Professional Documents
Culture Documents
https://doi.org/10.1038/s41560-021-00797-7
Although Li-based batteries have established a dominant role in the current energy-storage landscape, post-Li chemistries (for
example, Al or Zn) are emerging as promising candidates for next-generation rechargeable batteries. Electrochemical cells
using Al or Zn metal as the negative electrode are of interest for their potential low cost, intrinsic safety and sustainability.
Presently, such cells are considered impractical because the reversibility of the metal anode is poor and the amount of charge
stored is miniscule. Here we report that electrodes designed to promote strong oxygen-mediated chemical bonding between Al
deposits and the substrate enable a fine control of deposition morphology and provide exceptional reversibility (99.6–99.8%).
The reversibility is sustained over unusually long cycling times (>3,600 hours) and at areal capacities up to two orders of
magnitude higher than previously reported values. We show that these traits result from the elimination of fragile electron
transport pathways, and the non-planar deposition of Al via specific metal–substrate chemical bonding.
C
reating conformal metal coatings through electrochemical but ohmic heat generated by shorting poses added safety concerns
methods is important in multiple fields, including semicon- associated with thermal runaway14.
ductor manufacturing1, energy storage2, metal plating3 and so In theory, metal anodes that take advantage of micro- and/or
on. Although the governing chemistry is in principle simple—sol- nanopatterned conductive substrates that guarantee full access to
vated metal cations in a typically liquid electrolyte capture electrons electron and ion transport pathways throughout the plating/strip-
from an electrically conductive substrate, and are thereby reduced ping processes can surmount all of these challenges5,15. In conven-
into solid metallic form, for example, Mn+ (sol.) + ne− → M(s) tional battery configurations, ion transport occurs via interfacial
—the control over metal electrodeposition morphology remains contact between solid metal deposits and liquid electrolyte, which
challenging2,4. The search for fundamental solutions has recently means that control of the electrolyte volume and interface chem-
re-emerged as an area of scientific and technological interest because istry are sufficient to ensure a full ionic access. In contrast, elec-
of the role that the stable electrodeposition of metals in closed electro- tron transport relies solely on physical, solid–solid contact between
chemical cells plays in the stability and safety of rechargeable batteries the metal deposits and a conductive substrate. The hypothesis that
that utilize metal anodes, for example, Al, Li, Zn, Na and so on, to guides the current study is that the fragility of the electron trans-
achieve a higher energy and the low-cost storage of electrical energy. port pathways is the fundamental barrier to fully reversible metal
The working principle of such electrochemical cells is as follows: electrodeposition processes (Supplementary Fig. 1)6,10. Earlier
the reversible plating and stripping of metal at the anode enables work showed that, although electrodeposition of metals, such as
the reversible storage and release of electrical energy. In prac- Cu in patterned trenches for integrated circuits1,16 and Li and Na
tice, however, repeated formation and dissolution of the metallic in three-dimensional (3D) metal or carbon foams5,6, constrain the
phase is confounded by multiple problems that result in a porous, size of the electrodeposits, the transport length L and therefore the
non-planar deposition morphology5,6. A number of mechanisms electron transport timescale τ are broadly distributed, as the metal
have been proposed to explain these observations. The list includes is free to deposit in coarse, non-planar morphologies5.
intrinsic crystal anisotropy7, uneven solid–electrolyte interphase In this study, we investigated the electrodeposition of metals on
that drives heterogeneous growth4, concentration of electric field patterned substrates designed to strongly coordinate with the metal
lines at ‘hot spots’ for growth8 and growth limited by mass trans- deposit via surface chemical bonding (Fig. 1). Hypothetically,
port9. Porous metal deposits created in one plating cycle also have strong metal–substrate bonding increases the driving force for
a strong propensity to only partially dissolve in the next, to form nucleation to yield a uniform distribution of small metal nuclei.
