New design for highly durable infrared-reflective coatings

The fundamental challenge in designing durable infrared-reflective coatings is achieving the ideal combination of both high reflectivity and durability. Satisfying these competing demands is traditionally achieved by deposition of durable layers on highly reflective metals. We overturn the traditional logic of ‘first reflectivity and then durability’ and propose an alternative of ‘first durability and then reflectivity’: First, a transition-metal compound is selected as a durable base; then its reflectivity is improved by incorporating silver/gold to form an alloy or by overcoating a multilayer stack. Two validation experiments prove that the new strategy works extremely well: the coatings thus obtained have infrared reflectivities close to that of aluminum, and their hardness and acid and salt corrosion resistances are 27–50, 400–1 500 and 7 500–25 000 times that of aluminum. The traditional mirror coating (e.g., Al/SiO2 films) is more suitable for moderate environments, while our mirror coating that was obtained by the new strategy (e.g., an Ag-doped hafnium nitride film) is more suitable for harsh environments, such as ones with dust, windblown sand, moisture, acid rain or salt fog. This work opens up new opportunities for highly durable infrared-reflective coatings and rejuvenates the study of transition metal compounds in a completely new area of optics.


INTRODUCTION
Recently, there has been an increasing need for durable infraredreflective coatings on large infrared telescopes 1,2 , night-vision systems 3 , pointing and tracking systems 3 and other optical systems [4][5][6][7] . The ideal durable infrared-reflective coatings require not only high reflectivity in a particular infrared band but also durability to resist scratches and corrosion damage from long-term environmental exposure in environments with dust, windblown sand, moisture, acid rain and salt fog 8,9 . Unfortunately, ideal durable infrared-reflective coatings are hard to obtain, as reflectivity and durability are a classical dilemma to some extent, difficult to obtain simultaneously. In Al, for example, the abundance of free electrons renders an excellent infrared reflectivity, but the electrons and the metallic bonds result in high corrosion and a low hardness, leading to a poor durability 8,10 . In practice, Al reflective coatings are deposited in primary mirrors of a size up to 3.8 m in diameter in large infrared telescopes 1 , and those coatings have to be replaced every 2-3 years because of light scattering and reflectivity degradation resulting from scratches and corrosion. Replacing the reflectors costs extremely large amounts of time and money 9 . The sheer size of these enormous optical systems alone causes operational difficulties in the cleaning of the old coatings and the re-deposition of new coatings, thus making the process time-consuming and expensive [9][10][11] . Currently, achieving highly durable infrared-reflective coatings is an open and urgent challenge.
Traditionally, after the high infrared-reflective metal films such as Al, Ag and Au are deposited (step 1) (Figure 1a), transparent, hard metal oxide layers, such as SiO 2 and Al 2 O 3 26,27 , are deposited on top (step 2) for protection, as these reflective metal films cannot withstand corrosion, scratches and other environmental damage. This route follows the typical logic of 'achieving reflectivity first and then coating to protect'. As such, oxide films with a very high hardness 28 are deposited on very soft metal films 12 . To avoid reflectivity loss and poor adhesion, the hard protection layer has to be very thin, usually a few dozen nanometers in thickness; thus the protection layer is often ineffective at resisting scratch damage 9,10,29 . Additionally, oxide films of such a thickness usually contain pinholes that undermine the protection against corrosion 10 . Therefore, the metal/oxide combination obtained from the traditional route is far from the ideal durability [8][9][10][11]29 . The traditional route also results in processing complexity because it usually involves the preparations of a variety of films to meet the multiple needs of high reflectivity, high durability and good adhesion 8,10 .
In this paper, we overturn the traditional 'first reflectivity and then durability' logic and propose a new strategy of 'first durability and then reflectivity' (Figure 1a). Both statements of 'first durability and then reflectivity' and 'first reflectivity and then durability' refer to the design logic rather than the actual sequence during the preparation process. We recommend transition metal compounds, TMRs (TM = Ti, Zr, Hf, V, Nb, Ta, Cr, Mo, W; R = B, C, N), as the durable base material and then improve its reflectivity by incorporating silver/gold to form an alloy or by overcoating a multilayer stack to achieve the integration of durability and reflectivity. We use hafnium nitride (HfN x ) to verify these two methods, and the experimental and theoretical results are exciting, which proves that our new strategy works wonderfully and simplistically.

