Full Text

Turn on search term navigation

ARTICLE

Received 6 Jan 2015 | Accepted 26 Mar 2015 | Published 2 Jun 2015

Hot carriers (HC) generated by surface plasmon polaritons (SPPs) in noble metals are promising for application in optoelectronics, plasmonics and renewable energy. However, existing models fail to explain key quantitative details of SPP-to-HC conversion experiments. Here we develop a quantum mechanical framework and apply rst-principles calculations to study the energy distribution and scattering processes of HCs generated by SPPs in Au and Ag. We nd that the relative positions of the s and d bands of noble metals regulate the energy distribution and mean free path of the HCs, and that the electronphonon interaction controls HC energy loss and transport. Our results prescribe optimal conditions for HC generation and extraction, and invalidate previously employed free-electron-like models. Our work combines density functional theory, GW and electronphonon calculations to provide microscopic insight into HC generation and ultrafast dynamics in noble metals.

DOI: 10.1038/ncomms8044 OPEN

Theory and computation of hot carriers generated by surface plasmon polaritons in noble metals

Marco Bernardi1,2,*, Jamal Mustafa1,2,*, Jeffrey B. Neaton1,2,3,4 & Steven G. Louie1,2

1 Department of Physics, University of California at Berkeley, 366 LeConte Hall #7300, Berkeley, California 94720, USA. 2 Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA. 3 Molecular Foundry, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA. 4 Kavli Institute for Energy Nanosciences at Berkeley, Berkeley, California 94720, USA. * These authors contributed equally to this work. Correspondence and requests for materials should be addressed to S.G.L. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 6:7044 | DOI: 10.1038/ncomms8044 | http://www.nature.com/naturecommunications

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8044

Surface plasmon polaritons (SPPs) are electron collective excitations generated by light at the interface between a metal and a dielectric1. Modelling of SPPs is dominated by

approaches that employ classical electromagnetism and account empirically for the properties of the materials supporting the SPP. Yet, the inherently quantum mechanical nature of SPPs2,3 becomes manifest in their scattering and decay processes in bulk and nanoscale materials. A key example is the decay of SPPs to electron-hole pairs, a process whose crucial importance is twofold: rst, it is a main energy loss mechanism of SPPs in metals, currently limiting the applicability of plasmonics4,5; second, it leads to generation of hot electrons and holes with application in several branches of applied physics6. Recent experiments reported the extraction of such SPP-induced hot carriers (HCs) in metals before they thermalizetypically over nanometer lengths and sub-picosecond timesby injecting the HCs over a Schottky barrier into a semiconductor or oxide79, or by transferring them into surface adsorbates to perform chemical reactions with large activation barriers6,10,11. The vast literature on HCs in noble metals generated by intense light pulses1214 (instead of SPPs), together with recent calculations on SPP-to-HC conversion15,16, help interpret these recent experiments on SPP-induced HCs.

Lack of predictive and accurate quantum mechanical approaches to study HCs generated by SPPs or photons has led to ambiguity in the microscopic interpretation of experiments involving HCs in Au and Ag, two mainstay materials in plasmonics4. In particular, recent studies indicate that SPP decay excites electronic transitions from occupied states close to the Fermi energy (EF)10, implying that most of the SPP energy goes into hot electrons rather than hot holes, and that nanocrystals with o10 nm diameter are necessary to obtain

HCs with signicant energy17 (for example, 12 eV away from EF). The mean free paths (MFPs) of HCs in noble metals appear to be short10, though a quantitative estimate of the MFPs and their dependence on crystal direction is not available. In addition, it is often assumed that HC scattering and thermalization are dominated by Auger processes10,14, since thermalization induced by phonons occurs on a slower time scale14. As shown in this work, all these conclusions must be revised.

Here, we develop a quantum mechanical framework to study the energy distribution of HCs generated by SPPs and photons in Au and Ag, and employ ab initio calculations of the electronphonon (eph) and electron-electron (ee) interactions to study the MFPs and relaxation times of HCs within 5 eV of the Fermi energy. Our approach is free of experimental input, and combines density functional theory (DFT)18, the GW (where G is the Green function, W is the screened Coulomb potential, and GW is the diagram employed for the electron self energy) method19, and ab initio eph calculations20. Our ability to use extremely ne grids for Brillouin zone (BZ) integrations allows us to resolve HC scattering with unprecedented accuracy21. We nd that the

interband transition threshold (between d and s states) denes two regimes for HC generation and transport. The decay of SPPs with energy lower than the interband threshold leads to generation of long-lived HCs with long isotropic MFPs of up to 40 nm and energy within 12 eV of EF. On the other hand, decay of SPPs with energy higher than the interband threshold leads to generation of short-lived hot holes in d states with anisotropic and short (sub 5 nm) MFPs, and hot electrons with only o1 eV energy above EF.

The regime characterized by SPP energy below the interband threshold is better suited to employ HCs in applications requiring long MFPs, and allows one to optimize HC generation by tuning the SPP energy. These results represent an important rst step to understand SPP decay and energy loss, and to control HC generation and transport in bulk and nanostructured noble metals.

ResultsTheory of SPPelectron coupling. Coupling of SPPs or photons to electrons in materials can be described in the framework of many-body perturbation theory22. The lowest order Feynman diagram for this coupling process, shown in Table 1, describes a boson (here, SPP or photon) coupling to the electron gas through the electronic polarizability w and a coupling matrix element g.

This diagram is analogous to the lowest order coupling of phonons to electrons23,24. The decay rate G(qp,Ep) for a boson of momentum :qp and energy EP to electronhole pairs is proportional to the imaginary part of the self-energy22:

Gqp; Ep

1p Im

" #

fn;k fn

;k q

Xn;n ;k

jgn;n ;kqp; Epj2

Epqp En

;k q En;k iZ

where En,k and En

0;k qp are, respectively, the quasiparticle

energies of the hot hole in a state with band n and crystal momentum k, and hot electron in a state with band n0 and crystal momentum k qp, as generated in the boson decay

process. In addition, fn,k and fn0;k qp are Fermi occupations, Z is a

small broadening, and the sum is contributed by electronic transitions from occupied to empty states that differ by :qp in crystal momentum and Ep in energy. If the matrix elements have a weak band and k dependence, the decay rate of a SPP or photon is proportional to the imaginary part of the electronic polarizability, Imw(qp,EP).

Here, we employ a quantity proportional to Imw(qp,EP), the

nite momentum joint density of states (FM-JDOS) J(qp,EP),

dened as the number of states per unit energy separated by momentum :qp and energy Ep in the quasiparticle bandstructure:

Jqp; Ep

1p Imwqp; Ep

1p Im

Xn;n0;kfn;k fn0;k qp Epqp En0;k qp En;k iZ

" #

Table 1 | Bosonelectron coupling and matrix elements.