the so-called ‘dead’ and/or ‘orphaned’ metals6,10–12. Over repeated The uniformly distributed nuclei enable compact, uniform deposi-
cycling, the accumulated metal deposits bridge the interelectrode tion, which produces deposits with a controlled thickness, L, dur-
space to short-circuit the cell10,13. All of these processes are obviously ing the growth phase. The straightforward effect of such control
fatal for the stable long-term operation of an electrochemical cell, is that the electron transport timescale τ ≈ L2/D, which sets the
Department of Materials Science and Engineering, Cornell University, Ithaca, NY, USA. 2Energy and Photon Sciences Directorate, Brookhaven National
1
Laboratory, Upton, NY, USA. 3Robert Frederick Smith School of Chemical and Biomolecular Engineering, Cornell University, NY, USA. 4Department of
Chemistry, State University of New York at Stony Brook, Stony Brook, NY, USA. 5Department of Materials Science and Chemical Engineering,
State University of New York at Stony Brook, Stony Brook, NY, USA. ✉e-mail: laa25@cornell.edu
Nano-Al L Micro-Al
Carbon
Nano-Al
Carbon Carbon
Fig. 1 | Metal–substrate bonding-induced regulation of electrodeposition. Strong surface bonding between Al and a conductive fibrillar carbon substrate
facilitates electronic transport and influences the morphological evolution of the Al anode. a, At the onset of electrodeposition, Al forms the interfacial
Al–O–C chemical bonds with the surface of carbon fibres. b, Guided by the chemical bonding interaction, Al grows laterally to form a uniform, compact
deposition layer that comprises nanoscale grains on the carbon fibres. c, After the available carbon surface is fully covered, the subsequent Al deposits
form microsized particles embedded among the carbon fibres. L denotes the electron transport length.
a b c
Time (h)
0 10 20 30 40 50 60
Al plating/stripping CE (%) 110
100 Short
90
80 d
70
60 100 µm
0 2 4 6 8 10 12 14 16
Cycle no.
e f g
e– e– e– e–
100 µm
Al GF SS Al GF SS
Fig. 2 | The propensity of Al anodes to exhibit heterogeneous growth on conventional substrates. a, CE measured in Al plating/stripping reactions at a
capacity of 3.2 mAh cm–2. Two layers of a GF separator were needed to prevent shorting. b, SEM analysis of Al electrodeposits formed on a planar steel foil
electrode at 4 mA cm–2 and for a capacity of 0.8 mAh cm–2. c,d, Elemental maps of Al (c) and Fe (d) from an EDS analysis of the electrode surface shown
in b. e, Schematic diagram showing the failure mechanism of Al electrodes. A higher areal deposition capacity and the accumulation of dead metal in the
GF membrane over multiple plating/striping cycles result in the proliferation of non-planar metal deposits in the interelectrode space. f, SEM image of Al
electrodeposits formed on a non-planar Ni foam electrode at a rate of 4 mA cm–2 and capacity of 1.0 mAh cm–2. g,h, Elemental maps of Al (g) and Ni (h)
from an EDS analysis of the electrode surface shown in f. SS, inert stainless-steel electrode.
above 99.5%. We note also that even with multiple separators, the Ni foam substrate (Fig. 2f–h) has, at most, a negligible impact on
3.2 mAh cm–2 cell fails after only ~60 hours. Together, (1) the rapid the deposition morphology. Consequently, using Ni foam as the
battery shorting and (2) the low plating/stripping reversibility at anode substrate exhibited a rather limited improvement on the elec-
elevated areal capacities impose fundamental, pressing challenges trochemical performance (Extended Data Fig. 1). The first clue to
for practical Al batteries. understanding the aggressive non-planar deposition of Al is that
To diagnose the underlying problems, we performed scanning glass fibre (GF) (SiO2) membranes, rather than organic polymers,
electron microscopy (SEM) to characterize the electrodeposit are used as the separator. This choice stems from the fact that typi-
morphology. The SEM results in Fig. 2b show that the Al deposits cal polyolefin separators are chemically unstable in the imidazolium
exhibit a heterogeneous, non-planar morphology. Consistent with chloride + AlCl3 electrolyte melt. Early literature reports show that
prior reports24, the Al deposits were as large as 20–50 μm, sparsely there is a possibility for a strong chemical affinity between metallic
distributed on the substrate and isolated from each other. This led Al and SiO2 via an interfacial reaction: 4Al + 3SiO2 → 2Al2O3 + 3Si
to large and broadly distributed L and τ values, which produced (refs 26,27), which implies that, in addition to the other factors that
sluggish, incomplete metal stripping, as illustrated in Fig. 2b–d. The may drive non-planar metal electrodeposition, a chemical driving
low plating/stripping efficiency and the battery shorting in the Al force favours the growth of Al electrodeposits into the separator.