Film growth
All the films were deposited simultaneously onto optical glass and single-crystal Si (001) substrates using magnetron sputtering. The work pressure of the discharge gas was maintained at 1.0 Pa for all the deposition process. Before the deposition, the glass and Si (001) substrates were successively cleaned in acetone, alcohol and deionized water using the ultrasonic cube, and then they were introduced into the vacuum chamber (base pressure of 4 × 10 − 4 Pa) of the sputtering system. For the preparation of HfN x films, a pure Hf target was chosen as the cathode and the mixture of N 2 and Ar was used as the discharge gas. During the deposition, the flow rate ratios of N 2 /(Ar+N 2 ) were increased from 3 to 100%, which was controlled by the mass flow controllers. The stoichiometry (x) of the HfN x films was changed from 1.039 to 1.396. The other preparation parameters were maintained constant as follows. RF power applied to the Hf target: 150 W; substrate bias: − 80 V; substrate temperature: 200°C. By varying the ratios of N 2 /(Ar+N 2 ), we prepared HfN-only multilayer films (see Supplementary Section 7 for the parameter selections and preparation details). Additionally, we prepared Ag-doped HfN x films by cosputtering the Hf and Ag target in the gas mixture of N 2 and Ar. During the deposition, the content of Ag was controlled by changing the RF power applied to the Ag target from 0 to 100 W, while the DC power applied to the Hf target was kept constant at 150 W. Other preparation parameters were kept constant as follows. substrate bias: − 160 V; substrate temperature: 200°C; Ar flow rate: 80 sccm; N 2 flow rate: 2.8 sccm; sample rotation rate: 5 r min − 1 . For a better comparison, an undoped HfN x film was also prepared under the same deposition conditions, except that the RF power of the Ag target was turned off. We prepared Al/SiO 2 films by sputtering Al and then SiO 2 target in the discharge gas of Ar. The preparation parameters were kept constant as follows. Ar flow rate: 80 sccm; DC power applied to the Al target: 60 W; RF power to the SiO 2 target: 100 W.  Reflectivity and durability in one Figure 1 Design rules for durable infrared-reflective coatings. (a) The current practice and a new strategy for the combination of high reflectivity and high durability. The current practice is 'first reflectivity and then durability', that is, after the high infrared-reflective metal films are deposited first, transparent hard metal oxide layers are deposited on top for durability enhancement. The new strategy emphasizes 'first durability and then reflectivity', that is, finding a material first that satisfies the durability requirement and then modifying it to obtain a satisfactory reflectivity. The routes are to select a transition metal compound (TMR) as a durable base and then improve its reflectivity by overcoating a multilayer stack or by incorporating silver/gold to form an alloy or a nanocomposite, where TMR represents borides, carbides and nitrides of the groups IVB, VB and VIB transition metals. (b) Hardness 12,13 and infrared reflectivity 14-25 at a wavelength of 2000 nm of TMR as well as Al, Au and Ag, where hafnium nitride (a red pentagram) is chosen as the durable base for the validation experiments.
