Vanadium oxides have attracted much attention over the past few decades because of their thermochromic properties.^{[1– 4]} VO₂ exhibits an abrupt semiconductor metal transition at about 68  ° C, accompanied by sharp drop in resistivity and IR transmittance.^{[5, 6]} The optical transmission and reflection spectra of V₂O₅ can be changed by heating or laser irradiation due to the thermochromism, which shows potential application in the field of information storage or displays.^[7] The physical and chemical parameters of vanadium oxides can be modified by doping with other elements. For example, doping VO₂ thin films with transition metals is known to largely affect the properties of the metal– insulator transition (MIT).^{[8, 9]}

Thin films doped with vanadium oxides can be prepared by various methods, including reactive sputtering, ^[10] sol-gel method, ^[11] chemical vapor deposition, ^[12] ion implantation, ^[13] pulsed laser deposition, ^[14] etc. Among these methods, reactive sputtering has gained significant importance because it enables control of the structure, composition, and properties of VO_x films by adjusting the deposition conditions. This process is very nonlinear and usually exhibits a hysteresis behavior with respect to the reactive gas flow. In order to understand this phenomena, Berg et al.^{[15, 16]} proposed a qualitative model to describe the reactive sputtering deposition at steady state. Recently, the extensions to this model have been published, ^{[17– 22]} such as deposition with two gases.^{[21, 22]}

However, to our knowledge, there are no studies on reactive sputtering of multi-component thin films such as VO_x using a doping gas. Studies of nitrogen-doped vanadium oxide films are attracting more attention since it potentially eliminates the hysteresis effect and improves the film properties.

In this paper, the effect of the reactive gas mixture on the hysteresis effect during the reactive sputtering deposition of N-doped VO_x thin films is investigated both from experimental data and numerical modeling. Based on the modeling results, the effect of doped nitrogen on the compound fractions at the collecting area is also discussed. The overall composition of N-doped VO_x is generally a mixture of V, V₂O₃, VO₂, V₂O₅, and VN. During simulations, the film structure is assumed to be composed of a mixture of V, VN, V₂O₅, and suboxides such as V₂O₃ and VO₂. Some other phenomena, such as replacement reactions caused by oxidation of the nitride layers and reduction of compounds caused by sputtering, are also included into the model.

The mathematical model describing the sputtering system is shown in Fig.  1. The total input rate of the reactive gas (oxygen) is denoted as Q_tot. The target surface can then be divided into five parts (see Fig.  1). The fractions of V, V₂O₃, VO₂, V₂O₅, and VN at the target will be called θ _t1, θ _t2, θ _t3, θ _t4, and θ _tN, respectively. θ _s1, θ _s2, θ _s3, θ _s4, and θ _sN are fractions of collecting area covered by V, V₂O₃, VO₂, V₂O₅, and VN, respectively.

It is assumed that the gettering of reactive gas (O₂ and N₂ onto metal contributes to the formation of suboxide V₂O₃ or VN on the target, depending on the probability of these two reactions, which will increase θ _t2 or θ _tN, respectively. Further attachment (by oxygen) onto this suboxide V₂O₃ contributes to form the VO₂, causing an increase in θ _t3 but a decrease in θ _t2. The superoxide V₂O₅ is formed by the reaction of the oxygen and VO₂. For the metal vanadium atoms sputtered from the target, if the atom comes from the fraction of θ _t1 and deposits on the substrate surface which previously consisted of VO_x (V₂O₃, VO₂, V₂O₅, and VN), then the fraction of θ _s1 will increase and the fraction of (1-θ _s1) will decrease. Furthermore, when the VO_x particle sputtered from target deposit on the substrate which previously consisted of VO_i (i! = x), then the corresponding fraction of θ _si will decrease and the fraction of VO_x will increase. However, the fraction of flux (J/q)Y_tiθ _tiA_t deposited onto the fraction θ _si of the collecting area A_s will not change the value of θ _s1. Adding elemental non-reacted target atoms to an area that already consists of non-reacted atoms also does not change θ _s1. As we have mentioned, it is assumed that oxygen can convert nitride to oxide, with a certain probability. Therefore, β ₁, β ₂, and β ₃ are the replacement coefficient for replacement of compound VN by compound of V₂O₃, VO₂, and V₂O₅, respectively.

