Plasmonics
Plasmonics
Plasmonics
hot carriers in plasmonic nanoparticles: when quantization does and
does not matter?
Jacob B Khurgin (1), Uriel Levy (2)
(1)
Department of Electrical and Computer Engineering, Johns Hopkins University, Baltimore, Maryland
21218, United States
(2)
Department of Applied Physics, The Benin School of Engineering and Computer Science, The Center
for Nanoscience and Nanotechnology, The Hebrew University of Jerusalem, Jerusalem, 91904, Israel
Abstract: Plasmon‐assisted hot carrier processes in metal nanoparticles can be described either classically
or using the full strength of quantum mechanics. We reconfirm that from the practical applications point
of view, when it comes to description of the decay of plasmons in nanoparticles, classical description is
sufficiently adequate for all but the smallest of the nanoparticles. At the same time, the electron
temperature rise in nanoparticles is discrete (quantized) and neglecting this fact can lead to significant
underestimating of hot carrier assisted effects, such as photocatalysis.
Introduction
The field of plasmonics[1, 2], active since early 21st century, have seen most recently two extremely
interesting developments, both prompted by the realization that due to the inherently high losses [3]
occurring in the metal, plasmonic devices will always remain at a disadvantage when it comes to many
“conventional” nanophotonic devices, be that interconnects, sources, modulators, and so on. Therefore,
it has become imperative to find a niche (or niches) where plasmonics will be competitive with existing
dielectric‐based nanophotonics. One such niche involves using so‐called hot carriers generated by
plasmon decay [4, 5] for photo‐detection [6], energy harvesting[7], or for initiation of chemical reactions
[8, 9]. In these applications, the strong loss in metal, which is typically thought of as the major drawback
of plasmonics, is turned into its strength. The other niche where plasmonics may thrive and where activity
is rapidly growing is a broadly‐defined area of “quantum plasmonics”[10‐12]. The operational principle of
quantum plasmonic devices involves single quanta (photons or plasmons), and the relatively large loss
can be tolerated as long as significant increase in the speed of plasmon generation is attained[13]. The
developments in these two niches have been smoothly proceeding along two parallel tracks, but now and
then the claims of “quantum phenomena” have made their way into the description of hot carriers’ action
1
[14, 15]. Since “quantum effects” are rather loosely defined, these claims can always be admissible, as
every known phenomenon in the microscopic world can indeed be accurately described using the laws of
quantum theory. But, from practical point of view, it is far more important to address an entirely different
question of whether invoking the laws of quantum physics improves one’s understanding of hot carrier
phenomena? What is even more crucial is to determine when and where these phenomena cannot be
adequately described classically and quantum theory becomes indispensable? This work is an attempt to
answer the last question.
As of today, there already exists an imposing body of work [16‐18] in which quantum theory
(ranging from particle in the box [19] approach all the way to density functional [20]) has been applied to
surface plasmon polaritons (SPPs) on nanoparticles of various shapes and sizes. Practically all these
theoretical works [21, 22] as well as experimental data [23, 24] indicate that while quantum corrections
do matter, they do not change the picture qualitatively and for the most part can be taken into account
by introducing size‐dependent damping constant [23, 25]. There is little need to replicate these
calculations. The present effort, not being a review article, is intended to neither pay tribute to all the
prior works nor to criticize them, and the reader who is primarily looking to find out whether a particular
work has been given a proper accolade, may be disappointed. The intent here is to use the briefest,
simplest and most transparent of arguments to explain to a broad audience why the impact of quantum
effects on hot carrier generation have been so far found to be quite limited, and then to lead that audience
towards the discovery that quantum effects do play a very important role, but not exactly where they
habitually have been looked for.
With that in mind, we consider first the decay of SPPs in a metal nanoparticle that produces hot
carriers. While it is expected that as the size of nanoparticle decreases, the quantum effects must come
into play sooner or later we show that the quantum effects become significant only when the size of a
nanoparticle d approaches the value of d min vF / f where vF is the Fermi velocity and f is optical
frequency. For visible light it amounts to only 2‐3 nanometers. Up to that limit light absorption and SPP
decay can be adequately described by what essentially is a Drude formula with a simple continuous 1/f2
frequency and 1/d size dependencies of damping constant. At the same time, when it comes to statistics
of SPPs and carriers engendered by the latter’s decay, unexpectedly, quantum mechanics must be always
invoked since even nanoparticles as large as 200nm under realistic illumination never have more than a
single SPP residing on each of them (and most often no SPP at all). That leads to interesting (and in a sense
surprising) observation that instant increase of the electron temperature in nanoparticles depends only
on their volumes and not on the intensity of the illumination (within wide limits). Moreover, under
2
continuous illumination the instant (“quantized”) increase of electron temperature in smaller
nanoparticles can be orders of magnitude higher than average electron temperature rise evaluated
classically, with important consequences for various hot carrier applications. Below, the basis for these
conclusions are given in greater detail.