Highly durable infrared-reflective coatings C Hu et al

Film characterization
A high-resolution transmission electron microscopy (JEM-2100F, JEOL, Tokyo, Japan) and grazing-incidence X-ray diffraction measurements (D8tools, Cu Kα, Bruker, Karlsruhe, Germany) were used to characterize the structures of films. An X-ray photoelectron spectroscopy (XPS, VG ESCA LAB MKII, Thermo Fisher Scientific, Waltham, MA, USA) with a monochromatized Al Kα (1486.6 eV) X-ray source were carried out to determine the stoichiometry x, core-level spectra and valence band spectra of the HfN x films. Before the measurement of XPS, all the samples were subjected to a 180 s Ar + cleaning procedure to remove the surface carbon and oxygen. A Dektak surface profiler and four-point probe measurements were used to determine the thickness d and DC resistivity ρ of the films, respectively. A UV-visible spectrometer (Lambda 950, Perkin Elmer, US) was employed to obtain UVvisible reflectivity and transmission spectra, and an FTIR spectrometer (Perkin Elmer Spectrum One B type, Perkin Elmer, US) was employed to measure infrared reflectivity spectra. By analyzing the transmission spectra, we obtained the refractive indices and absorption coefficients, using a procedure reported by Swanepoel 30 . According to the Tauc equation, we plotted (αhν) 2 against the photon energy hν and calculated the optical gaps. An nanoindenter (MTS XP, MTS, US) was used to evaluate the hardness, where a continuous-stiffness-measurement mode was used. Hall-effect measurements (HL5550) were performed to obtain the concentrations of the free electrons of the films. An energy dispersive spectrometer equipped in a field-emission scanning electron microscope (SU8010, Hitachi, Tokyo, Japan) was used to determine the chemical composition of the Ag-doped HfN x films. The corrosion behaviors of the samples were measured using the Tafel curves in acid and base media corresponding to 0.5 mol l − 1 H 2 SO 4 and 3.5 wt.% NaCl solutions, respectively. An electrochemical workstation (CHI660E) was used to perform the corrosion measurements, which was connected to a three-electrode electrochemical reactor. The working electrode, reference electrode and auxiliary electrode were the coated samples, calomel and a Pt sheet, respectively. Before the measurements, the samples were tested in an open circuit potential mode for 6 min with a scanning rate of 1 mV s − 1 . Furthermore, the corrosion behavior of the films in a seawater environment was evaluated using salt-bath experiments. The samples were immersed in a 5.0 wt.% NaCl solution and incubated at 35 ± 1°C for 1, 5, 10, 15, 30, 60, 120 and 180 min and 10 days (14 400 min). A reflection fluorescence microscopy (DM 2500M, Leica, Wetzlar, Germany) was employed to observe surface corrosion morphology.

First-principles calculations
The present calculations were performed by the method of projector augmented-wave pseudopotentials with density functional theory coded in the Vienna ab inito simulation package 31,32 . For the electrons' exchange correlation energy, the Perdew-Burke-Ernzerhof function was used 33 . The kinetic energy cutoff is chosen as 550 eV for the plane wave expansion. The Brillouin zones were sampled with Monkhorst-Pack method. In order to make sure the convergence of total energy at 1 meV per atom level, the Γ-centered high-density k-point grid sets were chosen. In the calculation, we have considered the effect of spin polarization. Detailed modeling process can be seen in Supplementary Sections 3,4 and 10.

RESULTS AND DISCUSSION
Optical design for achieving highly durable infrared-reflective coatings To achieve highly durable infrared-reflective coatings, we propose a new strategy of 'first durability and then reflectivity' (Figure 1a), in other words, first finding a material that satisfies the durability requirement and then modifying it to obtain the required infrared reflectivity. These two steps are illustrated in detail as follows.