The following simulation parameters are used in this work: α _t1 = 0.8, α _t2 = 0.8, α _t3 = 0.8, α _t4 = 0.8, α _tN = 0.6, α _s1 = 1, α _s2 = 1, α _s3 = 1, α _s4 = 1, α _sN = 0.8. Y_t1 = 0.6, Y_t2 = 0.3, Y_t3 = 0.2, Y_t4 = 0.2, Y_tN = 0.2. γ ₁ = 1, γ ₂ = 0.4, γ ₃ = 0.4, γ ₄ = 0.4, γ _N = 0.4, γ = 1. l₁ = 0, l₂ = 3, l₃ = 4, l₄ = 5, l₅ = 2. β ₁ = 0.08, β ₂ = 0.15, β ₃ = 0.08, U₀ = 300  V. Target area A_t = 260  cm², gas temperature T = 300  K, pumping speed S = 0.2  m³/s.

Fig.  1.

	Figure Option ViewDownloadNew Window
	Fig.  1. Schematic of the conditions at target and collecting area during reactive sputtering with two reactive gases.

By making the assumptions above, in combination with the simplifications presented earlier, the balance equations for the conditions on the target should be rewritten as

Here J is the ion-current density; q is the elementary electronic charge; Y_t2, Y_t3, and Y_t3 are the sputtering yield of the V₂O₃, VO₂, and V₂O₅, respectively; α _t1 is the probability (sticking coefficient) for an oxygen molecule to react with a vanadium atom that comes from the θ _t1 fraction of the target; α _tN is the sticking coefficient for nitrogen; A_t is the target area; l₂ and l₅ are the stoichiometry of the V₂O₃ and VN, respectively; and F₁ and F₂ represent oxygen and nitrogen molecules/(unit area and time), respectively.

Equation  (1) determines the balance of the vanadium atom fraction (θ _t1) on the target. For simplicity, the sputtered material from the surface fraction (J/q)Y_t2θ _t2A_t + (J/q)Y_t3θ _t3A_t + (J/q)Y_t4θ _t4A_t is assumed to be sputtered as molecules, irrespective of whether the material is ejected in atomic or molecular form. The only way that sputtered compound molecules may be replaced is by reactions between neutral reactive gas molecules (O₂, N₂ and the elemental non-reacted vanadium atoms,

Moreover, equation  (2) determines the balance of the fraction of θ _t2 on the target. Here, l₃ is the stoichiometry of the VO₂. The reaction between elemental target atoms and the reactive gas atoms (which occurs on the compound fraction θ _t1) will contribute to an increase in the value of θ _t2. Besides, the replacement of nitride by oxygen, which converts compound VN into compound V₂O₃, will also increase the fraction of V₂O₃. On the other hand, the compound V₂O₃ is sputtered from the surface fraction θ _t2, decreasing the fraction of θ _t2. Furthermore, attaching the oxygen to V₂O₃ will contribute to a decrease in θ _t2 as well.

Similar to Eqs.  (1) and (2), equations  (3)– (5) determine the balance of the fraction of θ _t3, θ _t4, and θ _tN, respectively. Here, l₄ is the stoichiometry of the V₂O₅. Note that all the compound fraction sums to unity, namely,

The expressions for θ _t1, θ _t2, θ _t3, θ _t4, and θ _tN can be obtained by solving Eqs.  (2)– (6).

At a steady state, the contributions supporting an increase in θ _si must be identical to the contributions that support a decrease in θ _si. This leads to the following balance equations for the collecting area A_s:

where α _s1, α _s2, and α _s3 are the probability (sticking coefficients) for the oxygen molecule to react with V, V₂O₃, and VO₂ on the substrate, respectively.