When does metal loss become quantized?
Absorption of light in metals at long wavelengths, for which the photon energy is below the interband
transition threshold (e.g. 350nm for Ag and 500 nm for Au) is due to intraband transitions of free carriers
that are forbidden by the momentum conservation rules. The reason metals absorb at all is that
momentum conservation is maintained with the assistance of electron scattering on phonons, defects and
impurities as well as on other electrons[3]. The combined rate of momentum damping in the bulk metal
is bulk (typically on the scale of 1014s‐1) and phenomenologically it can be introduced into Drude formula
on the metal surface or on a nanoparticle, the fraction of electric field energy in the metal is always
Phonons and imperfections, however, are not the only pathways enabling intraband transitions
as electron reflection from the sharp metal boundary can also provide momentum matching – this is the
basis of the so‐called Kreibig [26, 27] decay channel in nanoparticles that can be phenomenologically
quantified by augmenting the bulk scattering by the additional scattering on the surface at a rate
surf ~ vF / d equal to the frequency of collisions with the surface. A more detailed treatment [28, 29]
shows that momentum‐dependent (or nonlocal) dielectric function of metal ( k ) [30] acquires non‐zero
direct (unassisted) intraband transitions damping become possible. This mechanism is called Landau
damping (LD). In the presence of a sharp interface, the electric field of a plasmon does contains such large
3
wavevector components, which leads to the additional surface‐induced, or LD rate [31] LD 3 vF / d eff
8
, where 𝑑 is the effective “depth” of the field which is essentially the volume‐to‐surface ratio
d eff
metal
E (r ) 2 dV /
metal
E2 (r )dS (1)
surface
and E is the normal to the surface component of the electric field inside the metal. Eq. 1 leads to
LD 3 4 vF / d for both sphere of dimeter d and cube with a side d, which is very close to the result
phenomenologically derived by Kreibig [26, 27].
As one can see, LD becomes the dominant damping mechanism once d vF / bulk ~ Lmfp , Lmfp
being the mean free path of an electron within the metal. At this point, however, one should expect the
energy levels inside the nanoparticle to become quantized, with a discreet set of eigen‐energies and
transitions between them. This is what after all occurs in semiconductor nanoparticles‐quantum dots
(QDs) [32‐34] where quantization has been exploited in numerous practical devices such as lasers and
photodetectors. This raises a question of whether similar quantization effects can be observed in metal
nanoparticles. Furthermore, the LD estimate is based on nonlocality of the dielectric function (which can
be expressed as either ( k ) or (r1 , r2 ) and for the quantized structure the whole concept of dielectric
constant makes no sense and one must account for the polarizability of the entire particle instead. And
this poses a question: what is the impact of quantum confinement on the intraband absorption and
consequently on the SPP damping, and when does the latter starts deviating significantly from the one
predicted by Kreibig?
We now look into the decay of a uniform electric field E inside a cubic nanoparticle and more
specifically how it depends on the size of the nanoparticle d and on the oscillating frequency of the field
x and Fermi velocity projection on that direction is vF , x n / dm0 . Then the matrix element of
Hamiltonian for x‐polarized light is
1 mn
H mn eE mpq x npq dxdydz 4eEd 2 2 , (2)
2 (m n 2 )2
4
as long as m and n have different parities. Furthermore, since the transition energy is
for the energy density inside the metal U 0 E 2 d 3 / 2 we obtain:
4
e 2 vF , x
2
H mn 2 U (3)
0 d d 4 4
Since the separation between two nearest odd or nearest even states is
E 2 2 [ n 2 ( n 2) 2 ] / 2m0 d 2 2 vF , x / d , one can introduce a 1D density of states per unit
energy in the entire cube as x 1/ E . Next we can apply Fermi Golden rule to evaluate the rate of
energy loss by the field inside the cube as
3
dU 2 e 2 vF , x
Rabs 2
H mn x ( EF )d 2
3
( EF )U , (4)
dt mpq 0d 2
over the indices m, p, q i.e. over the angle so that | vF3 , x | vF3 cos3 vF3 / 4 . For spherical
conduction band one has the following relation between the plasma frequency and properties at Fermi
Interestingly, while the confinement damping follows essentially the same dependence as LD, its
value for the cube is twice as high. For this there is a simple explanation. Using LD theory one assumes
that the reflections on two opposite sides of the cube that enable intraband transitions occur independent
of each other so the squares of the transition amplitudes add up. But if the electron experiences no
collisions between the two walls, then the amplitudes of transition should add up coherently, rather than
the incoherently, hence the factor of two is due to quantum interference. Of course, since the scattering
is a statistical process, the damping rate changes gradually as the size decreases,
conf 3 4 [1 exp(d / Lmfp )]vF / d . But what is most relevant to our discussion is that the expression
for damping does not contain Planck’s constant, hence it can be perfectly well described classically.