Interaction Matrix element Diagram Plasmonelectron gPL

2pe2Ep=q2p

Photonelectron gPHn;n ;k

2 e h=m

2 1= 2EEpV

eq pn;n ;k qp 0

SPPelectron gSPPn;n;k

2 e

h=m

2 1=2EeffEpALz

eTMq pn;n ;k qp

SPP, surface plasmon polariton.

Square modulus of the coupling matrix element, g

j j2, and Feynman diagram for the self-energy of the boson in the lowest-order interaction for the three bosons considered hereplasmons, photons and

SPPswith electrons in a material. These interactions share a common diagram (right column) consisting of a polarizability bubble w multiplied by two coupling vertices g. The wiggly lines on the two sides of the self-energy loop g2w indicate the incoming and outgoing bosons.

2 NATURE COMMUNICATIONS | 6:7044 | DOI: 10.1038/ncomms8044 | http://www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8044 ARTICLE

The FM-JDOS is a measure of the phase space available for the decay of a SPP or photon with momentum :qp and energy Ep.

Greater FM-JDOS values generally would correspond to greater SPP decay and HC generation rates.

The coupling matrix elements in equation (1) can be derived with a range of theoretical approaches. We rst discuss the coupling of electrons to plasmonsthat is, collective electronic excitations in the absence of lightsince theories developed for plasmons have been occasionally applied to SPPs25. The so-called Landau damping26 treatment of the electron gas derives the decay rate Gp of a bulk plasmon in the weak coupling limit via the zeros of the complex macroscopic dielectric function eM of the metal close to the real energy axis, that is, EMqp; Ep i Gpqp; Ep 0. Due to the bosonic nature of plasmons, the decay rate G(qp,Ep) has the form given in equation (1), with coupling matrix elements26 gPLj j2 2pe2Ep=q2p.

In spite of its common use, Landau damping is rarely discussed in terms of the diagram in Table 1 and the coupling gPL, though this viewpoint elucidates its physical origin.

The coupling matrix elements for SPPs and photons can be derived by second quantization of the vector potential27 (see Methods). For photons, the coupling is given by the well-known dipole matrix element

jgPHn;n0;kj2 e =m2 1=2EEpV je

= EMED

. This

approach would enable rst-principles calculations of plasmonic losses and the MFPs of SPPs in materials4,5.

HC generation. To study HC generation from SPPs at a metaldielectric interface (see Fig. 1a), we compute the FM-JDOS for Au and Ag, as shown in Fig. 1b, for SPP energies up to 5 eV and increasing SPP wavevectors along the GL direction in the

BZ (that is, for SPPs propagating along the [111] crystallographic direction). Our calculations are carried out using GW band-structures (see Methods).

Detailed features of the bandstructures of Au and Ag, shown in Fig. 2, regulate HC generation and transport in these materials. The main features of the Au and Ag bandstructures are the presence of relatively dispersionless, occupied d bands with large associated electronic DOS, straddled by a free-electron-like s band. The top of the d bands is located at energy Eint below

the Fermi energy, where Eint is the threshold for interband (d to s) electronic transitions32,33. From our GW calculations, we nd Eint values of 2.3 eV in Au and 3.7 eV in Ag (see Fig. 2), in excellent agreement with experiment5,32,33. The free-electron-like s band spans an energy window between Eint below and over

qp pn;n0;kqp 0j2, where V

is the volume, eqp is the polarization vector, and pn;n0;kqp the

transition dipole matrix element. For SPPs at a metaldielectric interface, we apply a vector potential second quantization procedure28,29 (see Methods) to obtain the SPPelectron coupling jgSPPn;n0;kj2 given in Table 1. The SPPelectron coupling

has a form analogous to that for photons, with two important differences. First, the volume V is replaced by the SPP-eld volume VSPP A Lz, where A is the metaldielectric interface

area and Lz the decay length of the SPP in the metal. Since Lz is of order 10 nm at visible wavelength for noble metals1 and thus much shorter than the light absorption depth, the electric eld associated with the SPP is concentrated to a small volume VSPP at

the metaldielectric interface1, resulting in enhanced local

generation of HCs from SPPs compared with light. Second, the SPP wavevector qp is in general comparable to the size of the BZ, while for photons qpE0. The nite SPP wavevector can signicantly alter the phase space available for SPP decay, as dictated by the qp dependence of the FM-JDOS. We remark that additional momentum can be transferred by phonons and defects assisting the SPP scattering process.

For both SPP and photon decay, the FM-JDOS regulates the phase space for HC generation. Since the experimental conditions and the material-specic SPP dispersion relation dictate the SPP energy Ep and wavevector qp1,30,31, we study the FM-JDOS as a function of qp and Ep and arbitrarily treat them for the purpose of this work as independent variables; HC generation from light is obtained as the specic limit qp 0. We note that the SPP decay

rate Gp(qp,Ep) could also be computed by evaluating the coupling matrix elements explicitly for xed experimental conditions and dispersion relation1 Epqp c qp

Dielectric

Metal

h[afii9840]

SPP

Electron

Hole

Phonon

0.8

0.6

0.4

0.2

Au Ag

0.05 L0.1 L0.2 L0.3 L0.5 L

FM-JDOS

(transition per eV)

0.8

0.6

0.4

0.2

0.05 L0.1 L0.2 L0.3 L0.5 L

0 0 1 2 Eint Eint

3 4 5

Energy (eV)

0 0 1 2 3 4 5

Energy (eV)

Figure 1 | Hot carrier generation in Au and Ag. (a) Schematic of a metaldielectric interface under illumination, showing the decay of a SPP to a pair of hot electron and hole. The hot carriers are scattered and lose energy through phonon emission processes. The mean free path corresponds to the length of the black arrows, and the excited volume within the metal is shaded in grey. (b) The FM-JDOS of Au (left panel) and Ag (right panel) as a function of SPP energy, shown for several values of qp along the GL direction, up to a value of qp 0.5 L. The SPP wavevector is expressed in units of the L point in the BZ,

where L (p/a)(1,1,1), with a the lattice constant. The FM-JDOS for photons and SPPs with vanishingly small wavevector corresponds to the qp 0 case (curve labelled as G). The dashed lines indicate the value of the FM-JDOS low-energy tail for large wavevectors.

