1) The growth of the first stars was halted by ultraviolet radiation feedback from the stars themselves. 2) Radiation from the protostar evaporated the circumstellar accretion disk when the star's mass reached 43 times the mass of the Sun. 3) These massive primordial stars may help explain the lack of pair-instability supernovae signatures in ancient metal-poor stars.

1. Protostellar Feedback Halts the Growth of the First Stars in the
Universe
Takashi Hosokawa, et al.
Science 334, 1250 (2011);
DOI: 10.1126/science.1207433
This copy is for your personal, non-commercial use only.
If you wish to distribute this article to others, you can order high-quality copies for your
colleagues, clients, or customers by clicking here.
Permission to republish or repurpose articles or portions of articles can be obtained by
following the guidelines here.
The following resources related to this article are available online at
Downloaded from www.sciencemag.org on December 1, 2011
www.sciencemag.org (this infomation is current as of December 1, 2011 ):
Updated information and services, including high-resolution figures, can be found in the online
version of this article at:
http://www.sciencemag.org/content/334/6060/1250.full.html
Supporting Online Material can be found at:
http://www.sciencemag.org/content/suppl/2011/11/09/science.1207433.DC1.html
This article cites 60 articles, 5 of which can be accessed free:
http://www.sciencemag.org/content/334/6060/1250.full.html#ref-list-1
This article has been cited by 1 articles hosted by HighWire Press; see:
http://www.sciencemag.org/content/334/6060/1250.full.html#related-urls
This article appears in the following subject collections:
Astronomy
http://www.sciencemag.org/cgi/collection/astronomy
Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the
American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright
2011 by the American Association for the Advancement of Science; all rights reserved. The title Science is a
registered trademark of AAAS.

2. REPORTS
tostar was calculated directly from the mass in-
flux through the sink-cell boundary, whereas the
Protostellar Feedback Halts the Growth luminosity from the protostar, which controls
the radiative feedback, was calculated consistent-
ly from the protostellar model by using the de-
of the First Stars in the Universe rived accretion rate.
At its birth, a very small protostar of ∼0.01
Takashi Hosokawa,1,2* Kazuyuki Omukai,2 Naoki Yoshida,3 Harold W. Yorke1 M⊙ was surrounded by a molecular gas envelope
of ∼1 M⊙, which was quickly accreted onto the
The first stars fundamentally transformed the early universe by emitting the first light and by protostar. Atomic gas further out initially had
producing the first heavy elements. These effects were predetermined by the mass distribution of too much angular momentum to be accreted di-
the first stars, which is thought to have been fixed by a complex interplay of gas accretion and rectly, and a circumstellar disk formed. The in-
protostellar radiation. We performed radiation-hydrodynamics simulations that followed the growth falling atomic gas first hit the disk plane roughly
of a primordial protostar through to the early stages as a star with thermonuclear burning. The vertically at supersonic velocities. A shock front
circumstellar accretion disk was evaporated by ultraviolet radiation from the star when its mass was formed; behind the shock, the gas cooled and
43 times that of the Sun. Such massive primordial stars, in contrast to the often-postulated settled onto the disk, and its hydrogen was con-
extremely massive stars, may help explain the fact that there are no signatures of the pair-instability verted to the molecular form via rapid gas-phase
Downloaded from www.sciencemag.org on December 1, 2011
supernovae in abundance patterns of metal-poor stars in our galaxy.
heoretical studies and detailed computer the ionized region in its vicinity grows, and even-
T simulations show that the cradles of the
first stars were dense concentrations of
primordial gas, with masses of ~1000 that of the
tually the circumstellar disk is directly exposed
to the stellar ionizing radiation. The gas on the
disk surface is photoionized and heated and evap-
Sun. Such gas clouds formed through radiative orates away from the star-disk system. A semi-
cooling, with hydrogen molecules at the center analytical model of this process predicts that this
of a dark matter halo of 106 solar mass (M⊙), photoevaporation quenches the accretion flow
when the age of the universe was a few hundred of the disk material and puts an end to the stellar
million years (1). growth (8). However, the interplay between the
According to our current understanding of accretion flow and the stellar radiation is high-
star formation, a gas cloud’s dense core grav- ly dynamical and complex.