To verify these simple estimations, we perform a numerical modeling of the decay process where
we plot the decay rate of energy inside the metal Rabs ( ) as a function of frequency for different cube
diameters in Fig1.a. The rate is calculated by adding up oscillator strengths of all the transitions and then
5
broadening them by ~ 60meV to account for the bulk scattering. As one can see, the decay rate
simply describes the loss of energy in the metal and not the absorption of incident radiation by the
nanoparticles – hence the absence of resonances. Shown in Fig. 1b is the plasmon damping rate due to
would expect from the most simple Kreibig’s [26, 27]theory as well as from more detailed LD decay theory
based on Lindhard’s formula[29]. Deviations occur only for smaller cube sizes – one can attribute the
decreased absorption at low energies for d=2nm to quantization of energy levels in the vicinity of Fermi
level and appearance of energy gap. Finally, in Fig.1c we plot the plasmon decay rate averaged over the
dependence with numerical value close to 3 2 vF / d .
1017 1015 1015
2 nm (b) (c)
Plasmon decay rate(s -1)
(a)
Plasmon decay rate (s -1 )
4 nm 2 nm
Field decay rate (s )
-1
4 nm
8 nm 8 nm
1016
16 nm
16 nm 1014 1014
32 nm
1015 32 nm
64 nm
64 nm
decay rate due to confinement averaged over photon energies from 0.5 to 2eV vs cube size. The dependence is roughly 1/d in
agreement with the phenomenological Kreibig’s theory.
According to this simple exercise, the quantization becomes important when the smallest
averaging over all directions) exceeds photon energy which immediately leads do the condition
d (2 / 3 )vF / . Phenomenologically, this condition simply means that the electron reaches the
6
surface within less than a single optical period, so the “free electron” description is no longer valid. It can
be just as easily interpreted from Lindhardt’s theory, if one notices that it corresponds to the situation
when the breadth of spatial spectrum of the electric field k ~ d 1 exceeds the offset of the LD in
Lindhard’s formula, / vF which makes the approximations made in [28, 29] invalid. For the ~1‐1.5eV
energy range this corresponds to about 2‐3 nm and it is easy to see from Fig.1 that it is there that the
curves start to deviate from simple estimate and show an oscillatory pattern. Therefore, when it comes
to the effect of size reduction on the absorption by nanostructures, for as long as dimensions exceed a
couple of nanometers, the “quantum effects” are quite adequately taken into account by the broadening
of the SPP resonance also accompanied by relatively small frequency shift. At smaller dimensions a more
discrete spectrum emerges, but given how broad the resonances are, that plays very little role, especially
since in many practical applications such as photovoltaics and photo catalysis the incoming light is
broadband. Calculations involving full quantum mechanical model [15, 19, 35‐37]lead to essentially the
same limits of applicability of classical expressions, but we have arrived at this conclusion without
resorting to any complicated quantum mechanical calculations, by rather straightforward and physically
transparent reasoning. Clearly, the striking distinctions between metal nanoparticles and QD’s can be also
explained simply by the fact that while a few nm in size QD is typically occupied by only a few electrons,
in metal nanoparticles of similar sizes the number of electrons is measured by hundreds and thousands,
so the collective modes of electrons in metal bear very little resemblance to the single electron transitions
in quantized semiconductor structures as first noted in [38].
When do the SPPs and hot carriers start following quantum rules?