NATURE COMMUNICATIONS | 6:7044 | DOI: 10.1038/ncomms8044 | http://www.nature.com/naturecommunications

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8044

Hot carrier population distribution (a.u.) vs Energy

Hot carrier distributionHot carrier distribution

0.9 eV0.1 L

1.7 eV0.2 L

1.9 eV 4

0.2 L

Energy (eV)

2.5 eV0.5 L

5.0 eV0.5 L

L X K

2.0 eV0.5 L

0 1 2 3 4 5

5 0 1 2 3 4 5

4Energy (eV)

Energy (eV)

0.9 eV0.1 L

Hot carrier distribution Hot carrier distribution

1.9 eV0.2 L

3.1 eV0.3 L

Energy (eV)

3.0 eV0.5 L

4.0 eV0.5 L

5.0 eV0.5 L

0 1 2 3 4 5

X K

Energy (eV)

Wavevector

Figure 2 | Energy distribution of hot carriers. Energy distribution of hot electrons (red) and hot holes (blue) in Au (a) and Ag (b) generated by SPP decay. Each plot refers to a specic SPP energy and wavevector. The right panel shows the GW bandstructures of Au and Ag, respectively, in a and b, along with a schematic of the momentum and energy conserving excitation of a hot electronhole pair resulting from the decay of a SPP with wavevector qp and energy

Ep. The shaded area indicates the energy range of the d states. The energy difference between the Fermi energy and the top of the d bands corresponds to the interband transition threshold. The Fermi energy is the zero of the energy axis.

4 eV above EF, and hence dominates the electronic properties of Au and Ag near EF.

The FM-JDOS in Fig. 1b follows similar trends in Au and Ag, with deviations due to the different Eint values in the two materials. We rst discuss the FM-JDOS with qp 0, which

quanties the phase space for HC generation for both SPPs with very small wavevectors and photons. The FM-JDOS curve with qp 0 (G) has a zero value for SPP energies up to Eint, indicating

the absence of possible SPP decay into HCs in this energy range. For SPP energies greater than Eint, the FM-JDOS increases

linearly with energy, indicating that a large phase space opens up for HC generation. In this regime, SPP damping is strong, and HC generation is intense and dominated by formation of hot d holes collecting most of the energy from the SPP, as discussed below.

For small SPP wavevectors along the GL direction up to B0.2 L, we nd a peak in the FM-JDOS corresponding to SPP decay through intraband transitions within the s band. The energy position of the peak increases linearly with SPP wavevector, and for large enough wavevectors merges with the FM-JDOS above Eint. In Au, the intraband peak is below Eint for qp values up to B0.2 L, and in Ag for qp values up to B0.3 L due to the higher Eint value in Ag. The presence of an intraband peak associated with ss transitions can be understood by examining

the energy difference between states with momentum k and k qp in the free-electron-like s band:

E k qp

E k

22m 2k qp q2p

For small wavevector qp, the term quadratic in qp can be neglected, while the term linear in qp increases by B 2p2

p = 10ma2

1 eV (a is the lattice constant) for an

increase of qp by 1/10 the GL distance. The computed FM-JDOS conrms these trends, as seen by the increase in the energy of the intraband peak by B1 eV in going from 0.1 L to 0.2 L and from0.2 L to 0.3 L. In this small wavevector regime, generation of HCs is appreciable only if the SPP energy and wavevector are matched to the intraband FM-JDOS peak, namely if Ep(qp)E( 2k qp)/m

for a set of k points in the BZ. When this condition is met, intense HC generation from intraband transitions can occur, resulting in an almost equal apportioning of the SPP energy between hot electrons and holes, as discussed below.

For SPP wavevectors larger than B0.3 L, the FM-JDOS resembles that for qp 0 except from a at tail that extends to

low SPP energy, as shown by the dashed lines in Fig. 1. This tail is due to generation of HCs from intraband transitions with large transferred momentum from the decay of SPPs with wavevector

4 NATURE COMMUNICATIONS | 6:7044 | DOI: 10.1038/ncomms8044 | http://www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8044 ARTICLE

comparable to the size of the BZ. The FM-JDOS for qp40.5 L is unchanged and nearly identical to the 0.5 L case.

To summarize, we nd two HC generation regimes. For SPPs with energy EpoEint, optimal generation can be achieved using

SPPs with relatively small wavevector (less than B1/5 the BZ size) that are matched with the FM-JDOS peak, while HC generation is relatively weak for SPPs with larger momentum. For SPPs with energy Ep4Eint, HC generation is intense regardless of the SPP momentum. We nd identical trends for SPPs with wavevector in the GX high-symmetry direction, that is, propagating along the [100] direction. Removal of the electron gas approximation used in simplied treatments of SPPs in noble metals17 is necessary to capture the delicate interplay found here between the energetics of the s and d states and intraband (ss) and interband (ds) transitions.

Energy distribution of HCs. Figure 2 shows the energy distribution of HCs generated by specically assumed SPPs with energy increasing in small steps from B1 to 5 eV and a range of wavevectors in the GL direction, along with the GW band-structures. From now on throughout the manuscript, all HC energies are referenced to the Fermi energy (EF), and the energy of the hot hole is E, that is, the direction of increasing energy in

the bandstructure is downward for holes and upward for electrons. The energy distribution histograms identify the HCs generated in the two regimes discussed above. For SPPs with energy lower than Eint, hot holes and electrons are generated with a roughly uniform energy distribution as a result of the ss intra-band transitions. The HC energy distributions shown in Fig. 2 in this energy range pertain to SPPs with energy and wavevectors matched to the FM-JDOS intraband peaks (see Fig. 1). While the details of the energy distribution in this regime are sensitive to the SPP energy Ep, the overall trends indicate that HCs are generated with an average energy of BEp/2.

For SPPs with energy greater than Eint, we nd a change to a

different HC generation regime. At the onset of the interband transitions for SPP energy of 2.3 eV in Au and 3.7 eV in Ag, on SPP decay the hot holes absorb most of the SPP energy, creating a spike in the hot hole population at the top of the d band. On the other hand, hot electrons have only modest energies in this regime. At higher SPP energies of B5 eV, the HC population become approximately uniformly distributed over a wide energy

range, although a peak in the hot d hole population is still present in Ag due to its higher Eint. These trends are common to both Au and Ag, and we have veried that Cu, a material less commonly used in plasmonics5 and not discussed here, follows identical HC generation and energy distribution trends as Au and Ag.