itationally contracts in a nonhomologous run- To identify the exact mechanism that halts
away fashion, in which the densest parts become the growth of a first star and to determine its final
denser faster than does the rest of the cloud (2–5). mass, we applied a method used for studying
In a primordial gas cloud, one or a few embryo the present-day massive star formation (9, 10)
protostars are formed near the center (2, 6). The to the case of the formation of the first stars in a
initial mass of these embryo protostars is only proper cosmological context. We followed the
≅0.01 M⊙; the bulk of the dense core material radiation hydrodynamic evolution in the vicin-
remains in the surrounding envelope and is sub- ity of an accreting protostar, incorporating ther-
sequently drawn toward the protostar (or proto- mal and chemical processes in the primordial gas
stars) through gravity. With the typical angular in a direct manner. We also followed the evolu-
momentum of dense cores, the centrifugal barrier tion of the central protostar self-consistently by
allows only a small amount of infalling gas to solving the detailed structure of the stellar inte-
accrete directly onto the star. Instead, a circum- rior with zero metallicity as well as the accretion
stellar disk is formed, and the gas is accreted flow near the stellar surface [supporting online
onto the central star through the disk (7). The material (SOM) text] (11, 12). We configured the
final mass of these first stars is fixed when the initial conditions by using the results of a three-
mass accretion terminates. However, when and dimensional (3D) cosmological simulation, which
how this termination occurs are largely unknown. followed the entire history from primeval density
Because the luminosity increases rapidly with fluctuations to the birth of a small seed protostar
protostellar mass, radiative feedback is expected at the cosmological redshift 14 (6). Specifical-
to regulate the mass accretion and ultimately shut ly, when the maximum particle number density
off the accretion flow, setting the final mass of reached 106 cm−3 in the cosmological simulation,
the first stars. A primordial star more massive than we considered a gravitationally bound sphere of
Fig. 1. (A to D) Evolution of the stellar radius,
a few tens times the mass of the Sun radiates a radius 0.3 pc around the density peak, which en- bolometric luminosity, evolutionary time scales, and
copious amount of hydrogen-ionizing photons closed a total gas mass of ≅300 M⊙. We reduced ionizing (hν ≥ 13.6 eV) and dissociating (11 eV ≤
(>13.6 eV). As an accreting star grows in mass, the 3D data to an axisymmetric structure by av- hν ≤ 13.6 eV) photon number luminosity. In (B),
eraging over azimuthal angles. the total luminosity Ltot (black) is the sum of the
1
Jet Propulsion Laboratory, California Institute of Technology, The system was evolved until the central stellar luminosity L* (red) and accretion luminosity
Pasadena, CA 91109, USA. 2Department of Physics, Kyoto Uni- particle number density reached 1012 cm−3. We
versity, Kyoto 606-8502, Japan. 3Institute for the Physics and Lacc (blue). The KH time scale tKH (red) and accre-
Mathematics of the Universe, Todai Institutes for Advanced then introduced a sink cell of size ≅10 astro- tion time scale tacc (blue) are depicted in (C). The
Study, University of Tokyo, Kashiwa, Chiba 277-8568, Japan. nomical units (AU) and followed the subsequent yellow and blue backgrounds denote the adiabatic
*To whom correspondence should be addressed. E-mail: evolution of the central protostar with a stellar accretion phase and KH contraction phase in the
hosokwtk@gmail.com evolution code. The accretion rate onto the pro- protostellar evolution.
1250 2 DECEMBER 2011 VOL 334 SCIENCE www.sciencemag.org

3. REPORTS
three-body reactions. The molecular disk ex- In the early accretion phase, the stellar radi- (UV) flux rapidly rose (Fig. 1D). Thus, ionization
tended out to ≅400 AU from the protostar, when us remained almost constant at ≅50 solar radius and heating by UV photons became important
the stellar mass was 10 M⊙. Accretion onto the (R⊙) (Fig. 1A). The stellar luminosity was sub- already in the KH contraction stage.