Having not discovered any significant quantum effects when it comes to description of the larger electrons
whose collective motion comprises SPP it now makes sense to discuss whether there is any quantization
when it comes to the hot carriers engendered by the decay of SPP. To do so, let us find out the number
of SPPs that reside at a given nanoparticle at a given time. One way to find out this number is to introduce
dependent parameter of the order of 1 (for spherical nanoparticle A=3) and is the total SPP damping
field in the SPP (or any oscillator for that matter) gets enhanced by Q ~ / . In noble metals Q rarely
7
exceeds 10 in Au and 15 ‐20 in Ag, especially for smaller particles subject to LD. Therefore for visible light
10
we can estimate the irradiance required to sustain a single SPP on a nanoparticle is roughly I1 ~ 10 / V
where the volume is in nm3 and irradiance is in W/cm2. Then for a 50 nm nanoparticle a 100kW/cm2 input
power density is required to sustain a single SPP in the mode and for a 10nm nanoparticle the value is as
high as 10MW/cm2.
Now, these power densities are routine in the femtosecond and picosecond lasers used to study
dynamics of hot carriers in numerous works [39‐42]. Then multiple SPPs can be excited on each
nanoparticle which in their turn engender large number of hot carriers that then decay towards
equilibrium. However, these enormous power densities have very little to do with the real world scenarios
in which hot carrier based technologies are expected to play a role. If we consider applications such as
photo catalysis[43‐45] or photovoltaics[46], then the light source is the sun with irradiance of only about
0.1W/cm2 so even if one uses a powerful concentrator with 1,000 concentration ratio, the average
number of SPPs on a 10‐50 nm nanoparticle NSPP is anywhere between 10‐5 and 10‐3. Likewise, if one
considers photodetectors [6, 47], the typical power density arrives at the detector plane is typically way
below a kW/cm2 . Thus, even if an optical antenna[48] is employed to further concentrate the energy, no
more than a single SPP ever populates a given nanoparticle at a given time. To understand what it all
means, consider Fig.2a in which a rendering of nanoparticle array “frozen” at some time t0 is shown with
1
only one out of N SPP 1 nanoparticles having an SPP excited on it.
This perhaps unforeseen outcome is simply the consequence of a very short SPP lifetime
spp 1 ~ 10 fs . Even putting aside the limited incident power considerations, it is simply impossible to
sustain even a single SPP on a nanoparticle for a long time before it melts, the argument first made in
reference to “spaser” in [49]. If N SPP 1 is to be maintained, the power dissipated per nanoparticle would
have to be in the order of P1 ~ 20W amounting up to 1011W / cm3 power density in 50nm
nanoparticle. That would cause a drastic temperature rise of about 1000K in ~10ns, no matter how
efficient is the cooling, with dire consequences of immediate meltdown of the nanoparticle. So, clearly
the process of SPP generation and decay is discrete, or quantum (as evidenced from the presence of
Planck’s constant in the expression for NSPP) and this fact has important repercussions.
First of all, one cannot use the term “dephasing” [50] when looking at SPPs at practical excitation
power densities – a single SPP cannot de‐phase simply because it has no phase reference, i.e. there is no
second SPP on the same nanoparticle. Instead, it can only be annihilated. Second, SPP cannot gradually
lose its energy to the movement of electron‐hole pairs near the Fermi surface as claimed in a number of
8
works where the “Drude” [51, 52]or “friction” [19, 35] like mechanisms are somehow introduced. That
would violate the energy conservation or it would require the SPP to change frequency! The SPP decay is
quantum and once SPP is annihilated a single electron hole pair (or two of them if electron‐electron
scattering causes SPP decay) is created with entire (or nearly entire if a phonon has been involved) energy
is now contained in these hot carriers. It means that the discrete or quantum features have been also
transferred to the hot carriers and at a given time only a relatively few hot carriers are present.
A A A
Once the “primary” electron and hole have been generated, they thermalize due to rapid
electron‐electron collisions that occur on the rate of tens of fs [42, 53‐55]. Since in each scattering act the
energy of hot carrier is divided between three new, “secondary” carriers, it only takes a few collisions to
thermalize the electrons to the electron temperature exceeding the lattice temperature by Te [56]. This
elevated temperature will then persist for some time EL that is on a scale of hundreds of femtoseconds
because that is how long it takes for the electron‐phonon scattering to restore equilibrium with the lattice.
Note that EL is so long not because the electron‐phonon scattering rate is weak (it is not! [57]) but
because the energy transferred to lattice each time phonon scattering takes place is only a phonon
energy. Therefore the fraction of nanoparticles that have elevated electron temperature at a given time
is N hot N SPP EL , i.e. 10‐100 times N SPP . For the realistic irradiances that are less than a 10kW/cm2
N hot 1 and from this fact a rather important consequence for practical applications follows.
original nanoparticle “A” that had SPP at time t0 remains “hot”, i.e. has elevated electron temperature Te
9
t2 t0 EL the nanoparticle “A” has cooled down to the lattice temperature T, as shown in Fig. 2c, yet
there is a chance that an SPP got excited on another nanoparticle “B”, and the sequence of events repeats
itself.