Step I. Selection of a durable base. A high concentration of free electrons in a material induces a large plasma energy (Equation (1)), which in turn affects the dielectric function (Equation (2)). The relationship between the reflectivity and dielectric function can be described by Equations (2)(3)(4)(5) Refs. 34-37, where the functions ε(E), n(E), k(E) and R(E) represent the complex dielectric function, refractive index, extinction coefficient and reflectance versus photon energy E, respectively; the variables n, m*, ε 0 , ε ∞ , E p and Γ D represent the concentration of free electrons, effective electron mass, vacuum permittivity, background constant, plasma energy and relaxation energy, respectively; and ε 1 (E) and ε 2 (E) are the real part and the imaginary part of ε(E), respectively. According to these Equations, ε 2 1 E ð Þ and ε 2 2 E ð Þ increase with E p (Equation (2)), which induces an increase in n(E) and k(E) (Equations (3 and 4)) and a subsequent increase in R(E) (Equation (5)). The proportional relationship above between E p and R(E) has been well demonstrated by previous simulations and experiments 38 . Hence, a sufficiently high concentration of free electrons is necessary for a high infrared reflectivity; metals such as Al, Ag and Au are good examples 37,39 . A durable material is usually associated with high hardness, which is, in essence, proportional to the degree of covalent bonding and the bond strength 40 . Therefore, to achieve a high durability and a high reflectivity, the ideal candidate should contain strong covalent bonds and a high concentration of free electrons. The borides, carbides and nitrides of groups IVB (d 2 s 2 ), VB(d 3 s 2 ) and VIB (d 4 s 2 ) transition metals, hereinafter referred to as TMRs, where TM = Ti, Zr, Hf, V, Nb, Ta, Cr, Mo and W and R = B, C and N, have not only strong TM-R quasi-covalent bonds from the hybridization between the R_p and TM_d orbitals but also high concentrations of unbound d-orbital free electrons. The strong quasi-covalent bonds endow TMRs with a superior durability, including high bulk moduli 41 , high hardness [42][43][44][45][46] , high melting point [47][48][49] and corrosion and abrasion resistance [50][51][52][53] , and makes them well known as cutting-tool coating materials [54][55][56] . The presence of unbound d-orbital free electrons causes TMRs to have similar electrical properties and infrared reflectivity characteristics to pure metals 57 , causing them to be widely used in the fields of superconducting materials 58,59 and optoelectronics 60,61 . From these studies, it is known that TMRs have high durability and metal-like reflective characteristics ( Figure 1b); thus TMRs are a category of ideal candidates to achieve the ultimate aim of both a high reflectivity and a high durability.
Step II. Reflectivity enhancement. From Figure 1b, the hardness of a TMR (14-30 GPa) is much higher than that of the reflective metals Al, Ag and Au (~0.5 GPa), but their reflectivity (40-80%) is far below that of these metals (90-98%). This explains why TMRs were basically excluded from the 'radar screen' in the search for infrared-reflective coatings. How to significantly increase the reflectivity of a TMR will be the key to solving the issue of needing a high durability and a high reflectivity. Taking advantage of the structural characteristics of TMRs, we propose two methods to improve the reflectivity of TMR films. One method is to deposit multilayer films on a durable TMR base to boost the reflectivity (Figure 1). These multilayer films are obtained by alternately depositing a transparent layer A and a transparent layer B on a durable TMR film, namely, TMR/A/B/ ... /A/B/ (determination of the number of layers is discussed later). According to the optical interference principle 62,63 , the multilayers can achieve a very high reflectivity close to 100% in the vicinity of a target wavelength when the refractive index of layer A is far less than that of layer B, and the optical thickness nd (where n is the refractive index and d is the film thickness) of layer A and layer B both equal a quarter of the target wavelength. The other method is to introduce a metal with a high concentration of free electrons (e.g., Ag, Au) into a TMR film to form a metal-TMR alloy or a metal-TMR nanocomposite (Figure 1a). With a high concentration of free electrons, the alloys and nanocomposites are expected to have high reflectivities over the whole infrared range. Based on this principle, we incorporate gold or silver into a TMR film and explore the relationship among the composition, structure, reflectivity and durability.
Verification experiments for achieving highly durable infraredreflective coatings From Figure 1b, the near-stoichiometric HfN films (HfN x , x = N:Hf) in a rock salt structure have not only a high infrared reflectance of 75% but also a high hardness of 22.6 GPa, very close to idealistic highly durable infrared-reflective coatings. Near-stoichiometric HfN x is thus our first choice as the base material to carry out the below mentioned two aspects of studies to boost infrared reflectivity: (I) depositing multilayers on top of the HfN x film to boost the infrared reflectivity at a single wavelength; (II) doping silver into the HfN x film to boost the infrared reflectivity over a wide range of wavelengths.