anodes are therefore concluded to be consequences of the local- This growth mode was confirmed using SEM and energy dispersive
ized, aggressive growth of the Al electrodeposits normal to the sub- spectroscopy (EDS) analysis on either a stainless steel or a Ni foam
strate plane. We here follow the literature21 and loosely refer to the substrate (Supplementary Figs. 5 and 6), which explains the speed
non-planar Al electrodeposit growth as dendritic, but show below with which Al is able to short-circuit cells even when multiple layers
that it is unusually aggressive and possesses previously unreported of separator membranes are used or when the deposition is per-
characteristics that appear to arise from the strong affinity of Al for formed using a Ni foam substrate.
the separator (Fig. 2e).
The speed with which our Al electrodes fail by short-circuiting— Uniform Al deposition enabled by O-mediated metal–
even at conventional electrode areal capacities—exceeds typical val- substrate bonding
ues for Li anodes21, which implies that the materials physics that This discovery poses a perhaps obvious question: could the pro-
control the non-planar electrodeposition of Al are quite different. pensity of Al to grow into the separator be countered by creating
It also explains the rather low areal capacity limits explored in pre- metal–substrate bonding that is strong enough to chemically usher
vious studies. We note further that whereas deposition of Li in a Al into a patterned substrate? This question can be theoretically
3D porous material, such as a Ni foam, can substantially extend reasoned by an analysis of the nucleation process in electrodeposi-
the time-to-failure due to short-circuiting25, deposition of Al in a tion, which plays a pivotal role in determining the size, uniformity
50
40
No. of particles
30
20
10
0
500 nm 50 100 150 200 250 300 350
Particle diameter (nm)
c d
500 nm 500 nm
e f
Al
5 µm 5 µm
C
g h
C(002)
Intensity (a.u.)
Al(200)
C(100) Al(111)
90 Al(200)
Al(111) C(100)
60
25 30 35 40 45 50
C(002) 30
2θ (°)
Fig. 3 | Microstructure of Al metal deposits formed on a substrate with strong metal–substrate bonding. a,c–f, SEM analysis of the Al deposit
morphology at varying areal capacities: 0.2 (a), 1 (c), 3 (d), 3 (e) and 4 (f) mAh cm–2. Inset (e): SEM and corresponding EDS mapping of a selected region
in e. b, Particle size distribution at 0.2 mAh cm–2 (average = 139 nm). g, 2D XRD pattern of Al deposits on carbon fibres. h, Corresponding integrated XRD
line scan for the material in g. a.u., arbitrary units.
and distribution of electrodeposits8. On the application of a volt- can be dissociated into two steps: (1) electrochemical reduction of
age, the initial process is the desolvation of the coordinated species the metal, Mn+ + ne− → M(free), and (2) binding with the sub-
in the electrolyte as they approach the negatively charged electrode strate, M(free) + ∗ → M∗, which directly depends on the interaction
surface28,29. In this desolvation process, the metal-containing spe- between the metal and the substrate; the nucleation overpotential is
cies are converted into a cationic form by losing the coordinating determined by the free energy change ΔG of the latter30. Classical
ions and/or molecules. The subsequent step is the electrochemical nucleation theory then implies that strong metal–substrate bonding
adsorption of the species: Mn+ + ne− + ∗ → M∗, where * refers results in a smaller critical nucleate size, rc = −2γ/ΔG, where γ is the
to the adsorption site and M* to the adsorbed metal. This reaction energy penalty for creating a unit area of the metal surface and ΔG
a b c
Data Data 24,000 Data
12,000 Peak 1 Peak 1 Peak 1
Peak 2 3,000 Background Background
Intensity (c.p.s.)
Intensity (c.p.s.)
Intensity (c.p.s.)