The value of Te Te T can be easily found as Te / celV where the electron specific heat is
cel 2 k B2TN e / 2 EF [58], N e ~ 60 nm 3 is the electron density and EF is the Fermi energy ( EF
5.5eV for both Au and Ag) . The rise in temperature does not depend on the incident power and is
determined solely by photon energy and the volume – if this is not a quantum effect, we are indeed at a
loss to say what is? We then obtain Te ~ 1.6 104 / V ( K ) where the volume is in nanometers cube. In
large nanoparticles (e.g. d>25nm), the temperature rises by less than 2K. Yet, for smaller nanoparticles,
say 5nm length the temperature rise can be by as high as 200K, and for 3nm nanoparticle it exceeds 1000K.
In the discretized (or quantum) picture the electron temperature is nearly instantly elevated by
Te for the time EL after each SPP generating event while these events occur with a frequency NSPP i.e.
N hot is a duty cycle. Therefore, the average rise of electron temperature, the one which could be calculated
result that depends on neither photon energy nor the particle size. The average, or classical rise of
electron temperature is much less than an instant one [43, 59], but it takes place continuously and not in
short bursts. This can be seen from Fig.3 where both discrete or instant rise in electron temperature Te
the nanoparticle size
3
10
102
Electron Temperature Rise(K)
1
10
Instant
0
10
Average Iin =10,000W/cm2
-1
10
Average Iin =1,000W/cm2
10-2
Average Iin =100W/cm2
-3
10
5 10 15 20 25 30 35 40 45
Now, let us see what difference does “quantization” make? We posit that the desired hot carrier
induced reaction can be only undertaken by a carrier with an energy higher than a certain barrier energy
taken relative to the Fermi energy. Then, when the carriers are at equilibrium with the lattice the rate
10
at which the desired action occurs is R0 exp( / k BT ) and when the average temperature is elevated
to Te T Te with a duty cycle N hot the time‐averaged or “classical” reaction rate becomes
Te Te
Rav exp( / k BTe ) R0 R0 T
and the enhancement K av Rav / R0 R0 T
When the granularity is
cycle Nhot , so that the “discrete” or “quantum” enhancement is
In Fig.4a the dependence of the reaction rate enhancement K dis 1 on heating duty cycle N hot for
different values of K av is shown. The horizontal line corresponds to a 100% enhancement. For relatively
estimate of the reaction rate. But for small duty cycle (i.e. for small nanoparticles) the enhancement
increases rapidly and for really small duty cycles the difference made by quantization becomes enormous.
To put it into a different reference frame we plot the enhancement vs nanoparticle size for different
barrier heights in Fig.4 b and c for two different incident irradiances. As one can see, for relatively high
barriers and small nanoparticles the difference between the quantized and classical results is indeed
enormous.
105 105
(a) 105 Φ=1.2eV
104 Φ=1.2eV 104 (c)
104
(b)
Φ=1.0eV Iin=1000W/cm2
103 Φ=1.0eV Iin=100W/cm2 103
103
Enhancement Kdis‐1
Enhancement Kdis‐1
Enhancement Kdis‐1
The discrete character of hot carrier generation is manifested in significant enhancement of desired hot
carrier action in comparison to classical one. Therefore, while the classical theory with pitifully small rise
in electron temperature cannot possibly explain how the rate of thermionic processes can be enhanced
by so much [43, 59], the quantum theory offers a very plausible explanation. Of course, in case when the
barrier is very high and R0 is small, neither classical nor quantum theory offer any explanation. It can be
the action of ballistic (or the first generation) hot carriers that have not yet experienced a single collision
11
[56] , which is precisely what happens in Schottky photodetectors[6, 47]. Or it can be due to rise in the
lattice temperature, i.e. not be a hot carrier effect in the first place[59].