protostar proceeded through this molecular disk stantially lower than the accretion luminosity When the stellar mass was 20 M⊙, an ionized
as angular momentum was transported outward. (Fig. 1B), and the KH time scale was much longer region rapidly expanded in a bipolar shape per-
The accretion rate onto the protostar was ≅1.6 × than the accretion time (Fig. 1C). Consequent- pendicular to the disk, where gas was cleared
10−3 M⊙ year−1 at that moment. ly, entropy carried by the accreted gas accumu- away (Fig. 2A). At this moment, the disk ex-
The evolution of the central protostar is de- lated at the stellar surface nearly without loss. tended out to ≅600 AU. The disk was self-shielded
termined by competition between mass growth During this quasi-adiabatic stage (M* < 7 M⊙), against the stellar H2-dissociating (11.2 eV ≤ hν ≤
by accretion and radiative energy loss from the the luminosity L* increased with stellar mass. 13.6 eV) as well as the ionizing radiation. The
stellar interior. The time scale for the former When the star grew to 8 M⊙, the KH time scale ionized atomic hydrogen (H II) region continued
˙
is the accretion time scale tacc ≡ M*/M , where finally fell below the accretion time scale (Fig. to grow and finally broke out of the accreting
˙
M* is the mass of the protostar and M is the 1C). After this, the protostar began its so-called envelope. At M* ≅ 25 M⊙, the size of the bipolar
mass accretion rate, whereas that for the latter KH contraction, in which it gradually contracted H II region exceeded 0.1 pc (Fig. 2B). Because
is the Kelvin-Helmholtz (KH) time scale tKH ≡ as it radiated its energy away (Fig. 1A). The stel- of the high pressure of the heated ionized gas,
G M*2/R*L*, where L* is the luminosity from lar luminosity was the main component of the the opening angle of the ionized region also
the stellar interior, R* is the radius of the pro- total luminosity after this evolutionary stage (Fig. increased as the star grew (Fig. 2C). Shocks
tostar, and G is the gravitational constant. The 1B). The stellar luminosity L* increased, and propagated into the envelope preceding the ex-
Downloaded from www.sciencemag.org on December 1, 2011
total luminosity of the protostar Ltot is the sum stellar radius R* decreased, as the stellar mass pansion of the ionized region. The shocked gas
of the stellar luminosity L* and accretion lumi- increased. As a result, the stellar effective tem- was accelerated outward at a velocity of several
˙
nosity Lacc ≡ GM*M/R* (13). perature Teff º (L*/R*2)1/4 and the ultraviolet kilometers per second. The shock even reached
regions shielded against direct stellar UV irra-
diation. The outflowing gas stopped the infall
of material from the envelope onto the disk (fig.
S7). Without the replenishment of disk mate-
rial from the envelope, the accretion rate onto
the protostar decreased (Fig. 3). In addition,
the absence of accreting material onto the cir-
cumstellar disk means that the disk was ex-
posed to the intense ionizing radiation from
the star. The resulting photoevaporation of disk
gas also reduced the accretion rate onto the
protostar. The photoevaporated gas escaped to-
ward the polar direction within the ionized re-
gion. The typical velocity of the flow was several
tens of kilometers per second, comparable with
the sound speed of the ionized gas, which was
high enough for the evaporating flow to escape
from the gravitational potential well of the dark
matter halo.
When central nuclear hydrogen burning first
commenced at a stellar mass of 35 M⊙, it was via
the proton-proton (pp)–chain normally associ-
ated with low-mass stars. The primordial mate-
rial does not have the nuclear catalysts necessary
for carbon–nitrogen–oxygen (CNO)–cycle hy-
drogen burning. Because the pp-chain alone can-
not produce nuclear energy at the rate necessary
to cover the radiative energy loss from the stel-
lar surface, the star continues to contract until
central temperatures and densities attain values
that enable the 3-a process of helium burning
(11). The product of helium burning is carbon,
and once the relative mass abundance of carbon
reaches ∼10−12, CNO-cycle hydrogen burning
takes over as the principal source of nuclear en-
ergy production, albeit at much higher central
densities and temperatures than those of stars
with solar abundances. These first-generation
ZAMS (zero-age main sequence) stars are thus
Fig. 2. UV radiative feedback from the primordial protostar. The spatial distributions of gas temper- more compact and hotter than are their present-
ature (left), number density (right), and velocity (right, arrows) are presented in each panel for the central day counterparts of equal mass (14). The subse-
regions of the computational domain. The four panels show snapshots at times when the stellar mass quent evolution of the accreting star followed
is M* = 20 M⊙ (A), 25 M⊙ (B), 35 M⊙ (C), and 42 M⊙ (D). The elapsed time since the birth of the along the ZAMS mass-radius relationship (Fig.
