution for the curvature, which is sharply peaked about zero:
80
Figure 24. The basic WMAP3 confidence contours on the key
cosmological parameters for flat scalar-only models (from Spergel et al.
2006).
Table 2. Constraints on the basic 6-parameter model (flat; no tensors) from WMAP
in combination with 2dFGRS in each case.
ingredients (tensors; w 6= −1). But so far these are not required, and there is a very
well specified 6-parameter standard model, as shown in figure 24 and Table 2.
The impressive thing here is the specification of a relatively low optical depth
due to reionization, leading to evidence in favour of ns < 1; exact scale-invariance would
81
need a larger optical depth, and thus stronger large-scale polarization than observed.
The detection of tilt (a roughly 3.5σ rejection of the ns = 1 model) has to be considered
an impressive success for inflation, given that such deviations from scale invariance were
a clear prediction. So should we consider inflation to be proved? Perhaps not yet, but
one is certainly encouraged to look more closely at the tensor signal.
82
Figure 26. The marginalized WMAP3 confidence contours on the
inflationary r − n plane (revised version of plot from Spergel et al. 2006).
One of the most radical conclusions of recent cosmological research has been the
necessity for a non-zero vacuum density. This was detected on the assumption that
Einstein’s cosmological constant, Λ, might contribute to the energy budget of the
universe. But if this ingredient is a reality, it raises many questions about the physical
origin of the vacuum energy; as we will see, a variety of models may lead to something
similar in effect to Λ, and the general term dark energy is used to describe these.
The properties of dark energy can be probed by the same means that we used
to deduce its existence in the first place: via its effect on the expansion history of the
universe. The vacuum density is included in the Friedmann equation, independent of
the equation of state
8πG
Ṙ2 − ρ R2 = −kc2 . (304)
3
At the outset, then we should be very clear that the deduced existence of dark energy
depends on the correctness of the Friedmann equation, and this is not guaranteed.
Possibly we have the wrong theory of gravity, and we have to replace the Friedmann
equation by something else. Alternative models do exist, particularly in the context
of extra dimensions, and these must be borne in mind. Nevertheless, as a practical
framework, it makes sense to stick with the Friedmann equation and see if we can get
consistent results. If this programme fails, we may be led in the direction of more radical
change.
To insert vacuum energy into the Friedmann equation, we need the equation of
state
w ≡ p/ρ c2 (305)
8πGρ
= Ωv a−3(w+1) . (306)
3H02
83
More generally, we can allow w to vary; in this case, we should regard −3(w + 1) as
d ln ρ/d ln a, so that
Z
8πGρ
= Ωv exp −3(w(a) + 1) d ln a . (307)
3H02
Some complete dynamical model is needed to calculate w(a). Given the lack of a unique
model, a common empirical parameterization is
2 −1
trip − t0 ≃ H |1 + w|−1 (1 − Ωm )−1/2 . (310)
3 0
ȧ
δ̈ + 2 δ̇ = 4πGρ0 δ. (312)
a
The vacuum energy manifests itself in the factor of H in the ‘Hubble drag’ term
2(ȧ/a)δ̇. For flat models with w = −1, we have seen that the growing mode for density
perturbations is approximately as g(a) ∝ aΩ(a)0.23 . If w is made more negative, this
makes the growth law closer to the Einstein–de Sitter g(a) ∝ a (for very large negative
w, the vacuum was unimportant until very recently). Therefore, increasing w (making
it less negative) has an effect in the same sense as decreasing Ωm . As shown in figure
27, the degeneracy between variations in Ωm and w thus has the opposite sign to the
degeneracy in D(z). Ideally, one would therefore try to observe both effects.
84
Figure 27. Perturbation around Ωm = 0.25 of distance-redshift and
growth-redshift relations. Solid line shows the effect of increase in w;
dashed line the effect of increase in Ωm
What are the best ways to measure w? We have seen that the two main signatures
are alterations to the distance-redshift relation and the perturbation growth rate. It is
possible to use both of these effects in the framework we have been discussing: observing
the perturbed universe in both the CMB and large-scale structure.