Conclusions
The lesson that may be drawn out of this unassuming exercise is rather short and plain, but hopefully not
inconsequential. The sheer fact that one can employ full quantum narrative to precisely describe a given
process does not necessarily mean that doing so will reveal anything that could not be adequately
described using rudimentary classical approach. This platitude can be applied to any task, and this study
of hot carrier processes in plasmonic nanoparticles is no different. When it comes to depiction of metal
nanoparticles, the metal is perfectly well described classically using a Drude model with damping constant
modified by surface collisions for all but the smallest (less than 2‐3nm) nanoparticles. Even for smaller
nanoparticles quantum correction amounts to very little if the incoming radiation is broadband. At the
same time, granular, or quantum character of SPP excitation and decay cannot be neglected. The “spiking”
or shot‐noise‐like nature of the rise and fall of electron temperature may lead to orders of magnitude
difference in the rate of certain hot‐carrier processes, such as for instance photo catalysis. With any lack,
this plea for judicious use of quantum theory will be heard by the community and be of some use to it.
Acknowledgement
Authors acknowledge support of Air Force Office of Scientific (AFOSR) Research Award # FA9550‐16‐
10362 ‐and the Israeli−USA Binational Science Foundation (BSF). JBK also appreciates indispensable
assistance and critique offered by Prof. P. Noir and sharp comments by the newest post‐doctoral member
of his group Ms. S. Artois.
1. Stockman, M.I., Nanoplasmonics: past, present, and glimpse into future. Optics Express, 2011.
19(22): p. 22029‐22106.
2. Maier, S.A., Plasmonics : fundamentals and applications. 2007, New York: Springer. xxiv, 223 p.
3. Khurgin, J.B., How to deal with the loss in plasmonics and metamaterials. Nature
Nanotechnology, 2015. 10(1): p. 2‐6.
4. Moskovits, M., The case for plasmon‐derived hot carrier devices. Nature Nanotechnology, 2015.
10(1).
12
5. Brongersma, M.L., N.J. Halas, and P. Nordlander, Plasmon‐induced hot carrier science and
technology. Nature Nanotechnology, 2015. 10(1): p. 25‐34.
6. Goykhman, I., et al., Waveguide based compact silicon Schottky photodetector with enhanced
responsivity in the telecom spectral band. Optics Express, 2012. 20(27): p. 28594‐28602.
7. Smith, J.G., J.A. Faucheaux, and P.K. Jain, Plasmon resonances for solar energy harvesting: A
mechanistic outlook. Nano Today, 2015. 10(1): p. 67‐80.
8. Christopher, P., H.L. Xin, and S. Linic, Visible‐light‐enhanced catalytic oxidation reactions on
plasmonic silver nanostructures. Nature Chemistry, 2011. 3(6): p. 467‐472.
9. Linic, S., P. Christopher, and D.B. Ingram, Plasmonic‐metal nanostructures for efficient
conversion of solar to chemical energy. Nature Materials, 2011. 10(12): p. 911‐921.
10. Quantum plasmonics. 2016, New York, NY: Springer Berlin Heidelberg. pages cm.
11. Tame, M.S., et al., Quantum plasmonics. Nature Physics, 2013. 9(6): p. 329‐340.
12. Bozhevolnyi, S.I. and J.B. Khurgin, The case for quantum plasmonics. Nature Photonics, 2017.
11(7): p. 398‐400.
13. Bogdanov, S.I., A. Boltasseva, and V.M. Shalaev, Overcoming quantum decoherence with
plasmonics. Science, 2019. 364(6440): p. 532‐533.
14. Román Castellanos, L., O. Hess, and J. Lischner, Single plasmon hot carrier generation in metallic
nanoparticles. Communications Physics, 2019. 2(1): p. 47.
15. Esteban, R., et al., Bridging quantum and classical plasmonics with a quantum‐corrected model.
Nature Communications, 2012. 3.
16. Manjavacas, A., et al., Plasmon‐Induced Hot Carriers in Metallic Nanoparticles. Acs Nano, 2014.
8(8): p. 7630‐7638.
17. Hartland, G.V., et al., What's so Hot about Electrons in Metal Nanoparticles? Acs Energy Letters,
2017. 2(7): p. 1641‐1653.
18. Dal Forno, S., L. Ranno, and J. Lischner, Material, Size, and Environment Dependence of Plasmon‐
Induced Hot Carriers in Metallic Nanoparticles. Journal of Physical Chemistry C, 2018. 122(15): p.
8517‐8527.
19. Govorov, A.O., H. Zhang, and Y.K. Gun'ko, Theory of Photoinjection of Hot Plasmonic Carriers
from Metal Nanostructures into Semiconductors and Surface Molecules. Journal of Physical
Chemistry C, 2013. 117(32): p. 16616‐16631.