Peak 3 Fit 18,000 Fit
Peak 4
8,000 Background
Fit 2,000
12,000
4,000 1,000
6,000
d e f
Data Data 24,000 Data
12,000 Peak 1 Peak 1 Peak 1
Peak 2 3,000 Peak 2 Peak 2
Intensity (c.p.s.)
Intensity (c.p.s.)
Intensity (c.p.s.)
Peak 3 Peak 3 Background
Background 18,000
Peak 4 Fit
Peak 5 Fit
8,000 2,000
Background
Fit 12,000
4,000 1,000
6,000
292 290 288 286 284 282 80 78 76 74 72 538 536 534 532 530 528
Binding energy (eV) Binding energy (eV) Binding energy (eV)
Fig. 4 | XPS characterization of the metal–substrate interaction. a–f, C 1s (a,d), Al 2p (b,e) and O 1s (c,f) spectra of Al samples in carbon fibre electrodes
in the as-deposited state (a–c) and the sputtered state (d–f). The peaks that correspond to the Al–O–C bond are shaded by grey and indicated in the plots
by vertical black lines. On sputtering, surface materials are removed from the samples by the incident Ar+ ion beam. Peak assignments are tabulated in
Supplementary Table 2. c.p.s., cycles per second.
is the Gibbs free energy associated with the formation of the con- altering the microstructure of the substrate. On these bases, in a
densed metal phase during the earliest stages of nucleation. non-planar architecture made of interwoven carbon fibres, the pos-
The implementation of this concept in a battery anode is not sible chemical metal–substrate bonding between C and Al mediated
straightforward, because the nucleation occurs on an electroni- by oxygen is then hypothetically able to promote uniform Al growth
cally conducting substrate, which means it does not strictly become on individual carbon fibres (Fig. 1), as opposed to the macroscopic
homogeneous. As discussed above, the formed metal nuclei may outer surface of the architecture, as shown in Supplementary Figs.
also make strong bonds with the separator, substrate or both, which 5 and 6.
will affect their shape. Additionally, to achieve a high electrode Figure 3 reports the main results of the present study. The Al
reversibility, the geometrical and chemical integrity of the substrate metal electrodeposition morphology on interwoven carbon fibres
must remain intact over repeated cycles of metal plating/stripping. was investigated using SEM. In stark contrast to the Al deposited
On this basis, a screening framework for promising substrate mate- on planar stainless steel and on non-planar Ni foam, for which
rial is proposed: the substrate should be electrically conductive, the deposit sizes are quite variable with average values of L on the
mechanically robust and capable of forming interfacial metal–sub- order tens of micrometres, the Al electrodeposits on the carbon
strate bonding. It is important that the chemical reaction between fibre surface with L typically in the range 100–200 nm in the nucle-
the substrate and the metal is limited within the interface, rather ation stage (Fig. 3a,b); this that for a fixed D, τ could be reduced
than penetrates into the bulk. A bulk reaction, owing to the crys- by a factor of 10,000 or more. As the deposition capacity increases,
tallographic and volume changes associated, inevitably gives rise to the nanoscale Al crystals grow laterally (Fig. 3c–e and EDS maps
morphological instabilities of the substrate over cycling. in Supplementary Fig. 9), that is expand along the carbon sur-
In screening for potential substrate materials, free-standing face to generate a conformal Al coating layer of thickness on the
interwoven carbon fibres (SEM image in Supplementary Fig. 7) are order of 102 nm (for example, 160 nm, see Supplementary Fig.
singled out because previous studies show that these materials form 10) that appear to strictly follow the surface curvature (Fig. 3d).
multiple types of interfacial covalent bonds, which include Al–C, Supplementary Fig. 10 also shows the morphology of the Al deposit
Al–O–C and so on, with Al metal during electroplating31,32. The cal- layer that faces the carbon fibre—nanoscale Al grains are seen to
culated binding energy of Al–O–C is as high as a few electronvolts31, merge into a compact layer. The Al deposition morphology is highly
which means that the driving force for Al to grow on the substrate uniform across a macroscopic area, as shown in Supplementary Fig.