HC scattering processes. While ample experimental13,32 and theoretical14 data exists on HCs in noble metals, characterization of the MFPs and relaxation times of HCs using ab initio theory has been limited by the absence of accurate eph calculations including all phonon modes over the entire BZ. First-principles calculations combining eph and ee interactions are necessary to characterize the anisotropic and energy-dependent MFPs of HCs in materials, as shown in our recent work on HCs in semiconductors21. We compute the scattering rate (and its inverse, the relaxation time) for the eph and ee interactions from the imaginary part of the respective self-energies, Im e ph and Im e e, and resolve these quantities for each electronic state with band n and k-point on very ne BZ grids (see Methods and ref. 21). The total relaxation times tn,k,

combining the eph and ee interactions, are obtained as

tn;k

2= Im e en;k Im e phn;k

h i

Figure 3 shows the total relaxation times tn,k for HCs in Au and Ag with energy up to 5 eV. The total relaxation times decay rapidly away from EF following a volcano-shaped trend, with the peak of the volcano centred close to EF. In Ag, we nd an additional peak in the relaxation times at the top of the d bands, in agreement with recent photoemission experiments34. Analysis of the eph and ee scattering rates (see the bottom panels in Fig. 3) highlights the differences between the energy windows spanned by the s and d states, and elucidates the origin of the relaxation time trends. In the energy range spanned by the s states, the ee scattering rates (that is, Im e e) form a parabola with minimum value of zero at EF as predicted by Fermi liquid theory14, while the eph scattering rates are relatively constant and exhibit a minimum at the onset of the d states and a maximum 12 eV above EF. The ee scattering rates become greater than the eph B2 eV away from EF, thus indicating that Auger and impact ionization processes included in the ee term dominate HC scattering only 2 eV or more away from EF, while eph dominates HC scattering within 2 eV of EF. Combining the

30 30

Au Ag

elphelel elph

elel

Relaxation time (fs)

Im(meV)

Relaxation time (fs)

200

150

100

Im(meV)

200

150

100

0 4 2 0 2 4

Energy (eV)

Figure 3 | Ultrafast hot carrier scattering. Relaxation time and scattering rates for HCs in Au (a) and Ag (b) within 5 eV of the Fermi energy. The top panels show the total relaxation time, accounting for both the ee and the eph interactions. The bottom panels show the scattering rate, expressed as the imaginary part of the self-energy, Im , for the ee and eph interactions. EF is the zero of the energy axis, as shown by the dashed line. Electrons (holes)

possess positive (negative) energies, and the shaded area indicates the energy window spanned by the d states.

0 4 2 0 2 4

Energy (eV)

NATURE COMMUNICATIONS | 6:7044 | DOI: 10.1038/ncomms8044 | http://www.nature.com/naturecommunications

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8044

two scattering mechanisms leads to total relaxation times with a broad maximum centred at EF and a rapid decay 12 eV away from EF as the parabolic ee scattering rate becomes dominant (see Fig. 3). In the energy window spanned by the d bands, the large increase in the electronic DOS causes very strong eph scattering, and the localized nature of the d states leads to deviations from the free-electron parabolic trends for ee scattering. The total relaxation times show that hot holes arising from d states lose energy on a sub-5-fs time scale, thus making such hot d holes challenging to extract before thermalization. Finally, our calculations yield comparable time scale and scattering rate for eph and ee scattering, in contrast with the common heuristic assumption that scattering by phonons occurs on a much slower time scale14. We thus conclude that previous models including only ee scattering14 and missing the important eph component are incomplete for understanding HC relaxation times and photoemission linewidths in noble metals.

To validate our results, we compare our relaxation times at the Fermi energy in Au and Ag with those inferred from carrier transport measurements. Kopitzki35 combined the Drude model with room temperature resistivity measurements, and obtained semiempirical relaxation times of 29 fs for Au and 41 fs for Ag, in close agreement with those obtained by Ashcroft and Mermin36 with the same approach. Averaging our relaxation times over a small energy window around EF yields relaxation times of 24 fs for Au and 27 fs for Ag. While the agreement is excellent for Au and good for Ag, our data shows a large spread in the relaxation times at EF, a feature not captured by the Drude model and stemming from the Fermi surface anisotropy. It is therefore puzzling that the Drude relaxation times33,35 are widely used in plasmonics4,5, given that their physical meaning is questionable

when applied to non-equilibrium situations involving SPPs. We conclude that further investigation on this point is necessary, and remark that ours is the rst truly ab initio determination of the relaxation times on the Fermi surface of noble metals, to be discussed in detail elsewhere.

HC MFP. Next, we turn our attention to HC transport. Figure 4 shows the MFP for the three crystallographic directions [100], [110] and [111] within 5 eV of the Fermi energy. The MFPs are obtained by multiplying the total relaxation times by the carrier velocities computed from the slope of the GW bandstructures (see Methods). Our computed MFPs are of the order 1040 nm in the energy window spanned by the s states, and much shorter (15 nm) for the d states. The MFPs of the s states exhibit an energy dependence with the volcano shape also seen in the relaxation times, with deviations coming from the different band velocities along each direction. The MFPs of hot holes with 12 eV energy are nearly isotropic, while the MFPs of hot holes with o1 eV energy are longer for the [100] and [110] directions. The MFPs of hot electrons are highly anisotropic due to the absence of electronic states up to 4 eV above EF in the [111] direction and at energy 41

2 eV in the [100] direction. The longest hot electron MFPs are in the [110] direction up to 4 eV above EF. We note that the energy derivative of the MFPs at EF at is negative, consistent with the positive thermopower in Au and Ag37. In the energy window spanned by the d states, the MFPs are of order 5 nm in the [111] direction and 1 nm in the other directions, and thus highly anisotropic with longer MFPs in the [111] direction. The isotropic MFPs for s holes and anisotropic for d holes are consistent with ndings in ballistic electron energy microscopy experiments in noble metals, where it was found that d holes injected by a scanning

40 5

[100] [110] [111]

Au d states

6 d states

Mean free path (nm)

0 5 4.5 4

4 3

4 2 0 2 4 Energy (eV)

Energy (eV)

Mean free path (nm)

[100] [110] [111]

0 4 2 0 2 4

Energy (eV)

Figure 4 | Hot carrier mean free path. The hot carrier MFPs along the [100], [110] and [111] crystallographic directions in Au (a) and Ag (b). The MFPs for the d states are expanded in the right panels.

6 NATURE COMMUNICATIONS | 6:7044 | DOI: 10.1038/ncomms8044 | http://www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8044 ARTICLE

Fermi energy

s states

Tunable HC generation

Energy

Intense HC generation

Short-lived d holes Low energy s electrons

Interband (d s) Intraband (s s)

E up to 2 eV, 2040 nm MFPs

Long-lived s electrons and holes

d states

Figure 5 | Regimes for hot carrier generation and transport in noble metals. The graph shows two HC generation regimes, for intraband (ss) and interband (ds) transitions induced by SPPs with energy EpoEint or

Ep4Eint, respectively. The regime with EpoEint offers a trade-off between optimal HC generation, moderate HC energies up to B2 eV, and longer MFPs and relaxation times than the interband regime. The dashed line indicates the Fermi energy.

tunneling microscope tip propagate along narrow cones and span small volumes, while s holes span large volumes isotropically38.