20. Govorov, A.O. and H. Zhang, Kinetic Density Functional Theory for Plasmonic Nanostructures:
Breaking of the Plasmon Peak in the Quantum Regime and Generation of Hot Electrons. Journal
of Physical Chemistry C, 2015. 119(11): p. 6181‐6194.
21. Douglas‐Gallardo, O.A., et al., Plasmon‐induced hot‐carrier generation differences in gold and
silver nanoclusters. Nanoscale, 2019. 11(17): p. 8604‐8615.
22. Jain, P.K., et al., Calculated absorption and scattering properties of gold nanoparticles of
different size, shape, and composition: Applications in biological imaging and biomedicine.
Journal of Physical Chemistry B, 2006. 110(14): p. 7238‐7248.
23. Berciaud, S., et al., Observation of Intrinsic Size Effects in the Optical Response of Individual Gold
Nanoparticles. Nano Letters, 2005. 5(3): p. 515‐518.
24. van Dijk, M.A., et al., Absorption and scattering microscopy of single metal nanoparticles.
Physical Chemistry Chemical Physics, 2006. 8(30): p. 3486‐3495.
25. Averitt, R.D., D. Sarkar, and N.J. Halas, Plasmon resonance shifts of Au‐coated Au2S nanoshells:
Insight into multicomponent nanoparticle growth. Physical Review Letters, 1997. 78(22): p.
4217‐4220.
26. Kreibig, U. and L. Genzel, Optical‐Absorption of Small Metallic Particles. Surface Science, 1985.
156(Jun): p. 678‐700.
13
27. Kreibig, U. and M. Vollmer, Optical properties of metal clusters. Springer series in materials
science. 1995, Berlin ; New York: Springer. xx, 532 p.
28. Khurgin, J.B., Ultimate limit of field confinement by surface plasmon polaritons. Faraday
Discussions, 2015. 178: p. 109‐122.
29. Khurgin, J., et al., Landau Damping and Limit to Field Confinement and Enhancement in
Plasmonic Dimers. Acs Photonics, 2017. 4(11): p. 2871‐2880.
30. Lindhard, J., On the Properties of a Gas of Charged Particles. Matematisk‐Fysiske Meddelelser
Kongelige Danske Videnskabernes Selskab, 1954. 28(8): p. 1‐57.
31. Khurgin, J.B., Hot carriers generated by plasmons: where are they generated and where do they
go from there? Faraday Discussions, 2019. 214: p. 35‐58.
32. Bandyopadhyay, S. and H.S. Nalwa, Quantum dots and nanowires. 2003, Stevenson Ranch, Calif.:
American Scientific Publishers. xix, 423 p.
33. Tartakovskiĭ, A.G., Quantum dots : optics, electron transport, and future applications. 2012,
Cambridge ; New York: Cambridge University Press. xviii, 358 p.
34. Wang, Z.M., Quantum dot devices. Lecture notes in nanoscale science and technology. 2012,
New York: Springer. xiii, 370 pages.
35. Besteiro, L.V., et al., Understanding Hot‐Electron Generation and Plasmon Relaxation in Metal
Nanocrystals: Quantum and Classical Mechanisms. Acs Photonics, 2017. 4(11): p. 2759‐2781.
36. Zuloaga, J., E. Prodan, and P. Nordlander, Quantum Description of the Plasmon Resonances of a
Nanoparticle Dimer. Nano Letters, 2009. 9(2): p. 887‐891.
37. Link, S. and M.A. El‐Sayed, Shape and size dependence of radiative, non‐radiative and
photothermal properties of gold nanocrystals. International Reviews in Physical Chemistry, 2000.
19(3): p. 409‐453.
38. Silberberg, Y. and T. Sands, Optical‐Properties of Metallic Quantum‐Wells. Ieee Journal of
Quantum Electronics, 1992. 28(7): p. 1663‐1669.
39. Mueller, B.Y. and B. Rethfeld, Relaxation dynamics in laser‐excited metals under nonequilibrium
conditions. Physical Review B, 2013. 87(3).
40. Rethfeld, B., et al., Ultrafast dynamics of nonequilibrium electrons in metals under femtosecond
laser irradiation. Physical Review B, 2002. 65(21).
41. Schmuttenmaer, C.A., et al., Femtosecond Studies of Carrier Relaxation Processes at Single‐
Crystal Metal‐Surfaces. Laser Techniques for Surface Science, 1994. 2125: p. 98‐106.
42. Petek, H. and S. Ogawa, Femtosecond time‐resolved two‐photon photoemission studies of
electron dynamics in metals. Progress in Surface Science, 1997. 56(4): p. 239‐310.