primordial protostar is labeled in each panel. 1A). By the time the star attained 40 M⊙, the
www.sciencemag.org SCIENCE VOL 334 2 DECEMBER 2011 1251

4. REPORTS
Fig. 3. Evolution of the 0.01 stantially. Thus, our radiation-hydrodynamic cal-
accretion rate onto the pri- culations predict final masses systematically lower
Acc. Rate: M ( M / yr )
mordial protostar. The blue than those of the semianalytic models.
line indicates the evolu- Although the results described above provide
tion, which includes the a complete picture of how a primordial protostar
effect of UV radiative feed- 0.001
No feedback regulates and terminates its growth, there are a
back from the protostar. (a) few key quantities that determine the strength of
The red line indicates a nu- (b) the feedback effect, as suggested by the semi-
merical experiment with (c)
analytic model (8). With smaller initial rotation
no UV feedback. The open of the natal dense core, the density in the en-
and solid circles denote 1e-4 (d)
velope near the polar directions would be higher.
the characteristic epochs feedback Hydrogen recombination occurs rapidly in the
of the protostellar evolu-
dense gas, which prevents the breakout of the
tion, beginning of the KH
contraction, and the pro- 0 10 20 30 40 50 60 ionized region. Gas accretion can last for a longer
tostar’s arrival to the ZAMS. time in this case and would form more massive
Stellar Mass: M ( M )
Fig. 2, A to D, shows the stars (SOM text). Nevertheless, even considering
snapshots at the moments marked here with asterisks. variations among dark halos bearing the first
stars, a substantial fraction of the first stars should
Downloaded from www.sciencemag.org on December 1, 2011
be less massive than 100 M⊙ and end their lives
as ordinary core-collapse supernovae. Gas ac-
entire region above and below the disk (Fig. 2D) than the first stars because of a different gas ther- cretion might not be completely halted in a few
was ionized. Mass accretion was terminated when mal evolution, with additional radiative cooling exceptional cases, thus leading to the formation
the stellar mass was 43 M⊙ (Fig. 3). via H2 and HD molecules (21, 22). However, this of a small number of extremely massive stars
The entire evolution described above took mode of star formation is suppressed even with that are >100 M⊙ in the early universe (26). Black
about 0.1 million years from the birth of the em- weak H2 photodissociating background radia- holes left after such very massive stars’ deaths
bryo protostar to the termination of the accretion. tion (22). If so suppressed, the formation process might have grown up to be the supermassive
The star is expected to live another few million of the later-generation primordial stars would black holes lying in galaxies.
years before exhausting all available nuclear fuel be similar to that of the very first stars, and most Recent 3D cosmological simulations showed
and exploding as a core-collapse supernova (15). primordial stars could have experienced the evo- that a primordial gas cloud breaks up into several
Our calculations show that the first stars reg- lution presented in this article regardless of their embryo protostars in an early phase (27, 28).