surface is rather large. This characteristic was previously utilized 11, without any observable evidence of aggressive and/or dendritic
to electroplate a uniform Al precoating layer on carbon, which growth. Microstructural characterization of the morphology after
improved the Al/C wettability in subsequent metallurgical manu- thermal annealing shows that surface diffusion is not as effective a
facturing of Al/C composite materials31. This carbon-based mate- mechanism for smoothing the Al electrodeposits (Supplementary
rial is also of specific interest here because the bonding occurs via Fig. 12); rather, the initial uniform nucleation promoted by metal–
interfacial interaction at oxygen-enriched defective sites on the sur- substrate bonding plays a dominant role. These results confirm that
face rather than via bulk phase transition. Of note is that the Raman the strong metal–substrate bonding is able to (1) guide a metal elec-
spectrum of the interwoven carbon fibres shows both the D band trodeposition precisely using the surface topography of a strongly
and G band (Supplementary Fig. 8). This means that the carbon interacting non-planar architecture and (2) promote the uniform
contains a considerable amount of defects, which reportedly pro- nanoscale nucleation and lateral growth of metal deposits to create
motes the bonding with Al (ref. 31). X-ray photon electron spectros- a compact film.
copy (XPS) characterization (Extended Data Fig. 2) shows that the We increased the deposition capacity beyond 3 mAh cm–2 and
exposure of carbon fibres to the imidazolium chloride + AlCl3 elec- observed that the growth of Al enters a second stage (Fig. 1): growth
trolyte remarkably increases the level of oxygen enrichment without of microscale Al crystals on the compact nanoscale Al layer (Fig. 3f).
Al CE(%)
105 0.6
100 no.1
95 CE = 99.8% no.50 (500 h)
90
0.4 no.125 (1,250 h)
0 1,000 2,000 3,000 Time (h)
no.200 (2,000 h)
b
Potential (V)
0 200 400 600 800 Cycle no. no.360 (3,600 h)
105 0.2
Al CE(%)
100
95 CE = 99.6%
90 0.0
0 1,000 2,000 3,000 Time (h)
100
95 CE = 99.6% 0 2 4 6 8
90
0 1,000 2,000 3,000 Time (h) Capacity (mAh/cm2)
Fig. 5 | Electrochemical cycling behaviour of structured Al electrodes in galvanostatic plating/stripping experiments. a–c, Al plating/stripping efficiency
measured at areal capacities of 0.8 (a), 3.2 (b) and 8 (c) mAh cm–2. d, Voltage profiles measured during Al plating/stripping at a very high areal capacity
of 8 mAh cm–2. The current densities used for the respective measurements are 4 (a) and 1.6 (b,c) mA cm–2.
X-ray diffraction (XRD) (Fig. 3g,h) verified that only Al crystals on the sample surface was etched away. In this case, obvious new
were present on the carbon substrate fibres. The microscale Al peaks emerged (C 1s, 283.5 eV; Al 2p, 74.7–74.9 eV; O 1s, 531.9 eV),
deposits gradually filled the space among the carbon fibres (mor- which are ascribed to Al–O–C bonding. XPS peak assignments are
phology at 8 mAh cm–2 shown in Supplementary Fig. 13). The tabulated in Supplementary Table 2. The observation is consistent
microscale Al deposits were intimately connected to the compact with that of previous work31. The metal–substrate bonding is also
nanoscale Al layer and to each other. This observation confirms evidently mediated by the surface oxidation species of the carbon.
that the metal–substrate bonding has a decisive influence on the In contrast, XPS measurements of Al electrodeposits on control
Al deposition morphology. After the substrate surface was fully substrates suggest that the Al–O–C bonding and other metal–sub-
occupied by the compact nanoscale Al layer, the newly deposited Al strate bonding do not exist on stainless steel and Ni foam (Extended
crystals resumed the intrinsic growth mode at the micrometre scale. Data Fig. 4).