43. Sivan, Y., I.W. Un, and Y. Dubi, Assistance of metal nanoparticles in photocatalysis ‐ nothing
more than a classical heat source. Faraday Discussions, 2019. 214: p. 215‐233.
44. Zhang, C., et al., Al‐Pd Nanodisk Heterodimers as Antenna‐Reactor Photocatalysts. Nano Letters,
2016. 16(10): p. 6677‐6682.
45. Mukherjee, S., et al., Hot Electrons Do the Impossible: Plasmon‐Induced Dissociation of H‐2 on
Au. Nano Letters, 2013. 13(1): p. 240‐247.
46. Atwater, H.A. and A. Polman, Plasmonics for improved photovoltaic devices (vol 9, pg 205, 2010).
Nature Materials, 2010. 9(10): p. 865‐865.
47. Goykhman, I., et al., Locally Oxidized Silicon Surface‐Plasmon Schottky Detector for Telecom
Regime. Nano Letters, 2011. 11(6): p. 2219‐2224.
48. Knight, M.W., et al., Photodetection with Active Optical Antennas. Science, 2011. 332(6030): p.
702‐704.
49. Khurgin, J.B. and G. Sun, Injection pumped single mode surface plasmon generators: threshold,
linewidth, and coherence. Optics Express, 2012. 20(14): p. 15309‐15325.
14
50. Bosbach, J., et al., Ultrafast dephasing of surface plasmon excitation in silver nanoparticles:
Influence of particle size, shape, and chemical surrounding. Physical Review Letters, 2002.
89(25).
51. Narang, P., R. Sundararaman, and H.A. Atwater, Plasmonic hot carrier dynamics in solid‐state
and chemical systems for energy conversion. Nanophotonics, 2016. 5(1): p. 96‐111.
52. Brown, A.M., et al., Nonradiative Plasmon Decay and Hot Carrier Dynamics: Effects of Phonons,
Surfaces, and Geometry. Acs Nano, 2016. 10(1): p. 957‐966.
53. Fann, W.S., et al., Electron Thermalization in Gold. Physical Review B, 1992. 46(20): p. 13592‐
13595.
54. Fann, W.S., et al., Direct Measurement of Nonequilibrium Electron‐Energy Distributions in
Subpicosecond Laser‐Heated Gold‐Films. Physical Review Letters, 1992. 68(18): p. 2834‐2837.
55. Bauer, M., A. Marienfeld, and M. Aeschlimann, Hot electron lifetimes in metals probed by time‐
resolved two‐photon photoemission. Progress in Surface Science, 2015. 90(3): p. 319‐376.
56. Khurgin Jacob, B., Fundamental limits of hot carrier injection from metal in nanoplasmonics, in
Nanophotonics. 2019.
57. Ono, S., Thermalization in simple metals: Role of electron‐phonon and phonon‐phonon
scattering. Physical Review B, 2018. 97(5).
58. Ziman, J.M., Electrons and phonons : the theory of transport phenomena in solids. Oxford classic
texts in the physical sciences. 2001, Oxford, New York: Clarendon Press, Oxford University Press
59. Dubi, Y. and Y. Sivan, "Hot" electrons in metallic nanostructures‐non‐thermal carriers or heating?
Light‐Science & Applications, 2019. 8.
15
Figure Captions
of frequency for different cube sizes. The energy loss rate is inversely proportional to the square of
frequency. (b) SPP polariton decay rate due to confinement conf vs frequency for different cube sizes.
This rate is nearly frequency (wavelength) independent. (c) SPP polariton decay rate due to confinement
averaged over photon energies from 0.5 to 2eV vs cube size. The dependence is roughly 1/d in agreement
with the phenomenological Kreibig’s theory.
Figure 2 Evolution of SSPs and hot carriers in the ensemble of nanoparticles. (a) An SPP is generated on
one out of great many nanoparticles (b) SPP has decayed and the electron temperature Te of the
nanoparticle A has risen above lattice temperature T (c) The electron temperature of nanoparticle A has
reached equilibrium with the lattice.
Figure 3. Instant (solid line) and Average (dashed lines) electron temperature rise vs the size of the
nanoparticle
Figure 4 (a) Enhancement of the hot electron induced reaction as a function of heating duty cycle (the
fraction of time the nanoparticle contains hot carriers) (b,c) Enhancement of the hot carrier induced
reaction vs. the size of nanoparticle for two different incident irradiances
16