DiscussionThe data obtained so far allow us to formulate optimal conditions to generate HCs and extract or utilize them before thermalization. Figure 5 summarizes HC generation and transport in Au and Ag by dening two regimes for HC generation from intraband (EpoEint) and interband (Ep4Eint) transitions induced by SPP decay. The energy distribution, relaxation time and MFP of the generated HCs are dramatically different in these two regimes. For EpoEint, the HCs consist of s holes and s electrons with average energies up to Eint/2 (that is, 12 eV) generated by weakly damped SPPs due to the small FM-JDOS values. HC generation can be optimized by matching the energy and momentum of the SPP to the FM-JDOS intraband peak described above and shown in Fig. 1. The HCs in this energy range possess relatively long MFPs of B1040 nm, and can thus be extracted before thermalization over B50 nm lengths in ideally pure samples. This regime is thus optimal for HC generation and extraction. For hot electrons, we predict that the (110) surface would be best suited to extract HCs due to the longer MFPs in the [110] direction. For hot holes with B12 eV energy, the (100) and (110) surfaces are predicted to enable optimal extraction. Our results further suggest that Ag may be better suited than Au for HC generation due to the wider energy window for intraband transitions, thus motivating studies of HC generation from SPPs in Ag. On the other hand, the strong SPP damping regime with Ep4Eint is non-optimal for extracting HCs, since in spite of strong HC generation the SPP energy is mostly transferred to short-lived hot d holes with B1 nm MFP, while hot electrons possess only low energy of o1 eV. These conditions prevent HC extraction in this regime, unless a nanometer-thick thin lm or nanostructured metal is employed.

Our data further show that nanostructuring the metal to sub-10-nm size is not required to obtain energetic HCs, and not needed to extract them unless HCs with 42 eV energy are desired. In addition, for SPP with energy lower than Eint, the SPP

energy is equally distributed between hot electrons and holes, and energy loss is dominated by eph rather than Auger scattering. These ndings address the misconceptions emerged in recent work as outlined above, and show that optimal HC generation is possible by carefully tuning the SPP energy and wavevector at noble metaldielectric interfaces. Finally, we note that our

approach is for now limited to SPP at metaldielectric interfaces, while localized surface plasmons in nanostructures need to be treated differently due to their localized nature39. We believe that the method presented in this work can be extended to study HC generation from localized surface plasmon modes, which will be the subject of future investigation.

In summary, we establish a theoretical framework to study SPP damping and HC generation and transport on the same footing using many-body perturbation theory. Our rst-principles calculations can accurately describe SPP-induced generation and ultrafast scattering of HCs in noble metals of use in plasmonics, photocatalysis, photovoltaics and optoelectronics. Our work highlights the interplay of the s and d bands in noble metals, and prescribe optimal experimental conditions for generation and extraction of HCs. Our approach paves the way to rst-principles calculations of SPP losses in materials4.

Methods

First-principles calculations. We carry out ab initio calculations on Au, Ag and Cu in the face-centred cubic structure with lattice parameters of 7.72, 7.71 and6.82 bohr, respectively. The ground-state electronic structure is computed within the local density approximation (LDA)40 of DFT using the QUANTUM ESPRESSO code41. We use norm-conserving pseudopotentials (which include semicore s and p states) to describe the core-valence interaction42, together with a plane-wave basis set with kinetic energy cutoff of 60 Ry for Ag and Au, and 240 Ry for Cu. The ee (that is, GW) and eph self-energies are computed separately, and then combined together.

The real and imaginary parts of the GW self-energy19 are computed using the Berkeley-GW code43. The real part Re GWn;k is obtained with a generalized

plasmonpole calculation19 on a 12 12 12 k-point grid and then interpolated

using Wannier functions (see below). Kinetic energy cutoffs of 50 Ry and 120 Ry are used, respectively, for the screened and bare Coulomb interactions, together with B1,000 empty bands and a 8 8 8 q-point grid for the dielectric screening.

The DFT eigenvalues En;k are corrected with GW self-energies to obtain the quasiparticle bandstructures, En;k En;k Re GWn;k, used in all calculations in this

work. The imaginary part of the GW self-energy, Im GWn;k, is computed using full-

frequency GW calculations43. Here Im GWn;k are evaluated on-shell at the LDA

eigenvalues En;k, and then plotted versus the corresponding GW eigenvalues En,k. We use 20 20 20 k-point grids to converge Im GWn;k within B10 meV. In

addition, we use kinetic energy cutoffs of 20 Ry and 120 Ry, respectively, for the screened and bare Coulomb interactions, together with B100 empty bands and a 20 20 20 q-point grid for the dielectric screening.

The imaginary part of the eph self-energy, Im e phn;k, is computed using the

EPW code20. We compute the self-consistent potential and KohnSham states on a 12 12 12 k-point grid using DFT, and lattice-dynamical properties with

DFPT44 on a 4 4 4 q-point grid. The eph matrix elements are rst computed

on these coarse grids, and then obtained on signicantly ner grids using an interpolation procedure based on Wannier functions as implemented in the EPW code20,24. Maximally localized Wannier functions45 are obtained starting from ve d orbitals on the metal atoms and two s orbitals, each in a tetrahedral interstitial point of the face-centred cubic lattice, for a total of seven wannierized bands (we skip the semicore states). The Wannier interpolated and DFT bandstructures agree within 5 meV in an energy window up to 5 eV above the Fermi energy. The Bloch-to-Wannier rotation matrices are then used to interpolate the GW bandstructures, which are used to compute Im e phn;k. The ne grids for the calculation of

Im e phn;k consist in a 40 40 40 k-point grid, and up to 1 million random phonon q-points in the BZ. Such ne grids allow us to converge Im e phn;k within 1 meV. Further details of our approach to compute the eph self-energy are discussed in ref. 21.

The FM-JDOS calculations are carried out using Wannier interpolated GW bandstructures, a 100 100 100 k-point grid and a small Lorentzian broadening

ZE25 meV. The energy distribution of the generated HCs in Fig. 2 are obtained by counting the momentum- and energy-conserving transitions in equation (2) as a function of occupied band (for hot holes) and empty band (for hot electrons) quasiparticle energies.