Achieving highly durable infrared-reflective coatings for a specific wavelength through multilayering. To verify the reflectivity boost at a specific target wavelength, we develop a novel homogeneous multilayer film with three significant optical characteristics consisting entirely of HfN x . The creation of this multilayer originates from the unique behavior of electron localization in HfN x and the resulting Highly durable infrared-reflective coatings C Hu et al tunable reflectivity/transmission properties. According to the measurements of the electron concentration (Figure 2a), the resistivity ( Figure 2b) and the optical gap (Figure 2c), we find that, as the N/ Hf ratio increases from 1:1 (a measured x value of 1.039) to 4:3 (a measured x value of 1.334), the free electrons of HfN x films are completely localized and the films transform from a metal to a semiconductor with an optical gap of approximately 2.5 eV (see Supplementary Section 1 for more detailed discussion). Our electron concentration ( Figure 2a) and resistivity measurements (Figure 2b) show that the electron localization goes through two stages of transition from slow to fast. In the first stage, the electron concentration decreases slowly from 1. 46  Understanding the electron localization process, we explore the effect of electron localization on the optical properties of the films. We determine from the transmission spectra (Figure 2d) that before the electron localization, the HfN x film (x = 1.039) is completely opaque, a typical metal-like characteristic, so that the transmittance of this sample is zero and there are no interference fringes. After the electron localization (x = 1.334 and 1.396), the films transform into transparent semiconductors, and thus interference fringes occur in the range of 500-2 500 nm (Figure 2d). According to the Equation 2nd = mλ ( ref. 30 ), a larger optical thickness nd (where n is the refractive index and d is the film thickness) produces more interference fringes. Therefore, the interference fringes of the film with x = 1.334 (nd = 2.78 × 625 nm) are more pronounced than those of a film with x = 1.396 (nd = 2.08 × 217 nm) due to the film's larger refractive index and film thickness. Furthermore, the refractive index is found to decrease from 2.78 to 2.08, while the extinction coefficient remains approximately 0 (o0.001) as x increases further from 1.334 to 1.396 (Figure 2e), which is attributed to the decrease in the average molecular polarizability of the film because the polarizability of N atoms (α N = 1.1 × 10 − 24 cm 3 ) ( ref. 64 ) is much smaller than that of Hf atoms (α Hf = 15.3 × 10 − 24 cm 3 ) ( ref. 65 ). These results suggest that the optical characteristics of the HfN x films are easily controllable by changing the stoichiometry x (Figure 2f): if x is approximately 1, the film is an opaque metal (OM); if x = 1.334-1.342, the film is a high-refractive-index transparent semiconductor (HT, n = 2.78-2.69); if x = 1.383-1.396 the film is a low-refractive-index transparent semiconductor (LT, n = 2.17-2.08).