Equation  (7) determines the balance of the fraction θ _s2 on the substrate. The fraction of flux (J/q)Y_t2θ _t2A_t that is deposited onto the fraction (1-θ _s2) will increase the value of θ _s2. Obviously, the reactions between O₂ and vanadium atoms at the substrate surface will increase θ _s2. Moreover, the replacement of nitride by oxygen 2β ₁F₁A_sθ _sN / l₂, which converts compound VN into compound V₂O₃, will also increase the fraction of V₂O₃. The fraction of flux (J/q)(Y_t1θ _t1 + Y_t3θ _t3 + Y_t4θ _t4 + Y_tNθ _tN deposited onto the fraction (θ _s2) will contribute to a decrease in θ _s2. The number of molecules which convert into VO₂ by oxygen is denoted by 2α _s2F₁A_sθ _s2 /(l₃ − l₂ ). The reaction between elemental non-reacted atoms and oxygen will also decrease the value of θ _s2.

Likewise, the steady-state equation for the fraction of VO₂, V₂O₅, and VN on the substrate may therefore be defined as

Finally, the sum of fractions (θ _s1, θ _s2, θ _s3, θ _s4, θ _sN) is equal to 1, like that in the case of the target surface,

Based on the fact that target voltage U depends both on the secondary emission coefficient γ and the degree of target poisoning. Pflug et al.^[23] suggested the following expression:

in which γ ₁, γ ₂, γ ₃, γ ₄, and γ _N are the secondary electron emission coefficients, respectively, at the V, V₂O₃, VO₂, V₂O₅, and VN on the target surface, and U₀ is a reference voltage required for keeping the equation dimensionless. Thus, by this equation, U is calculated as functions of θ _t1, θ _t2, … , θ _tN.

Compound formation by reaction between O₂ and vanadium atoms, compound V₂O₃, and compound VO₂ at the target will consume reactive gas molecules Q_t1, Q_t2, and Q_t3, reactively. Thus, the consumption (number of oxygen molecules per unit time) at the target Q_t can be obtained from

While the consumption Q_s at the collecting area A_s can be expressed by

Here, Q_s1 is reactive gas molecules consumed by oxygen molecules and vanadium atoms at collecting area. Q_s2 is the oxygen molecules consumed by V₂O₃ compound. The other part Q_s3 is the oxygen molecules consumed by the VO₂ compound.

The remaining part Q_p of the oxygen will escape from the processing chamber through the pumping system,

Here, P_O2 is the partial pressure of oxygen, N_a is the Avogadro constant, R_r is the general gas constant, T is the temperature, and S is the pumping speed.

The total flow rate of oxygen is denoted as Q_tot-O2

Similar to the oxygen gettering process, the flow rate of nitrogen is determined by

Here, P_N2 is the partial pressure of nitrogen. The expressions for oxygen flow (Q_{tot– O2}) and nitrogen flow (Q_{tot– N2}) versus target voltage (U) can be obtained by numerical calculation based on Eqs.  (1)– (17).

N-doped VO_x thin films were deposited by DC reactive magnetron sputtering in a custom-built vacuum system. A water-cooled rectangle planar vanadium metal target, with a purity level of 99.99% , was used in these experiments. The deposition chamber was evacuated to a base pressure of about 1.0× 10^-4 Pa by a combination of turbomolecular and rotary vane pumps. Argon, oxygen, and nitrogen flow were controlled by using independent mass flow controllers, and the argon flow was kept constant at 100  sccm. The oxygen flow and nitrogen flow were varied from 1.0 to 6.0  sccm and 0 to 4.0  sccm, respectively. In order to ensure reproducible results of the reactive sputtering between runs, the vanadium target was first pre-sputtered by argon plasma for 15 min, and then oxygen injection was kept on for 15 min until equilibrium is reached in the argon/oxygen mixture.