ulated their growth by their own radiation. They generations. One might argue that today’s ob- Each of these protostars would continue to grow
were not extremely massive, but rather similar in served metal-deficient stars formed after an epi- through mass accretion, but it remained uncertain
mass to the O-type stars in our Galaxy. This re- sode of star formation with nonzero metallicity. how and when the growth is halted. Our radiation-
solves a long-standing enigma regarding the el- For the star formation process to differ substan- hydrodynamics calculations explicitly show that
emental abundance patterns of the Galactic oldest tially from that of the very first stars, one would the parent gas cloud is evaporated by intense
metal-poor stars, which contain nucleosynthetic require metallicities in excess of [Fe/H] > −5 radiation from the central star when its mass is
signatures from the earliest generation of stars. (23, 24). Caffau and co-workers (25) report on ob- several tens of solar masses. The circumstellar
If a substantial number of first stars had masses servations of a metal-deficient star with [Fe/H] ≅ disks in our simulations are marginally stable
in excess of 100 M⊙, they would end their lives −5, but without the corresponding enhancement against gravitational fragmentation (fig. S5). We
through pair-instability supernovae (16, 17), ex- of carbon, nitrogen, and oxygen found in metal- expect that these disks—in a 3D simulation—
pelling heavy elements that would imprint a deficient stars. The abundance pattern of this star would evolve analogously to our numerical sim-
characteristic nucleosynthetic signature to the agrees with expectations from core-collapse su- ulations with assumed axial symmetry and have a
elemental abundances in metal-poor stars. How- pernovae, implying that it formed from gas en- similar time-averaged structure. It is conceivable
ever, no such signatures have been detected in hanced by material ejected from primordial stars that a few protostars are ejected dynamically
the metal-poor stars in the Galactic halo (18, 19). with masses less than 100 M⊙. from the parent cloud, to remain as low-mass
For example, the abundances of elements with We have performed radiation-hydrodynamic stars. Observationally, however, there have been
odd atomic numbers are generally reduced in simulations only for a single star-forming region no low-mass zero-metallicity stars discovered in
remnants of primordial supernovae. The odd- embedded in a cosmological simulation. Our se- the Galaxy. This fact suggests the limited for-
even contrast pattern expected in pair-instability lected dark halo was typical in mass, spin, and mation efficiency of such low-mass primordial
supernovae is much stronger than the observed formation epoch when compared with those in stars (19). Low-mass stars (<1 M⊙) and extreme-
patterns in Galactic metal-poor stars (17). More- other studies (5). The evolution presented here is ly high mass stars (>100 M⊙), if any, are thus a
over, pair-instability supernovae predict a small somewhat similar to that predicted by the semi- minor population among the first stars.
abundance ratio [Zn/Fe], but observed values are analytic model, in which the expansion of the Our self-consistent calculations show that
much larger (16). Detailed spectroscopic studies ionized region begins when the stellar mass is the characteristic mass of the first stars is several
of extremely metal-deficient stars indicate that ≅25 M⊙ and the final mass is ≅57 M⊙ (8), the tens of solar masses. Although this is less than
the metal-poor stars were born in an interstellar lowest final stellar mass predicted by the semi- that of the conventionally proclaimed extremely
medium that had been metal-enriched by super- analytic treatment. If input parameters of the massive stars (>100 M⊙), it is still much larger
novae of ordinary massive stars (20). semianalytic models are chosen to fit our initial than the characteristic mass of stars in our galaxy
Second-generation stars, which formed from gas cloud, however, the final mass should be (<1 M⊙) (29).
the primordial gas affected by radiative or me- higher, ≅90 M⊙ (SOM text). Our calculations fol-
chanical feedback from the first stars, could have low the dynamical response of the infalling gas References and Notes
dominated the metallicity of the young inter- onto the circumstellar disk. The expansion of 1. V. Bromm, N. Yoshida, L. Hernquist, C. F. McKee, Nature
459, 49 (2009).
stellar medium, which then spawned the observed the ionized region around the protostar generates 2. K. Omukai, R. Nishi, Astrophys. J. 508, 141 (1998).
Galactic halo stars. These second-generation stars a powerful outflow even behind the surrounding 3. T. Abel, G. L. Bryan, M. L. Norman, Science 295, 93
could have been more numerous but less massive disk. This effect reduces the accretion rate sub- (2002).