As a more direct control experiment, we artificially precoated the
carbon fibres with an interphase that blocked the possible Al–O–C Electrochemical performance of regulated metal anodes
bonding, and evaluated the Al electrodeposition morphology on To evaluate the reversibility of the Al deposited at these high areal
this ‘deactivated’ carbon fibre matrix (Supplementary Fig. 14). The capacities, we measured the plating/stripping CE of the Al elec-
SEM characterization shows that the Al deposition was no longer trodes in electrochemical cells. The CE reveals the ratio between the
uniform over the surface of the carbon fibres. This negative control amount of metal that can be stripped and the amount that is origi-
experiment validates our hypothesis that oxygen-enriched surface nally plated. The results are reported in Fig. 5. In stark contrast to
chemistry plays an indispensable role to realize the control over the low CE and rapid battery failure observed on a planar stainless
deposition morphology. steel substrate, anodes produced by Al electrodeposition in a pat-
Moreover, we devised physical simulation experiments to more terned substrate composed of carbon fibres manifested a high level
quantitatively understand the role played by a large driving force of reversibility (99.4–99.8%, stripped morphology in Supplementary
at the nucleation stage and its impact on the ultimate deposition Fig. 16) over a broad range of capacities (0.8–8 mAh cm–2). The
morphology (Extended Data Fig. 3, Supplementary Note 1 and results also showed that the cells maintain stable operations for
Supplementary Fig. 15). It was shown that with a nucleation driven more than 2,000 hours (Fig. 5a–c). The voltage profile in Fig. 5d
by a large overpotential of 2 V, uniform Al growth was achieved confirms that the plating/stripping reaction was stable, as evidenced
at the subsequent stage of deposition under conventional con- by the narrowing overpotential over the 3,600 hours of cycling, and
stant current conditions. This set of results confirms that a large does not show any trend towards deterioration. It is also notewor-
driving force imposed in the initial nucleation process serves as a thy that the CE at 8 mAh cm–2 was not compromised by the pres-
determinant factor in achieving uniform Al deposition. However, ence of microscale Al deposits, owing to what we suspect is a robust,
this physical approach is fundamentally associated with chemical interconnecting electronic pathway in the conductive non-planar
and hydrodynamic instabilities, which makes it unsuitable for bat- architecture that guarantees a good electrochemical access to the
tery applications (detailed discussion after Supplementary Fig. 15). metal and prevents the formation of dead or orphaned Al. When
Further molecular-level simulation of this process could be obtained paired with a cathode, this improvement could be clearly reflected
by joint density function theory methods33 that capture the detailed in proof-of-concept full cells at an elevated capacity of ~1 mAh cm–2,
kinetic steps near a solid–liquid interface, for example, desolvation, which is 1–2 orders of magnitudes higher than the values reported
bonding and interphase formation33–35. in the literature (Supplementary Figs. 17 and 18). Consistent with
To more thoroughly interrogate the nature of the metal–sub- the plating/stripping measurement, batteries using Al foil as the
strate bonding, we performed X-ray photoelectron spectroscopy. anode fail rapidly in <100 cycles. By contrast, ‘anode-free’ full bat-
Figure 4a–c shows the XPS results of the as-deposited sample teries with a carbon-fibre anode without any pre-stored Al showed
(3 mAh cm–2), which are consistent with the conclusion that no a stable cycling and excellent capacity retention over thousands of
possible metal–substrate bonding is detected on the surface. This cycles. These results of the full battery tests further confirm the high
observation is expected because the substrate architecture is fully reversibility of Al plating/stripping on carbon fibres when subject
covered by the compact deposition layer, which means that the to cycling.
metal–substrate interface is not exposed. Figure 4d–f shows the XPS As an extreme test of the approach, we performed plating/strip-
results of the sample after Ar+ sputtering, during which the material ping experiments at an unprecedentedly high current density, that
a b
5 µm 5 µm
c d
105
Shorting-induced
Al CE(%)
0.5 100
endless charging
95
40 mA; 0.4 mAh cm–2
0 90
Potential (V)
2nd cycle 95
40 mA; 1 mAh cm–2
–1.5 90
0.0 0.2 0.4 0.6 0.8 0 200 400 600 800 1,000 1,200
Areal capacity (mAh cm–2) Cycle no.