Using this tunability of the reflectivity/transmission properties, we develop multilayers consisting of hafnium-nitride-only films: OM/(LT/ HT) z , where z is the total number of repeating layers (Figure 3a and 3b). For the sake of simplicity, we refer to this structure as a 'multilayered film'. The reflectivity of a multilayered film depends on the total number of repeating layers. According to the principle of optical interference, the more repeating layers there are (i.e., the higher the value of z), the better the reflectivity enhancement is. However, increasing the number of repeating layers increases the deposition time and the processing difficulty. Everything considered, we set the total number of repeating layers as high as 6. Thus a periodic LT/HT stack with z = 6 is designed (see Supplementary Section 5). For the multilayers with λ 0 = 1 900 nm, the refractive indices of the LT and HT layers are n 1 = 2.08 at x = 1.396 and n 2 = 2.69 at x = 1.342, respectively. According to n 1 d 1 = n 2 d 2 = λ 0 /4, when λ 0 = 1900 nm, the thickness of the two layers are d 1 = 228 nm and d 2 = 177 nm. Similarly, for λ 0 = 4100 nm, n 1 = 2.17 at x = 1.383 and n 2 = 2.78 at x = 1.334, d 1 and d 2 are 472 and 369 nm, respectively (see Supplementary Section 6 for more details). To verify the above designs, we prepare the multilayered film of OM/(LT/HT) 6 and find that the reflectivity of the OM layer improves tremendously (Figure 3c and 3d). We obtain experimentally the refractive index of each layer by controlling the composition of the films according to the relationship between the refractive index and the composition (Figure 2e). We obtain the desired thickness of each film by adjusting the deposition time. The detailed deposition parameters for the two multilayered films are given in Supplementary Tables S1 and S2 in Supplementary Information (See Supplementary Section 7). By controlling the optical thickness nd of the LT and HT layers, we achieve an infrared reflectivity higher than that of Al at a targeted wavelength λ 0 . For example, when both n 1 d 1 and n 2 d 2 are equal to 1900/4 nm, the reflectivity of the multilayers is 99.0% at 1900 nm, exceeding that of the pure Al film (96.5% at 1900 nm) (Figure 3c). When both n 1 d 1 and n 2 d 2 are equal to 4100/4 nm, the reflectivity is 97.0% at 4100 nm, again higher than that of the pure Al film (96.8% at 4100 nm) (Figure 3d).
As an indication of durability, we characterize the hardness and the corrosion behavior of the multilayered film (see Supplementary Section 8). Because Al is the most commonly used infraredreflective coating material 8,9 , we conduct the same tests on an Al film as a comparison. The hardness of the multilayered film is 13.8 GPa, 27 times beyond that of the Al film (0.5 GPa) (Figure 3e). In a 0.5 mol l − 1 H 2 SO 4 solution, the corrosion current density of the multilayered film is 2.94 × 10 − 6 A cm − 2 , three orders of magnitude less than that of the Al film (4.69 × 10 − 3 A cm − 2 ) (Supplementary Figure S4). In other words, the acid corrosion resistance of the Highly durable infrared-reflective coatings C Hu et al multilayered film is 41500 times better than that of Al. In a salt solution (3.5 wt.% NaCl), the corrosion current density of the multilayered film is 6.76 × 10 − 6 A cm − 2 , in contrast to that of the Al film of 5.26 × 10 − 2 A cm − 2 (Figure 3e and Supplementary Figure S5). This fact means the multilayered film is 7500 times more corrosion resistant than Al in this salt solution. Additionally, during salt-bath experiments (see Supplementary Section 9), the Al film shows significant corrosion after immersion in a NaCl solution at 35°C for 5 min. When the duration time gradually increases to 180 min, the corrosion pits continue to increase (Supplementary Fig. S6). However, the multilayered film does not show any corrosion traces in the whole 10 days (14 400 min) of the salt-bath experiment ( Supplementary Fig. S6), indicating that the corrosion resistance of the multilayered film in a seawater environment is easily 3000 times that of Al. These results demonstrate that the HfN x -based multilayer film possesses a much higher scratch and corrosion resistances than pure Al. In conclusion, by exploiting the tunable properties between reflectivity and transmission induced by electron localization, we successfully achieve HfN x -only multilayer films with both high infrared reflectivity and high durability. Compared with the conventional multilayer stacking of a variety of materials, our approach involves only one material (HfN x ) with switching of the optical states achieved by changing the stoichiometry x (i.e., via the partial pressure of nitrogen only), thus greatly simplifying the manufacturing process. Additionally, as seen in Figures 3d and 4a, the reflectance (97%) of the multilayered HfN film at 4100 nm is higher than that of the Ag-doped HfN film (94%) at the same wavelength. This means that a multilayered HfN film has better reflective properties than an Ag-doped HfN film at a target wavelength or in a narrow band, which is very useful for many important optical applications (e.g., pointing and tracking optical systems 3 ).