1252 2 DECEMBER 2011 VOL 334 SCIENCE www.sciencemag.org

5. REPORTS
4. N. Yoshida, K. Omukai, L. Hernquist, T. Abel, Astrophys. J. 19. A. Frebel, J. L. Johnson, V. Bromm, Mon. Not. R. Astron. manuscript. T.H. appreciates the support by Fellowship
652, 6 (2006). Soc. 392, L50 (2009). of the Japan Society for the Promotion of Science for
5. B. W. O’Shea, M. L. Norman, Astrophys. J. 654, 66 20. N. Iwamoto, H. Umeda, N. Tominaga, K. Nomoto, K. Maeda, Research Abroad. The present work is supported in part by
(2007). Science 309, 451 (2005). the grants-in-aid by the Ministry of Education, Science and
6. N. Yoshida, K. Omukai, L. Hernquist, Science 321, 669 21. B. W. O'Shea, T. Abel, D. Whalen, M. L. Norman, Culture of Japan (19047004, 2168407, 21244021:KO,
(2008). Astrophys. J. 628, L5 (2005). 20674003:NY). Portions of this research were conducted at
7. H. W. Yorke, P. Bodenheimer, Astrophys. J. 525, 330 (1999). 22. N. Yoshida, K. Omukai, L. Hernquist, Astrophys. J. 667, the Jet Propulsion Laboratory, California Institute of Technology,
8. C. F. McKee, J. C. Tan, Astrophys. J. 681, 771 (2008). L117 (2007). which is supported by NASA. Data analysis was (in part)
9. H. W. Yorke, A. Welz, Astron. Astrophys. 315, 555 (1996). 23. [Fe/H] ≡ log(Fe/H)star − log(Fe/H) sun, where (Fe/H) is carried out on the general-purpose PC farm at Center for
10. H. W. Yorke, C. Sonnhalter, Astrophys. J. 569, 846 (2002). mass ratio of Fe to H. Computational Astrophysics (CfCA) of National Astronomical
11. K. Omukai, F. Palla, Astrophys. J. 589, 677 (2003). 24. T. Hosokawa, K. Omukai, Astrophys. J. 703, 1810 (2009). Observatory of Japan.
12. T. Hosokawa, K. Omukai, Astrophys. J. 691, 823 (2009). 25. E. Caffau et al., Nature 477, 67 (2011).
13. This definition of accretion luminosity includes the 26. T. Ohkubo, K. Nomoto, H. Umeda, N. Yoshida, S. Tsuruta, Supporting Online Material
mechanical luminosity of an accretion-driven wind, and Astrophys. J. 706, 1184 (2009). www.sciencemag.org/cgi/content/full/science.1207433/DC1
Materials and Methods
the ratio tKH/tacc is equal to the ratio Lacc/L*. 27. M. J. Turk, T. Abel, B. O’Shea, Science 325, 601
14. D. Ezer, A. G. W. Cameron, Astrophys. Space Sci. 14, 399 (2009). SOM Text
Figs. S1 to S7
(1971). 28. P. C. Clark et al., Science 331, 1040 (2011).
Tables S1 and S2
15. D. Schaerer, Astron. Astrophys. 382, 28 (2002). 29. G. Chabrier, Publ. Astron. Soc. Pac. 115, 763 (2003).
16. H. Umeda, K. Nomoto, Astrophys. J. 565, 385 (2002). References (30–64)
17. A. Heger, S. E. Woosley, Astrophys. J. 567, 532 (2002). Acknowledgments: We thank T. Nakamura, K. Nomoto, 25 April 2011; accepted 28 October 2011
18. J. Tumlinson, A. Venkatesan, J. M. Shull, Astrophys. J. S. Inutsuka, and N. Turner for stimulating discussions on this Published online 10 November 2011;
612, 602 (2004). topic. Comments by an anonymous referee helped improve the 10.1126/science.1207433
Downloaded from www.sciencemag.org on December 1, 2011
of two counter-oscillating sublattices within the
Entangling Macroscopic Diamonds diamond structure. The optical phonons are mac-
roscopic, persistent excitations distributed over
at Room Temperature ∼1016 atoms within the crystals. The phonons have
a very high carrier frequency of 40 THz owing
K. C. Lee,1* M. R. Sprague,1* B. J. Sussman,2 J. Nunn,1 N. K. Langford,1 X.-M. Jin,1,3 to the extremely strong interactions between neigh-
T. Champion,1 P. Michelberger,1 K. F. Reim,1 D. England,1 D. Jaksch,1,3 I. A. Walmsley1† boring atoms, giving rise to a mechanically stiff
lattice. This large energy, compatible with the ul-
Quantum entanglement in the motion of macroscopic solid bodies has implications both for trashort pulses in our experiment (bandwidth
quantum technologies and foundational studies of the boundary between the quantum and ∼7 THz), also eliminates the need for cooling or
classical worlds. Entanglement is usually fragile in room-temperature solids, owing to strong optical pumping, because thermal excitations are
interactions both internally and with the noisy environment. We generated motional entanglement negligible at room temperature. The specific ex-
between vibrational states of two spatially separated, millimeter-sized diamonds at room perimental protocol that we use is based on the