In the CMB, the main observable is the angle subtended by the horizon at last
scattering
This has the approximate scaling with cosmological parameters (for a flat universe)
85
The latter term comes from a convenient approximation for the current horizon size:
c −α(w)
D0 = 2 Ω . (315)
H0 m
At first sight, this looks bad: the single observable of the horizon angle depends on three
parameters (four, if we permit curvature). Thus, even in a flat model, we can only pin
down w if we know both Ωm and h.
However, if we have more detail on the CMB than just the main peak location,
then we have seen that the Ωm − h degeneracy is weakly broken, and that this situation
improves with information from large-scale structure, which yields an estimate of Ωm h.
In effect, we have two constraints on the Ωm −h plane that are consistent if w = −1, but
this is not the case for other values of w. In this way, the current combined constraints
from CMB plus alternative probes (LSS and the Supernova Hubble diagram) yield an
impressive accuracy:
+0.054
w = −0.926−0.053 , (316)
for a spatially flat model – see Spergel et al. (2006). The confidence contours are plotted
in detail in figure 28, and it is clear that so far there is very good consistency with a
simple cosmological constant. But as we will see, plenty of models exist in which some
deviation is predicted. The next goal of the global cosmology community is therefore
to push the errors on w down substantially – to about 1%. There is no guarantee that
this will yield any signal, but certainly it will cut down the range of viable models for
dark energy.
86
One of the future tools for improving the accuracy in w will be large-scale
structure. We have seen how this helps pin down the parameter degeneracies inherent in
a CMB-only analysis, but it also contains unique information from the acoustic horizon.
Earlier, we approximated this without considering how the speed of sound would depend
on the baryon density; a good approximation to the exact result is
This forms a standard measuring rod, as seen in the ‘baryon wiggles’ in the galaxy power
spectrum. In future galaxy surveys, the measurement of this signature as a function of
redshift will be a further useful geometrical probe.
The amplitude constraint from the CMB has been harder to implement.
Although WMAP provides an accurately determined normalization, it involves the
uncertain optical depth due to reionization:
The value of τ is constrained by large-angle polarization data, and lies close to 0.1
according to WMAP. The current accuracy would be useful if we had an accurate
independent estimate of σ8 . This can be attempted using the abundance of clusters of
galaxies and also gravitational lensing, although the test is not really properly mature
as yet.
11.3 Quintessence
The simplest physical model for dynamical vacuum energy is a scalar field. We
know from inflationary models that this can yield something close in properties to a
cosmological constant, and so we can immediately borrow the whole apparatus for
modelling vacuum energy at late times. This idea of scalar fields as a dynamical
substitute for Λ was first explored by Ratra & Peebles (1988). Of course, this means yet
another scalar field that is introduced without much or any motivation from fundamental
physics. This hypothetical field is given the fanciful name ‘quintessence’, implying a new
addition to the ancient Greek list of elements (fire, air, earth, water).
The Lagrangian density for a scalar field is as usual of the form of a kinetic minus
a potential term:
L = 12 ∂µ φ ∂ µ φ − V (φ). (319)
V (φ) = 1
2 m2 φ2 , (320)
where m is the mass of the field. However, as before we keep the potential function
general at this stage.
Suppose the Lagrangian has no explicit dependence on spacetime (i.e. it depends
on xµ only implicitly through the fields and their 4-derivatives). Noether’s theorem then
gives the energy–momentum tensor for the field as
T µν = ∂ µ φ∂ ν φ − g µν L. (321)
From this, we can read off the energy density and pressure:
87
If the field is constant both spatially and temporally, the equation of state is then
p = −ρ, as required if the scalar field is to act as a cosmological constant; note that
derivatives of the field spoil this identification.
For a homogeneous field, we have the equation of motion
φ̈ + 3H φ̇ + dV /dφ = 0, (323)
d ln ρ
= −3(1 + w) = −3φ̇2 /(φ̇2 /2 + V ), (324)
d ln a
following which the relations H = d ln a/dt and V̇ = φ̇V ′ can be used to change variables
to t, and the damped oscillator equation for φ follows. The solution of the equation of
motion becomes tractable if we make the slow-rolling approximation that |φ̈| is
negligible in comparison with |3H φ̇| and |dV /dφ|, so that
From this, we know that a sufficiently flat potential can provide a dynamical vacuum
that is arbitrarily close to a cosmological constant in its equation of state. However,
there are good reasons why we might want to imagine the slow-roll conditions being
violated in the case of dark energy.