The x-ray photoelectron spectroscopy (XPS) spectra were taken by using XSAM800 system (Kratos Ltd., Al Kα x-ray source without monochromatization, the base pressure of the ultrahigh vacuum chamber is lower than 3× 10^-7 Pa, binding energies (BE) are determined by using C1s peak at 284.8  eV as a reference, the resolution is 0.9  eV/10⁴ counts per second).

The numerical and experimental results for the target voltage as functions of oxygen flow rate at different N₂ flow rates are shown in Figs.  2(a) and 2(b). It can be seen that there is a pronounced difference in hysteresis width for the curves without N₂. The appearance of the hysteresis loop is caused by the large difference between the consumption of reactive gas during the increase and decrease of the O₂ flow rate. The increase of the reactive gas flow without N₂ in the gas mixture induces the formation of V₂O₅ on the target and collecting area. Because the sputtering yield of V₂O₅ is much lower than that of pure V, less metal will be deposited on the collecting area, and the chemisorption rate will decrease abruptly. Consequently, higher cathode voltage will be required to maintain the discharge and cause hysteresis. In addition, data from the target voltage as a function of the O₂ flow rate indicate the vanishing hysteresis after the addition of N₂ in the gas discharge. The experimental results in Fig.  2(b) confirm the numerical finding. After the addition of N₂ in the gas mixture, both gases will start to compete for compound formation. The sputtering yield of the nitride layers is much lower than that of pure V, but is still greater than V₂O₅. Therefore, it is expected that the transition to the oxide mode becomes smoother after the addition of N₂, because the nitrogen atoms contribute to avoiding instabilities in the transition region.

All of the compounds employed in the numerical model were confirmed by XPS. The XPS spectrum of film surface

Fig.  2.

	Figure Option ViewDownloadNew Window
	Fig.  2. Numerical results (a) and experimental curves (b) of the target voltage as a function of the O2 flow rate at N2 flow of 0, 2, and 4  sccm.

was fitted with mixed Lorentzian– Gaussian profiles after subtraction of a Shirley function using software XPSpeak4.1. The N_1s core level for film deposited at O₂ flow rate of 3.0  sccm, 3.2  sccm, 3.4  sccm, 3.8  sccm, and 5.0  sccm are shown in Figs.  3(a)– 3(e), respectively. According to the results, films deposited at low O₂ flow rate are mainly composed by V– N, N– C, and N– O. Increasing the O₂ flow rate, the fraction of VN is relatively reduced and the fraction of NO is gradually increased. At O₂ flow rate of 3.4  sccm, O atoms diffuse in the VN lattice and replace N atoms, therefore removing the compound of VN. Consequently, the fraction of VN drops to zero. At high oxygen flow, the fraction of NO continues to increase and the film becomes mostly covered by N– O (see Figs.  3(c)– 3(e)).

The V2p_3/2 core level, depicted in Fig.  4, was decomposed in four characteristic contributions: V⁰⁺ , V³⁺ , V⁴⁺ , and V⁵⁺ . These peaks are, respectively, assigned to V, V₂O₃, VO₂, and V₂O₅.

When the oxygen flow is fairly low (spectrum (a)), all of the four-species states are found. As the O₂ flow rate

Fig.  3.

	Figure Option ViewDownloadNew Window
	Fig.  3. XPS spectra for N1s core levels for film deposited at O2 flow rates of (a) 3.0  sccm, (b) 3.2  sccm, (c) 3.4  sccm, (d) 3.8  sccm, and (e) 5.0  sccm.

increases, the film is oxidized due to oxidation of VN and chemisorption of oxygen particles. With the increase in the oxygen flow (spectrum (b)), V⁰ gradually disappears due to the increased oxygen concentration. When oxygen flow is increased to around 4.2  sccm, the intensity of V³⁺ drops to negligible level. At very high oxygen flow (spectrum (d)), only V⁵⁺ is detected. From the XPS measurement analysis results shown in Figs.  3 and 4, it is also noted that the fraction of VN is quite lower than that of VO_x, indicating that the incorporation of N element into the growing film is very small. This phenomenon will also be seen in Fig.  5(b).

Fig.  4.