The MFPs are obtained using velocities computed from the slope of the GW bandstructure along the given high-symmetry direction, together with total relaxation times tn,k combining the eph and ee relaxation times with Matthiessens rule. The total relaxation times are obtained from Im e phn;k computed on grids of over 100 k points along each high-symmetry direction, and

Im GWn;k computed on a 20 20 20 k-grid and then averaged over k points to

generate an energy-dependent Im GWn;k evaluated at each energy En,k for which a calculation of Im e phn;k is performed.

NATURE COMMUNICATIONS | 6:7044 | DOI: 10.1038/ncomms8044 | http://www.nature.com/naturecommunications

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8044

Derivation of the coupling matrix elements. The derivation of the plasmon-electron matrix element in Table 1 within the Landau damping approximation is given in ref. 26 and will not be discussed. Here we derive the coupling matrix elements for the photonelectron and SPPelectron interactions given in Table 1. We use SI units throughout the derivations. The photon case is well known27 and discussed here as a starting point for the SPP case. The second quantized vector potential ^

A r; t

for a photon with momentum :qp, energy Eq op and

polarization unit vector eq can be obtained by associating the amplitude with a destruction operator ^aq and its complex conjugate with a creation operator ^ayq .

We use the vector potential with unit amplitude since we are interested in the coupling matrix element, and thus the transition rate per unit of incident power. It reads27:

Aq r; t fq eq ^aq ei q r o t

;k q 9 Only the rst term in equation (8) gives a non-zero contribution, and we get:

Gabsq

The inverse process of photon emission from an initial state |S;0i to a nal state j0; 1q i has a rate Gemq given by:

Gemq

Xn;n ;k e2 2Vm2eoq

eq pn;n ;k qp

2 fn;k 1 fn

;k q

d oq En;k En

;k q

Xn;n ;k e2 2Vm2Eoq

eq pn;n ;k qp

e q ^ayq e i q r o th i

The net rate Gq Gabsq Gemq is the decay rate of a photon with momentum

:qp and energy Ep oq into a hot electronhole pair. Writing the delta

function as a Lorentzian with broadening Z-0, namely d x

1p Im

1x iZ, we

2 fn ;k q 1 fn;k

oq En;k En

;k q

where fq is a normalization constant. Using the Coulomb gauge rAq 0, the

second quantized electric eld can be obtained from ^

E @

A=@t.

obtain the decay rate as:

eq pn;n ;k qp

The normalization constant fq is determined by equating the energy of the classical and quantized elds. The classical (cycle averaged) energy hUq i for a

photon with momentum :qp in a homogeneous and isotropic material with dielectric constant E is hUq i VEE2q =2, where V is the volume of the material.

This energy needs to be equal to its quantum counterpart, oq nq 1=2. In

addition, hUq i can be obtained by averaging the quantized electric eld

Xn;n ;k e2 2Vm2Eoq

2 1

p Im

fn;k fn

;k q

Ep qp

;k q En;k

5 10

@ ^

Aq =@t over the quantum state jnq i with nq photons. We get:

D E

oq nq

This expression is identical to equation (1) with the squared coupling matrix elements jgPHn;n ;kj2 in Table 1. Since for photons qpE0 compared with the size of the

BZ, the squared coupling matrix elements are:

gPHn;n ;k

e m

2EEpV eq pn;n ;k qp 0

as shown in Table 1. We note that for slowly varying matrix elements in the BZ, the photon decay rate is proportional to the zero-momentum JDOS commonly used in the theory of the optical properties of solids 46, that is, Gq qp 0; Ep.

The derivation of the coupling matrix elements jgSPPn;n;kj2 for SPPs follows similar steps to the photon case discussed above. Following Nkoma et al.28, we study the case of a SPP generated at the interface between a metal with frequency-dependent dielectric function EM o and a dielectric material with frequency-independent

dielectric constant ED. A similar second quantization procedure was also developed by Elson et al.29. It is well known that the intensity of the SPP wave decays exponentially in the direction perpendicular to the dielectric-metal interface1,30. To study HC generation, we focus here on the SPP eld in the metal, and its associated vector potential. Since SPP possess transverse magnetic polarization, their wavevector can be taken as real in the xy planehere the plane of the metal dielectric interface, with area Aand purely imaginary in the z direction normal to the interface. With the metal in the z40 space and the x-axis parallel to the propagation direction of the SPP wave, the SPP wavevector is qp (qp,x,0,iqp,z).

While in the main text, we dene qp (qp,x,0,0) as the propagation wavevector in

the xy plane, for the purpose of this derivation qp is dened differently. Here qp,x is a real positive number since the electron-induced dissipation effects we aim to nd would be encoded in its imaginary part, and qp,z is related to the decay length Lz of the SPP eld intensity by30 qp,z 1/(2Lz). The relative magnitude of the

components of qp are determined by the dielectric properties of the metal and dielectric1,30, which impose qp;x=qp;z

ED=EM

D E

nq ^

E2q

D E

2 2VEo2q f 2q nq

where in the last equation we used the time derivative of equation (4) the electric eld and the commutation rules for boson creation and annihilation operators,

^ aq ; ^ay

h i

1. Equating equations (5) and (6) yields a normalization constant

fq = 2VEoq

h i1=2

, and a vector potential:

Aq r; t

2Veoq

!1=2

eq ^aq ei q r o t

e q ^ay

q e i q r o t

h i

The photonelectron interaction Hamiltonian is Hint em

^ p, as obtained by

^ to rst order in the vector potential.

The second quantized elds associated with the electronic Bloch states are:

C r

expanding the kinetic energy j^

p e

Aq j2= 2m

Xn;kun;k r eik r^cn;k

Xn ;ku n ;k r e ik r^cyn ;k

(here, the real parts of ED and EM should be employed for consistency). Similarly, the components of the unit polarization vector eTMq are related by30 eTMqx=eTMqz j

ED=EM

where ^cn;k and ^cy

n ;k are, respectively, fermion annihilation and creation

operators satisfying the anticommutation rule ^cn;k^cy

n ;k

^cy

n ;k ^cn;k dn;n dk;k . The

second quantized interaction Hamiltonian can be written as:

H0 Z

dr Cy r

HintC r

The vector potential of the SPP wave in the metal, with unit amplitude as in the photon case, can be written in second quantized form as:

Aq r; t fq eTMq^aq ei q r o t

e m

Xn;n k;k

2VEoq

!1=2

^ q e i q r o t

h i

eTMq

where, as in the photon case, the normalization constant fq can be determined by energy considerations. We write the classical electric eld associated with the SPP in the metal as E E0ei q r o t

. Using a result in ref. 28, the cycle-averaged energy in the SPP eldwhich accounts for contributions from both the eld in the dielectric and the metalcan be written as hUq i AEeff E0

j j2=4qp;z, where

we dened an effective dielectric constant:

Eeff o

EM o ED EM o

1 V

dr ei k k q

r o t

u n ;k r

eq ^

pun;k r

^ n ;k ^cn;k^aq

1 V

dr e i k k q

r o t

u n ;k r

e q ^

pun;k r

^ n ;k ^cn;k^ay

Crystal momentum conservation requires k0 qp, where the plus and minus

sign apply for the rst and second integral above, respectively. The interacting Hamiltonian becomes:

Xn;n ;k e2 2Vm2Eoq

!1=2

@EM o

We use the correspondence principle for a quantum state jnq i with nq SPP

quanta. Similar to the photon case, the total energy in the SPP mode oq nq 1=2 can be equated to its classical counterpart, while the square

modulus of the electric eld jEq j2 can be replaced with its quantum mechanical

average hnq j

n ;k q ^cn;k^aq h:c:h i

eq pn;n ;k qp

e io t^

E2q jnq i, with

Eq @

Aq =@t. We obtain the two equations:

D E

where h.c. is the hermitian conjugate, and we dened the momentum matrix element as the integral pn;n ;kqp 1O

dru n ;k q r

eq ^

pun;k r

oq nq

over the

unit cell volume O, as commonly used in the study of optical absorption46. We obtain the photon decay rate Gq by applying Fermis golden rule. Photon absorption occurs with a transition rate Gabsq between an initial state j0; 1q 4,

consisting of the ground electronic state at room temperature and one photon with momentum qp, and a nal state |S;0i with an electron in the Bloch state |n0,k qpi,

a hole in the Bloch state |n,ki and no photons. The absorption rate is given by:

Gabsq

jhS; 0 j H j 0; 1q ij2d oq En;k En

D E

Eeff A

4qp;z nq

D E

e 2q

E2q

A qp;z

Eeff o2q f 2q nq

By equating equations (12) and (13), and using qp,z 1/(2Lz), we obtain the

8 NATURE COMMUNICATIONS | 6:7044 | DOI: 10.1038/ncomms8044 | http://www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8044 ARTICLE

normalization constant fq =2ALzEeff oq 1=2, and thus the SPP vector

potential:

Aq r; t

2ALzEeff oq

!1=2

^ ayq e i q r o t

h i

This vector potential is similar to the photon case, equation (7). The key differences are the decay in the z direction for the SPP eld, which introduces an effective SPP volume ALz in the vector potential, the appearance of an effective dielectric constant for the medium and the fact that the polarization is transverse magnetic with a direction imposed by the properties of the metal and dielectric material.

Since the form of the SPP vector potential is completely analogous to the one for photons, the decay rate of SPPs to hot electronhole pairs can be carried out following the same steps as in the photon case discussed above, with the substitutions V-ALz, eq ! eTMq, and allowing for a nite momentum1 qp with magnitude qp Ep= c

EMED

eTMq^aq ei q r o t

eTMq

= EM ED

. This leads to the SPPelectron coupling matrix element in Table 1:

gSPPn;n;k

2 qp; Ep

e m

2Eeff EpALz eTMq pn;n ;k qp

References

1. Barnes, W. L., Dereux, A. & Ebbesen, T. W. Surface plasmon subwavelength optics. Nature 424, 824830 (2003).

2. Tame, M. et al. Quantum plasmonics. Nat. Phys. 9, 329340 (2013).3. Fakonas, J. S., Lee, H., Kelaita, Y. A. & Atwater, H. A. Two-plasmon quantum interference. Nat. Photonics 8, 317320 (2014).

4. West, P. R. et al. Searching for better plasmonic materials. Laser Photonics Rev. 4, 795808 (2010).

5. Khurgin, J. B. & Boltasseva, A. Reecting upon the losses in plasmonics and metamaterials. MRS Bull. 37, 768779 (2012).

6. Brongersma, M. L., Halas, N. J. & Nordlander, P. Plasmon-induced hot carrier science and technology. Nat. Nanotechnol. 10, 2534 (2015).

7. Clavero, C. Plasmon-induced hot-electron generation at nanoparticle/metal-oxide interfaces for photovoltaic and photocatalytic devices. Nat. Photonics 8, 95103 (2014).

8. Chalabi, H. & Brongersma, M. L. Plasmonics: harvest season for hot electrons. Nat. Nanotechnol. 8, 229230 (2013).

9. Tisdale, W. A. et al. Hot-electron transfer from semiconductor nanocrystals. Science 328, 15431547 (2010).

10. Mukherjee, S. et al. Hot electrons do the impossible: Plasmon-induced dissociation of H2 on Au. Nano Lett. 13, 240247 (2012).

11. Linic, S., Christopher, P. & Ingram, D. B. Plasmonic-metal nanostructures for efcient conversion of solar to chemical energy. Nat. Mater. 10, 911921 (2011).

12. Frischkorn, C. & Wolf, M. Femtochemistry at metal surfaces: nonadiabatic reaction dynamics. Chem. Rev. 106, 42074233 (2006).

13. Valle, F. Ultrafast spectroscopy of metals. C. R. Acad. Sci. Ser. IV Phys. 2, 14691482 (2001).

14. Echenique, P., Pitarke, J., Chulkov, E. & Rubio, A. Theory of inelastic lifetimes of low-energy electrons in metals. Chem. Phys. 251, 135 (2000).

15. Sundararaman, R., Narang, P., Jermyn, A. S., Goddard, III W. A. & Atwater, H. A. Theoretical predictions for hot-carrier generation from surface plasmon decay. Nat. Commun. 5, 5788 (2014).

16. Manjavacas, A., Liu, J. G., Kulkarni, V. & Nordlander, P. Plasmon-induced hot carriers in metallic nanoparticles. ACS Nano 8, 76307638 (2014).

17. Govorov, A. O., Zhang, H. & Gunko, Y. K. Theory of photoinjection of hot plasmonic carriers from metal nanostructures into semiconductors and surface molecules. J. Phys. Chem. C 117, 1661616631 (2013).

18. Martin, R. M. Electronic Structrure: Basic Theory and Practial Methods (Cambridge Univ. Press, 2008).

19. Hybertsen, M. S. & Louie, S. G. Electron correlation in semiconductors and insulators: Band gaps and quasiparticle energies. Phys. Rev. B 34, 5390 (1986).

20. Noffsinger, J. et al. EPW: A program for calculating the electronphonon coupling using maximally localized Wannier functions. Comp. Phys. Commun. 181, 21402148 (2010).

21. Bernardi, M., Vigil-Fowler, D., Lischner, J., Neaton, J. B. & Louie, S. G. Ab initio study of hot carriers in the rst picosecond after sunlight absorption in Silicon. Phys. Rev. Lett. 112, 257402 (2014).