Achieving highly durable infrared-reflective coatings over a range of wavelengths through silver doping. To verify the reflectivity boost over a range of wavelengths, we dope Ag into near-stoichiometric HfN x film using co-sputtering. For simplicity, we refer to the Agdoped HfN film as 'the doped film'. We investigate the effect of the Ag content (C Ag ) on the structure and reflectivity. Our reflectivity measurements show that the average reflectance of the Ag-doped films in the range of 3-12 μm increases from 77 to 95% as the Ag content increases from 0 to 3.1%. However, when the Ag content further increases to 3.9%, the reflectance reduces to 80% due to an increase in the surface roughness. Additionally, it is found that the hardness of the films increases from 22.6 GPa to 25.4 GPa to 32.4 GPa as the Ag content increases from 0% to 3.1% to 3.9%. However, the hardness decreases significantly when more Ag is incorporated. Considering that the film with the Ag content of 3.1% has the highest infrared reflectivity and a relatively high hardness, we believe that the Ag doping of 3.1% is the best. Figure 4a shows the infrared reflectivity spectra of pure HfN (C Hf = 50.6%, C N = 49.4%) and the doped film (C Ag = 3.1%, C Hf = 50.2%, C N = 46.7%) together with that of an Al film as reference. The reflectivity of a pure HfN film is only 77%, Step 4 Step 3 Step 2 In fr ar ed lig ht Highly durable infrared-reflective coatings C Hu et al while that of the doped film increases abruptly to 95%, close to that of an Al film (97%) over a spectrum of 3 to 12 μm.
To explore the cause of the reflectivity enhancement, we characterize the structure of the samples. Figure 4c shows the grazing incident X-ray diffraction pattern for the pure and doped films, wherein both samples contain diffraction patterns attributed to the rock salt phase, showing that the pure and doped films have the same rock salt structure. Figure 4d shows lattice images from high-resolution transmission electron microscopy of the doped film, in which wellcrystallized nanograins are uniformly distributed on the film surface. does not have an obvious change. This is consistent with the measured resistivity of the pure (5.4 Ω) and doped films (5.3 Ω). Additional electron states among − 3.5 eV and − 2 eV are introduced by the doping (Figure 5c). These extra electrons induce the charge polarization near N atoms (Figure 5b). From the band structure (Figure 5a), these extra states have a good dispersion due to the hybridization with the nearby N atoms. They have an important contribution from the free electrons to the energy of the plasma. The visible-near-infrared reflectivity spectra (Figure 5d) shows that the plasma energy of a HfN film increases significantly from 2.6 to 3.0 eV after the introduction of Ag, which agrees well with the calculation above. Previous studies 66 have shown that the reflectivity of HfN films in the visible-infrared band depends on both the intraband transition that is related to the free electrons (described by a Drude part) and the interband transition that is related to the bound electrons (described by a Lorentz part). The reflectivity in the infrared region depends mainly on the intraband transition and is closely related to the plasma energy. The Ag-dopinginduced increase in the infrared reflectivity is attributed to an increase in the plasma energy and the blueshift of the reflective cutoff wavelength ( Figure 5), which expands the high-reflectivity region toward lower wavelengths. Highly durable infrared-reflective coatings C Hu et al In addition to the significant improvement of reflectivity, we find that the introduction of Ag increases the durability of the HfN film. In a 0.5 mol l − 1 H 2 SO 4 electrolyte solution, the corrosion current density of a pure HfN film is 1.81 × 10 − 5 A cm − 2 , whereas that of the doped film decreases to 1.14 × 10 − 5 A cm − 2 , only 2.4‰ of that of the pure Al film (4.69 × 10 − 3 A cm − 2 ) ( Supplementary Fig. S8 in Supplementary Section 11). In a 3.5 wt.% NaCl electrolyte solution, the corrosion current density of the pure HfN film is 5.83 × 10 − 6 A cm − 2 , while that of the doped film decreases to 1.90 × 10 − 6 A cm − 2 or only 0.04‰ of that of the pure Al film (5.26 × 10 − 2 A cm − 2 ) ( Supplementary Fig. S9 in Supplementary Section 11). Additionally, the introduction of Ag causes an increase in the hardness of the HfN film from 22.6 to 25.4 GPa, or 450 times the hardness of an Al film (0.5 GPa) due to the solid-solution strengthening effect (Figure 4b). Additionally, during salt-bath experiments (see Supplementary Section 12), the doped film does not reveal any corrosion characteristics within 10 days (14 400 min), whereas the Al film shows significant corrosion pits in only 5 min ( Supplementary Fig. S10), demonstrating that the doped film is much more corrosion resistant than Al in seawater. These results suggest that the Ag-doped HfN film not only has a high infrared reflectivity similar to Al but also has a much higher durability than Al. Therefore, it can be used as a highly effective durable infraredreflective coating.