Fig. 6 | Ultrahigh current density plating/stripping of Al. a,b, SEM analysis of Al plated at a current density of 40 mA cm–2 on stainless steel (a) and on
carbon fibres (b). c, Voltage profile showing that the Al||stainless steel cells fail through short-circuiting by the second cycle. d, CE versus cycle index for
the Al plating/stripping process on carbon fibres with metal–substrate bonding.
is, 40 mA cm–2. SEM characterization revealed that diffusion-limited prior literature reports. It is apparent that even without taking into
ramified electrodeposit structures with classical fractal and/or den- account the intrinsic advantages of Al metal, which include safety,
dritic morphologies formed on the planar electrode (Fig. 6a). In cost, ease of manufacturing and high earth-crust abundance, Al bat-
contrast, nanoscale, compact Al electrodeposition was maintained teries based on the proposed anode design show enormous promise.
on the carbon-based fibres (Fig. 6b). As expected, the cells using To explore the broader utility of our concept, we studied the elec-
the planar stainless-steel foil failed quickly, after <5 minutes, due trodeposition of Zn, another metallic battery anode material that
to the short-circuiting process (Fig. 6c). This can be compared with has aroused enormous research interest recently36–38, in a patterned
the electrochemical results (Fig. 6d), which showed high levels of substrate composed of interwoven carbon fibres (Supplementary
reversibility and long lifetimes (for example, CE = 99.96% over Fig. 19). The galvanostatic plating/stripping results demonstrate
60,000 cycles at a current density of 40 mA cm–2 and an areal capac- that, unlike Al, Zn has a low plating/stripping efficiency and slow
ity of 0.4 mAh cm–2) with the regulation produced by the metal– kinetics on carbon fibres, which indicates that the substrate archi-
substrate bonding. tecture alone is insufficient to achieve highly reversible deposition.
The stable plating/stripping behaviours of Al under the regula- We artificially introduced a strong metal–substrate interaction by
tion of metal–substrate bonding can be understood in a quantitative coating the substrate with graphene, which we recently reported
manner. As Al forms a compact layer on the surface of the carbon strongly coordinates with Zn by an epitaxial mechanism owing to
fibres, the electron transport length scale is maintained as small. As a minimized interfacial energy36. The plating/stripping reversibility
a quantitative estimation, the characteristic relaxation time τ of the and lifetime are improved remarkably by this coating. SEM results
nanoscale Al deposits formed in the patterned substrate is four orders showed that the presence of the strong metal–substrate interac-
of magnitudes smaller than that of the dendritic Al (L ≈ 160 versus tion via the graphene interphase can effectively promote the uni-
20 μm) formed in the planar case. This means that the nanoscale form deposition of nanoscale, plate-like Zn metal with a minimized
Al can be stripped around four orders of magnitude faster. In addi- electron transport length scale, in comparison with the microscale
tion to considerations based on transport, the non-planar and/or Zn deposits observed on bare carbon fibres (Supplementary Fig.
dendritic metal deposit is associated with a stronger morphologi- 19). Moreover, the effectiveness of the metal–substrate bonding
cal instability—as stripping occurs at the bottom of the deposits, can be demonstrated on planar substrates, which is a more con-
the whole structure detaches from the electrode6. Our observations, ventional configuration in commercial batteries. We evaluated the
therefore, suggest that an Al anode in which the metal is deposited electrochemical plating/stripping of Al on stainless steel coated
in a patterned substrate capable of forming strong bonds with Al are by carbon nanotubes with and without carboxylic side groups.
of potential immediate interest for applications in Al batteries that As shown in Supplementary Fig. 20, the planar substrates coated
operate under various conditions. In Table 1, we compare the elec- with carboxylic-functionalized carbon nanotubes manifested stable
trochemical performance achieved for our Al anodes with those of cycling and a high CE, in comparison with those of bare stainless
Extended Data Fig. 1 | Electrochemical plating/stripping behavior of Al metal on nonplanar nickel foam substrate. a, Coulombic efficiency obtained at
0.8 mAh, 4 mA/cm2. Voltage profiles of Al plating/stripping: (b) 0.8 mAh, 4 mA/cm2;(c) 3.2 mAh, 1.6 mAh/cm2 and (d) 8.0 mAh, 1.6 mA/cm2. The results
mean that the improvement made by using a nonplanar, inert architecture is very limited, particularly at practical capacities, that is 3.2 and 8 mAh/cm2.