ρφ = φ̇2 /2 + V
(326)
pφ = φ̇2 /2 − V.
This gives us two extreme equations of state: (i) vacuum-dominated, with V ≫ φ̇2 /2,
so that p = −ρ; (ii) kinetic-dominated, with V ≪ φ̇2 /2, so that p = ρ. In the first
case, we know that ρ does not alter as the universe expands, so the vacuum rapidly
tends to dominate over normal matter. In the second case, the equation of state is the
unusual Γ = 2, so we get the rapid behaviour ρ ∝ a−6 . If a quintessence-dominated
universe starts off with a large kinetic term relative to the potential, it may seem that
things should always evolve in the direction of being potential-dominated. However,
this ignores the detailed dynamics of the situation: for a suitable choice of potential, it
is possible to have a tracker field, in which the kinetic and potential terms remain
in a constant proportion, so that we can have ρ ∝ a−α , where α can be anything we
choose.
Putting this condition in the equation of motion shows that the potential is
required to be exponential in form. The Friedmann equation with ρ ∝ a−α requires
a ∝ t2/α , so we have ρ ∝ t−2 as usual. But now both V and φ̇2 must scale in the same
88
way as ρ, so that φ̇ ∝ 1/t. Both the φ̈ and 3H φ̇ terms are therefore proportional to
V , so an exponential potential solves the equation of motion. More importantly, we
can generalize to the case where the universe contains scalar field and ordinary matter.
Suppose the latter obeys ρm ∝ a−α ; it is then possible to have the scalar-field density
obeying the same ρ ∝ a−α law, provided
They show that these can yield an evolution in w(t) so that it switches from w ≃ 1/3
in the radiation era to w ≃ −1 today.
However, a degree of fine-tuning is still required, in that the trick only works
for M ∼ 1 meV, so there is no natural reason for tracking to cease at matter-
radiation equality. The idea of tracker fields thus does not remove completely the
puzzle concerning the level of present-day vacuum energy. But such models are at least
testable: because the Λ-like behaviour only switches on quite recently, it is hard to
complete the transition, and the prediction is of something around w ≃ −0.8 today. As
we have seen, this can be firmly ruled out with current data. These ideas about the
dynamical vacuum are therefore already interesting testable science.
k-essence In a sense, quintessence is only half the story. We started with the
usual Lagrangian for a simple massive scalar field, L = φ̇2 /2 − m2 φ2 /2 and generalized
the quadratic mass term to an arbitrary potential, V (φ). Why not take the same
liberties with the kinetic term? Even though such k-essence models lack the intuitive
analogies of quintessence, a Lagrangian can be anything we like. The simplest models
try to express things in terms of the normal kinetic expression
ρ = 2XL,X − L
(331)
P =L
f
w= . (332)
2Xdf /dX − f
89
For a normal kinetic term, this gives w = +1 if there is no potential. The equation of
motion is derived just by writing conservation of energy as for quintessence:
d ln ρ
= −3(1 + w). (333)
d ln a
What sort of k-essence Lagrangian will yield tracking? We want to fix w at the
value of the dominant component, which requires
d ln f
= (1 + 1/w)/2 ⇒ f (X) ∝ X (1+1/w)/2 . (334)
d ln X
Thus, a Lagrangian proportional to the square of the usual kinetic term will produce
tracking during the radiation era, but tracking in the matter era requires a step to
f (X) = 0 to be encountered just as the universe becomes matter dominated. This
is the opposite to the case of quintessence: now fine-tuning would be required in
order for tracking to be maintained. The real question is whether a simple model can
achieve sufficiently strong departure from tracking to get somewhere close to w = −1
in the matter era in an inevitable way. This seems to be controversial: Armendariz-
Picon, Mukhanov & Steinhardt (0006373) claimed that it could be done, but Malquarti,
Copeland & Liddle (0304277) disagreed. The issue, as with quintessence, is the extent
to which a tracking solution arises inevitably independent of initial conditions – i.e.
whether it is an attractor. This has certainly not been demonstrated.