temperature. By measuring strong nonclassical correlations between Raman-scattered photons, well-known DLCZ scheme (12) and previous
we showed that the quantum state of the diamonds has positive concurrence with 98% probability. pioneering experiments in cold atomic ensem-
Our results show that entanglement can persist in the classical context of moving macroscopic bles (13–16). We first create a phonon via spon-
solids in ambient conditions. taneous Raman scattering from a strong optical
pump pulse, an event that is simultaneously ac-
ur intuition about the nature of the phys- approach is to design systems with well-defined companied by the emission of a Stokes photon
O ical world is strongly conditioned by
the experience that macroscopic solids
move according to the rules of classical mechan-
and long-lived normal modes that can be selec-
tively excited, and then to cool them to remove
thermal noise and isolate them from the environ-
(red-shifted from the pump). After this interac-
tion, the joint state of the diamond and the
Stokes mode can be written as
ics. Quantum theory, however, asserts that super- ment. Substantial progress has been made toward  Ã
positions and entanglement are possible even for demonstrating strong quantum signatures in larger ys ≈ 1 þ es s† (ts )b† (ts ) vac ð1Þ
large objects. Therefore, exploration of the persist- systems—for example, in optomechanical (3–5),
ence of quantum correlations in the traditionally molecular (6), and superconducting (7) systems. where |es|2 1 is the scattering probability
classical realm is important for both fundamental Mechanical oscillators can now be cooled to the and |vac〉 = |vacopt〉 ⊗ |vacvib〉 is the joint op-
science and technology, because of the impli- thermal ground state (7–9). tical and vibrational vacuum state containing
cations for physics beyond conventional quan- A different approach is required, however, no photons and no phonons; s and b are the
tum mechanics (1) and for quantum information to reveal quantum features in the motion of “or- bosonic annihilation operators for the Stokes
processing, which requires sustained coherence dinary” solids in the high-entropy environment and phonon modes, respectively, evaluated at
across many particles (2). present at ambient conditions. Without a spe- the time ts when the pump pulse exits the dia-
The two main barriers for creating superpo- cially engineered system, measurements must be mond. Equation 1 describes an entangled state
sitions and entanglement in the mechanical mo- made on time scales shorter than the character- of optical and material modes, which is already
tion of macroscopic systems are strong internal istically fast coherence decay times of a real-world nonclassical. To entangle two diamonds, we simul-
interactions, which complicate the dynamics, and system. This can be achieved with ultrashort op- taneously pump two separate crystals—producing
strong coupling with the environment, which tical pulses. Recent studies of quantum coherence the state |Ys〉 = |ysL〉|ysR〉, where L and R denote
leads to short decoherence times. The standard in biology have used ultrafast probes and inter- the left and right diamonds, respectively—and
ference measurements to establish the persist- we combine their Stokes modes on a polarizing
1
Clarendon Laboratory, University of Oxford, Parks Road, Oxford ence of quantum behavior in naturally occurring beamsplitter (Fig. 1). The Stokes modes are caused
OX1 3PU, UK. 2National Research Council of Canada, Ottawa, bulk systems (10, 11). To probe entanglement, to interfere with relative phase ϕs by means of a
Ontario K1A 0R6, Canada. 3Centre for Quantum Technologies,
National University of Singapore, Singapore. however, it is also necessary to access the cor- half-wave plate and polarizer. Detection of a
*These authors contributed equally to this work.
relations of the excited modes. Stokes photon at a detector Ds placed behind
†To whom correspondence should be addressed. E-mail: We study excitations of the optical phonon the polarizer corresponds to application of the
walmsley@physics.ox.ac.uk mode in diamond, a bulk vibration consisting measurement (17) E = 〈vacopt|s(t′s), where t′s is
www.sciencemag.org SCIENCE VOL 334 2 DECEMBER 2011 1253