The traditional method of improving the durability of an Al film is to deposit protective layers of oxides on its top.  Table 1) are consistent with the previous experimental results 9,67 , demonstrating the reliability of our experiments. We compare the reflectivity, stability and cost of traditional mirror coating (Al/SiO 2 film) with our mirror coating (Ag-doped HfN film). The results are as follows. (1) The average reflectivities of an Ag-doped HfN film and Al/SiO 2 films are 95 and 96% at 3-12 μm, respectively ( Figure 6, Table 1). This fact indicates that their infrared reflectivities are very similar. (2) According to Table 1, the acid corrosion resistance, salt corrosion resistance, wear resistance and hardness of an Ag-doped HfN film are 25 times, 53 times, 10 times and 10 times those of an Al/SiO 2 film, respectively. Additionally, the reflectivity of an Al/SiO 2 film decreases by 55% after the 10-day salt water bath test, while the reflectivity of an Ag-doped HfN film decreases by only 3%.  costs approximately 150 dollars. Therefore, taking the cost of the raw materials into account, the Ag-doped HfN film is more expensive than the Al/SiO 2 film. However, the Ag-doped HfN film only needs to deposit a one-layer film, which is easier than the two-step preparation of the two-layer Al/SiO 2 film. Moreover, the Ag-doped HfN films have much higher durabilities and stabilities. Thus we believe that the cost of the two mirror coatings is very similar. These results suggest that the Al/SiO 2 films have slightly better reflectivities and a lower cost than the Ag-doped HfN films, but their durabilities and stabilities of the reflectivity are far lower than those of Ag-doped HfN films. Thus the traditional mirror coating is more suitable for moderate environments, while our mirror coating obtained by the new strategy is more suitable for harsh environments, such as those with dust, windblown sand, moisture, acid rain and salt fog.

CONCLUSIONS
We propose an alternative design approach of 'first durability and then reflectivity' to achieve the goal of both high durability and reflectivity.  We recommend transition metal compounds, TMR (TM = Ti, Zr, Hf, V, Nb, Ta, Cr, Mo, W; R = B, C, N), as the durable base material and then improve its reflectivity to achieve the integration of high durability and reflectivity. The two validation experiments on HfN prove that our new strategy and approaches work as expected. The coatings thus obtained not only have an infrared reflectivity performance close to that of aluminum but also have a far better durability. The traditional mirror coating (e.g., Al/SiO 2 films) is more suitable for moderate environments, while our mirror coating that is obtained by the new strategy (e.g., Ag-doped HfN film) is more suitable for harsh environments, such as those with dust, windblown sand, moisture, acid rain and salt fog. This research opens up whole new areas in developing durable infrared-reflective coatings. HfN is only 1 of the 27 TMR transition metal compounds; thus this study can be considered one point in the whole spectrum of possibilities, that is, zero dimension. By substituting any other transition metal for Hf and/or replacing N with B and C, a two-dimensional study can be mapped out. Furthermore, the inherent properties of each TMR can be improved by modulating its structure (multilayering, alloying, nanocompositing, etc.); thus new three-dimensional studies may emerge in this area. As such, new materials, new functions and new mechanisms are expected to be developed. This study has ushered the transition metal compounds from hard protective coatings (e.g., cutting-tool coatings) into a completely new area of optical coatings. This rejuvenates the study of the transition metal compounds.