22. Mahan, G. D. Many-Particle Physics (Springer, 2000).23. Schrieffer, J. R. Theory of superconductivity, Advanced book classics (Perseus Books, 1999).

24. Giustino, F., Cohen, M. L. & Louie, S. G. Electronphonon interaction using Wannier functions. Phys. Rev. B 76, 165108 (2007).

25. Grigorenko, A., Polini, M. & Novoselov, K. Graphene plasmonics. Nat. Photonics 6, 749758 (2012).

26. Pines, D. & Schrieffer, J. R. Approach to equilibrium of electrons, plasmons, and phonons in quantum and classical plasmas. Phys. Rev. 125, 804 (1962).

27. Loudon, R. The Quantum Theory of Light (Oxford Univ. Press, 2000).28. Nkoma, J., Loudon, R. & Tilley, D. Elementary properties of surface polaritons.J. Phys. C 7, 35473559 (1974).29. Elson, J. M. & Ritchie, R. H. Photon interactions at a rough metal surface. Phys. Rev. B 4, 41294138 (1971).

30. Raether, H. Surface Plasmons on Smooth and Rough Surfaces and on Gratings (Springer, 1988).

31. Zayats, A. V., Smolyaninov, I. I. & Maradudin, A. A. Nano-optics of surface plasmon polaritons. Phys. Rep. 408, 131314 (2005).

32. Cooper, B. R., Ehrenreich, H. & Philipp, H. R. Optical properties of noble metals. Phys. Rev. 138, A494A507 (1965).

33. Johnson, P. B. & Christy, R.-W. Optical constants of the noble metals. Phys. Rev. B 6, 43704379 (1972).

34. Paggel, J., Luh, D.-A., Miller, T. & Chiang, T.-C. Electronic-structure dependence of the electronphonon interaction in Ag. Phys. Rev. Lett. 92, 186803 (2004).

35. Kopitzki, K. & Herzog, P. Einfhrung in die Festkrperphysik (Springer, 2007).36. Ashcroft, N. W. & Mermin, N. D. Solid State Physics (Saunders College, 1976).37. Bourassa, R., Wang, S. & Lengeler, B. Energy dependence of the Fermi surface and thermoelectric power of the noble metals. Phys. Rev. B 18, 1533 (1978).

38. Blanco-Rey, M., Alducin, M., Juaristi, J. & de Andres, P. Diffusion of hydrogen in Pd assisted by inelastic ballistic hot electrons. Phys. Rev. Lett. 108, 115902 (2012).

39. Pitarke, J., Silkin, V., Chulkov, E. & Echenique, P. Theory of surface plasmons and surface-plasmon polaritons. Rep. Prog. Phys. 70, 1 (2007).

40. Perdew, J. P. & Zunger, A. Self-interaction correction to density-functional approximations for many-electron systems. Phys. Rev. B 23, 5048 (1981).

41. Giannozzi, P. et al. QUANTUM ESPRESSO: a modular and open-source software project for quantum simulations of materials. J. Phys. Condens. Matter 21, 395502 (2009).

42. Troullier, N. & Martins, J. L. Efcient pseudopotentials for plane-wave calculations. Phys. Rev. B 43, 1993 (1991).

43. Deslippe, J. et al. BerkeleyGW: A massively parallel computer package for the calculation of the quasiparticle and optical properties of materials and nanostructures. Comp. Phys. Commun. 183, 12691289 (2012).

44. Baroni, S., de Gironcoli, S., Dal Corso, A. & Giannozzi, P. Phonons and related crystal properties from density-functional perturbation theory. Rev. Mod. Phys. 73, 515 (2001).

45. Marzari, N. & Vanderbilt, D. Maximally localized generalized Wannier functions for composite energy bands. Phys. Rev. B 56, 12847 (1997).

46. Bassani, F. & Parravicini, G. P. Electronic States and Optical Transitions in Solids (Pergamon, 1975).

Acknowledgements

This research was supported by the SciDAC Program on Excited State Phenomena in Energy Materials funded by the US Department of Energy, Ofce of Basic Energy Sciences and by the Advanced Scientic Computing Research, under Contract No. DEAC02-05CH11231 at Lawrence Berkeley National Laboratory which provided for algorithm and code developments and simulations; and by the National Science Foundation under grant DMR 10-1006184 which provided for basic theory and formalism. Work at the Molecular Foundry was supported by the Ofce of Science, Ofce of Basic Energy Sciences and by the the US Department of Energy under Contract No. DE-AC02-05CH11231. This research used resources of the National Energy Research Scientic Computing Center, which is supported by the Ofce of Science of the US Department of Energy.

Author contributions

M.B. and J.M. carried out the calculations. All authors conceived the research and wrote the manuscript.

Additional information

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions

Web End =reprintsandpermissions /

How to cite this article: Bernardi, M. et al. Theory and computation of hot carriers generated by surface plasmon polaritons in noble metals. Nat. Commun. 6:7044doi: 10.1038/ncomms8044 (2015).

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

NATURE COMMUNICATIONS | 6:7044 | DOI: 10.1038/ncomms8044 | http://www.nature.com/naturecommunications

Word count: 9282

Show less

Abstract

Translate

Hot carriers (HC) generated by surface plasmon polaritons (SPPs) in noble metals are promising for application in optoelectronics, plasmonics and renewable energy. However, existing models fail to explain key quantitative details of SPP-to-HC conversion experiments. Here we develop a quantum mechanical framework and apply first-principles calculations to study the energy distribution and scattering processes of HCs generated by SPPs in Au and Ag. We find that the relative positions of the s and d bands of noble metals regulate the energy distribution and mean free path of the HCs, and that the electron-phonon interaction controls HC energy loss and transport. Our results prescribe optimal conditions for HC generation and extraction, and invalidate previously employed free-electron-like models. Our work combines density functional theory, GW and electron-phonon calculations to provide microscopic insight into HC generation and ultrafast dynamics in noble metals.

Details

Title

Theory and computation of hot carriers generated by surface plasmon polaritons in noble metals

Author

Bernardi, Marco; Mustafa, Jamal; Neaton, Jeffrey B; Louie, Steven G

Pages

7044

Publication year

2015

Publication date

Jun 2015

Publisher

Nature Publishing Group

e-ISSN

20411723

Source type

Scholarly Journal

Language of publication

English

DOI

https://doi.org/10.1038/ncomms8044

ProQuest document ID

1684974864

Theory and computation of hot carriers generated by surface plasmon polaritons in noble metals

Jump to:

Full Text

Abstract

Details

Suggested sources