	Figure Option ViewDownloadNew Window
	Fig.  4. V 2p3/2 lines of vanadium oxides film taken at different samples with different oxygen flow. Oxygen flow is (a) 2  sccm, (b) 3.8  sccm, (c) 4.2  sccm, and (d) 6  sccm.

The film stoichiometry obtained from modeling and experiment is shown in Fig.  5(a). The nitrogen flow is kept constant at 4  sccm. The fraction of θ _s4 increases with the increase of oxygen flow. In the region of oxygen flow up to 3.5  sccm, the fraction of θ _s4 is higher than θ _s3, and θ _s1 almost reaches zero. Furthermore, with the continued increasing amount of oxygen, θ _s4 increases asymptotically to one. Meanwhile, the fraction of metal V, V₂O₃, and VO₂ is reduced and eventually disappears. There are several reasons for the shape of these curves. For a zero flow of oxygen the target is to some extent oxidized and the rest is metallic, while at higher flows of oxygen the target is essentially oxidized. After the 3.5  sccm point, the amount of O₂ consumed to VO₂ surpasses the amount of O₂ gettered by deposited V and V₂O₃, resulting in a drastic increase in the fraction of θ _s4. It should also be noted that the fraction of VO₂ is generally higher than that of V₂O₃, which means that the films contain significantly more VO₂ than V₂O₃. This is a consequence of the higher reactivity of V₂O₃ than V with oxygen.

Another interesting feature observed from the modeling result is that θ _s1 and θ _s3, corresponding to the fraction of metal vanadium and V₂O₃, decrease dramatically in the region from 3.0 to 4.0  sccm. However, the fraction of VO₂ climbs down much slowly than V and V₂O₃ in the region from 3.0 to 4.0  sccm. This effect could be explained by the replacement reaction of VN by O₂ atoms because of higher reactivity of O₂ than N₂, which decreases the fraction of VN and increases fraction of VO₂.

These results show that the trends in the experiments are adequately described by the simulations. It is found that the error between experimental and modeling results is less than 15% over the whole figure, and is less than 8% for most values.

In order to study the effect of additional N₂ on the film composition, a direct comparison between the experimental and modeling results of VN and VO_x fractions as a function of oxygen flow rate is shown in Fig.  5(b). There appears to be a reasonable agreement between the numerical and experimental data. It is found that the error is less than 10% . Increasing the O₂ flow rate, we observe a decrease in the VN concentration and an increase in the VO_x content by XPS. According to Ref.  [24], if the film stoichiometry is governed by oxidation processes, then a complete oxidation of the film surface is expected at high O₂ flow rates once the formation of oxides is energetically more favorable than nitrides.

Fig.  5.

	Figure Option ViewDownloadNew Window
	Fig.  5. (a) Experimental numerical results of the elemental compositions (V, V2O3, VO2, and V2O5) versus oxygen flow. (b) Experimental numerical results of VN versus VOx concentration comparison.

In this paper the reactive sputtering deposition of N-doped VO_x has been studied both numerically and experimentally at different reactive gas mixtures. A kinetic model of reactive sputtering is proposed to modify Berg's model by taking into account the five species V, V₂O₃, VO₂, V₂O₅, and VN in the N-doped vanadium oxides film. The fraction of compound on the substrate surface has been investigated as a function of reactive gas flow. From simulation results, substrate surface coverage is estimated.

Vanadium oxide films with nitrogen as an additional reactive gas have been deposited by reactive magnetron sputtering using different oxygen flow rates. Film compositions obtained from the XPS analysis are compared to the composition obtained from simulations and we reach a satisfactory agreement. Results show that the compound of VN is only found at very low O concentration because the replacement reaction of VN by O₂ atoms owns higher oxygen flow rates and the incorporation of nitrogen into the growing film is quite small.

We also find that the smoother transition of the metallic mode to the oxide mode and the vanishing hysteresis after the addition of N₂ are mainly attributed to the reduction of the sputtering yield and the decrease of the reactive gas consumption in the metallic mode.