a b C=O C-O
c
Data Data Data
45,000
Intensity (cps)
15,000
0
290 288 286 284 290 288 286 284 290 288 286 284
Binding Energy (eV) Binding Energy (eV) Binding Energy (eV)
Extended Data Fig. 2 | XPS spectrum of interwoven carbon fibers. a, Pristine, (b) after exposure to IL+AlCl3 electrolyte, (c) after exposure to dimethyl
carbonate as a negative control. Peak assignments: 284.8 eV (C-C, C-H), 286 eV (C-O), 287.5–288 eV (C=O, O-C-O), 289 eV (O=C-O). The intensities at
286 and 287.5 ~ 288 eV suggest that the exposure to IL+AlCl3 notably increases the level of oxygen enrichment on carbon fibers.
Extended Data Fig. 3 | SEM images and EDS mapping of Al deposition morphology obtained using a sequential, two-step protocol. In Step I, a small
capacity (0.05 mAh/cm2) of Al is deposited at fixed overpotential, η values (for example, η = 0.05, 0.3, 1.0, 2.0, 3.0 V); In Step II, a greater areal capacity
(that is 0.45 mAh/cm2) of Al is galvanostatically deposited at a current density (that is 4 mA/cm2). SEM images of the Al deposition morphology obtained
using this protocol, with the η value equal to (a) 0.05 V, (b) 0.3 V, (c) 1.0 V, (d) 2.0 V, and (e) 3.0 V. f–o, the corresponding EDS mapping results. See
Supplementary Note 1 for a detailed discussion.
Intensity (cps)
Intensity (cps)
Peak 1 Peak 1 Peak 1
Peak 2 Peak 2 Peak 2
Peak 3 30,000 Peak 3 30,000 Peak 3
Peak 4 Peak 4 Peak 4
Intensity (cps)
Intensity (cps)
Peak 1 Peak 1 Peak 1
Peak 2
Peak 3
16,000 Peak 2
Peak 3
16,000 Peak 2
Peak 3
Peak 4 Peak 4 Peak 4
4,000
8,000 8,000
292 290 288 286 284 282 292 290 288 286 284 282 292 290 288 286 284 282
Intensity (cps)
Intensity (cps)
Peak 1 Peak 1 Peak 1
Background Background 36,000 Background
18,000 Fit Fit Fit
40,000
12,000 24,000
20,000
6,000 12,000
O 1s 24,000 Data 60,000 Data Data
Intensity (cps)
Intensity (cps)
Intensity (cps)
Peak 1 Peak 1 Peak 1
Peak 2 Peak 2 36,000 Background
18,000 Background
Fit
Background
Fit
Fit
40,000
12,000 24,000
20,000
6,000 12,000
538 536 534 532 530 528 538 536 534 532 530 528 538 536 534 532 530 528
Peak 1
Intensity (cps)
Intensity (cps)
Peak 1 Peak 1
3,000 Background Background 9,000 Background
Fit Fit Fit
8,000
2,000 6,000
4,000
1,000 3,000
Al 2p
Data 12,000 Data Data
Intensity (cps)
Peak 1
Intensity (cps)
Intensity (cps)
Peak 1 Peak 1
3,000 Peak 2 Peak 2 Peak 2
Peak 3 Background 9,000 Background
Background Fit Fit
2,000
Fit 8,000
6,000
1,000
4,000 3,000
80 78 76 74 72 80 78 76 74 72 80 78 76 74 72
Binding Energy (eV) Binding Energy (eV) Binding Energy (eV)
Extended Data Fig. 4. | XPS spectrum of Al electrodeposits on carbon fibers, stainless steel and nickel foam. XPS of Al deposited on (a)–(c) carbon
fibers, (d)–(f) stainless steel, and (g)–(i) Ni foam. (a)(d)(g) C 1 s spectra; (b)(e)(h) O 1 s spectra; (c)(f)(i) Al 2p spectra. Upper panels and lower panels
show spectra before and after Ar+ sputtering, respectively. After sputtering, the Al-O-C bonding was observed on samples where Al was deposited on
carbon fibers. On other samples, no characteristic metal-substrate covalent bonding is observable.