∂p
c2s = . (335)
∂ρ
∂p/∂X
c2s = (336)
∂ρ/∂X
i.e. ignoring perturbations in the field. The justification for this is that a gauge freedom
exists, and that δφ = 0 corresponds to the rest-frame of the vacuum fluid.
This means that, for quintessence, the sound speed is always cs = c. Even a
completely flat potential with initial condition φ̇ = 0 does not mimic a cosmological
constant. This only happens if the Lagrangian is set up completely lacking any kinetic
term. The low sound speeds in some k-essence models can have quite large effects on
the CMB anisotropies, and so can be probed observationally beyond just w and its
evolution.
Whether or not one finds the ‘essence’ approach compelling, there remains one big
problem. All the models are constructed using Lagrangians with a particular zero
level. All quintessence potentials tend to zero for large fields, and k-essence models
lack a potential altogether. They are therefore subject to the classical dilemma of the
cosmological constant: adding a pure constant to the Lagrangian has no affect on field
dynamics, but mimics a cosmological constant. With so many possible contributions to
this vacuum energy from the zero-point energies of different fields (if nothing else), it
seems contrived to force V (φ) to asymptote to zero without a reason.
90
This leads us in the direction of anthropic arguments, which are able to limit Λ to
some extent: if the universe had become vacuum-dominated at z > 1000, gravitational
instability would have been impossible – so that galaxies, stars and observers would not
have been possible (Weinberg 1989). Indeed, Weinberg made the astonishingly prescient
prediction on this basis that a non-zero vacuum density would be detected at Ωv of order
unity, since there was no reason for it to be much smaller.
many universes At first sight, this argument seems quite appealing, but it rapidly
leads us into deep waters. How can we talk about changing Λ? It has the value that
it has. We are implicitly invoking an ensemble picture in which there are many
universes with differing properties. This is a big step (although exciting, if this turns
out to be the only way to explain the vacuum level we see). In fact, the idea of an
ensemble emerges inevitably from the framework of inflationary cosmology, since the
fluctuations in the scalar field can affect the progress of inflation itself. We have used
this idea to look at the changes in when inflation ends – but fluctuations can affect the
field at all stages of its evolution. They can be thought of as adding a random-walk
element to the classical rolling of the scalar field down the trough defined by V (φ). In
cases where φ is too close to the origin for inflation to persist for sufficiently long, it is
possible for the quantum fluctuations to push φ further out – creating further inflation
in a self-sustaining process. This is the concept of stochastic eternal inflation
due to Linde. Sufficiently far from the origin, the random walk effect of fluctuations
becomes more marked and can overwhelm the classical downhill rolling. This means that
some regions of space can inflate for an indefinite time, and a single inflating universe
automatically breaks up into different bubbles with their own histories. Some random
subset of these eventually random-walk close enough to the origin that the classical end
of inflation can occur, thus creating a set of ‘universes’ each of which can potentially
host observers.
With this as a starting point, the question now becomes whether we can arrange
for the different members of this ensemble to have different values of Λ. This is easily
achieved. Let there be some quintessence field with a very flat potential, so that it
is capable of simulating Λ effectively. Quantum fluctuations during inflation can also
displace this field, so that each member of the multiverse would have a different Λ.
91
where we need both the prior distribution of Λ between different members of the
ensemble and how the chance of getting an observer is modified by Λ. The latter factor
should be proportional to the number of stars, which is generally take to be proportional
to the fraction of the baryons that are incorporated into nonlinear structures. We
can estimate this using the Press-Schechter apparatus to get the collapse fraction into
systems of a galaxy-scale mass. The exact definition of this is not very important, since
the CDM power spectrum is very flat on small scales: any mass at all close to 1012 M⊙
gives similar answers.
The more difficult part is the prior distribution of Λ, and a common argument
is to say that it has a uniform distribution – which seems reasonable enough if we are
to allow it to have either sign, but know that we will be interested in a very small
range near zero. This is the startling proposition of the anthropic model: the vacuum
density takes large ranges, and in almost all realizations, the values are comparable in
magnitude to the natural scale m4P ; such models are stupendously inimical to life.
We therefore have the simple model
where fc is the collapse fraction into galaxy-scale objects. For large values of Λ, growth
ceases at high redshift, and fc is exponentially suppressed. But things are less clear-cut
if Λ < 0. Here the universe eventually recollapses, and the high density means that the
collapse fraction always tends to unity. So why do we not observe Λ < 0? The answer
is that we have to cut off the calculation at late stages of recollapse: once the universe
becomes too hot, star-formation may be affected and in any case there is little time for
life to form.
With this proviso, figure 29 shows the posterior distribution of Λ conditional on
the existence of observers in the multiverse. We express things in natural units: if we
adopt the values Ωv = 0.75 and h = 0.73 for the key cosmological parameters, then
4
−27 −3 h̄ Ev
ρv = 7.51 × 10 kg m = , (339)
c h̄c
So is the anthropic explanation the correct one? Many people find the hypothesis
too radical: why postulate an infinity of universes in order to explain a detail of one of
them? Certainly, if an alternative explanation for the ‘why now’ problem existed in the
form of e.g. a naturally successful quintessence model, one might tend to prefer that.
But so far, there is no such alternative. The longer this situation persists, the more
we will be forced to accept that the universe we see can only be understood by making
proper allowance for our role as observers.
92
Figure 29. The collapse fraction as a function of the vacuum density,
which is assumed to give the relative weighting of different models. The
dashed line for negative density corresponds to the expanding phase only,
whereas the solid lines for negative density include the recollapse phase,
up to maximum temperatures of 10 K, 20 K, 30 K.
93
Advanced Cosmology: Summer 2007
A.1.
(a) Consider the universe at a GUT-scale temperature of kT = 1015 GeV. As-
suming a standard radiation-dominated model, estimate the comoving size of the
particle horizon at the GUT era (currently, kT = 10−3.6 eV). [5]
(b) Write down the integral for the current size of the particle horizon. If the
universe begins with a phase of inflation, during which the Hubble parameter, H,
is constant, how long must inflation continue in order to reconcile the GUT-scale
horizon with the present value of ∼ c/H0 ? [5]
(c) Write down the equation of motion obeyed by a homogeneous scalar field, φ,
commenting on the physical significance of the terms that appear. Explain how
the equation can be solved in the slow-roll approximation. [5]
(d) For a potential V (φ) ∝ φα , show that the solution to the slow-roll equation
is
φ/φinitial = (1 − t/tfinal )1/β ,
where β = 2 − α/2. Explain why this equation does not hold near φ = 0. [10]
(a) Show that, for a static model, density fluctuations grow exponentially as long
as the wavelength is sufficiently large, and explain physically why this is so. [5]
(b) Derive the solutions to the perturbation equation for a universe of critical
density, in the limit of infinitely long wavelength. [5]
1
(d) If the universe contains a homogeneous component in addition to matter that
can clump, the perturbation equation still applies – but ρ̄m does not include the
homogeneous component. Consider a flat universe that contains a mixture of hot
and cold dark matter: for sufficiently small wavelengths, the hot component can
be assumed to be uniform, with density parameter√Ωh . Show that fluctuations
in the cold component grow as δ ∝ tα , where α = ( 25 − 24Ωh − 1)/6. [8]
A.3.
(a) Write brief notes on the mechanisms that contribute to the anisotropies of
the cosmic microwave background. Include a sketch of the temperature power
spectrum and indicate the angular scales where each effect is important. [5]
(c) If the universe is reionized by UV radiation from galaxies and quasars, photons
from the CMB can be scattered by foreground matter. If the transition from
neutral to ionized is assumed to happen suddenly at redshift zc (≫ 1), calculate
the optical depth to last scattering as a function of zc (the Thomson cross-section
is σt = 6.652 × 10−29 m2 and the proton mass is mp = 1.673 × 10−27 kg). [10]
(d) Explain briefly why the effects of reionization on the CMB power spectrum
tend to be degenerate with tilt. [5]