Tuning the conductivity of vanadium dioxide films on silicon by swift heavy ion irradiation

Ion tracks are created when heavy ions of sufficiently high kinetic energy, usually above 1 MeV per nucleon, pass through solids. The large energy deposition along the ion trajectory causes electronic excitation and ionization processes followed by a rapid temperature spike caused by electron-phonon interaction within a localized cylindrical volume along the ion trajectory.^1,2 Depending on the material properties, this temperature increase may cause local melting followed by rapid cooling on picosecond time scales. In many solids, in particular in insulators, these processes create a cylindrical damage zone along the ion path consisting of defect-rich, amorphous or otherwise modified material.^3,4 Amorphous track formation was observed for a number of insulator materials such as α-quartz,⁵ point and extended damage regions along the ion tracks are seen e.g. for irradiated compound semiconductors,⁶ α-Al₂O_3,⁷ doped vanadium sesquioxide (V_1-xCr_x)₂O₃⁸ and magnetite.⁹ 

It was shown by Kokabi et al. that swift heavy ion irradiation of Cr-doped vanadium sesquioxide [(V_1-xCr_x)₂O₃] influences the metal–semiconductor transition temperature and changes the electrical properties of both the metallic and semiconducting phase.^8,10 (V_1-xCr_x)₂O₃ exhibits an intrinsic low temperature metal-semiconductor transition at about 170 K as well as a second high temperature transition between 170K and 470K depending on Cr dopant concentration. In these studies, (V_1-xCr_x)₂O₃ bulk samples were irradiated with 6 GeV ²⁰⁸Pb ions at fluences between 4·10¹¹ and 5·10¹² ions/cm². The corresponding electronic energy loss along the ion path is about 37 keV/nm. TEM revealed discontinuous tracks with damage regions. From XRD relative peak intensities an average track diameter of about 2.7 nm was calculated.⁸ After swift heavy ion irradiation a shift of the 170 K transition towards lower temperatures, and for lower fluences (non-overlapping tracks) a decreasing resistivity for both the semiconducting and the metallic state occurs. At higher fluences the resistivity of the metallic state starts to increase. The high temperature transition gradually disappears with increasing ion fluence.¹⁰ The observations were explained by stresses exerted by the ion tracks on the neighboring pristine material, assuming that the ion tracks are damage regions of lower density. In the case of the low temperature transition, the stresses stabilize the metallic phase, which is also stable under external pressure.

It was also shown that stresses, generated by damage regions of swift heavy ion tracks, have a strong influence on the surrounding materials properties. For example, magneto-elastic effects and changes in the magnetic properties were observed for swift heavy ion irradiated magnetite⁹ and hexagonal ferrites, like barium hexaferrites.¹¹ After irradiation with either 3.8 GeV ¹²⁹Xe or 6 GeV ²⁰⁸Pb ions, corresponding to electronic energy loss of about 20 and 35 keV/nm, respectively, clear magneto-elastic effects are observed. The irradiation of barium hexaferrite with fluences between 1·10¹¹ and 7·10¹² ions/cm² caused a magnetic texture with magnetization oriented parallel to the tracks. This anisotropy was attributed to the stresses induced in the “halos” of the ion tracks.¹¹ In the case of magnetite a decrease in saturation magnetisation and a preferential orientation of the hyperfine magnetic field was observed for fluences < 1·10¹² ions/cm², i.e. no significant track overlap.⁹ Again, this behaviour was attributed to extended defects along the discontinuous latent ion tracks, generating stress in the surrounding regions. It should be noted, that in these studies of Kobaki et al.,^8,10 Costantini et al.¹¹ and Meillon et al.⁹ rather high fluences were used, corresponding to 10-100 % coverage of the sample surface area with a significant fraction of overlapping ion tracks. In our present studies, the ion fluences are much lower and the total track area takes up at most 0.3 % of the sample surface area, i.e. we investigate the effect of well separated individual ion tracks.

A unique material is diamond-like carbon, in particular tetrahedrally bonded amorphous carbon (ta-C) with high sp³ bond fraction of about 80%, because the as-grown material has a high resistivity and can be locally converted into graphitic sp²-bonded conducting ion tracks of typically 8 nm in diameter upon swift heavy ion irradiation.^12,13 Each ion track produces a 1-3 nm high hillock on the surface which can be easily identified using atomic force microscopy. The current contrast between on-track and off-track regions measured using conductive atomic force microscopy is about 10⁴-10⁵.¹⁴ The track formation in carbon materials was simulated using molecular dynamics,¹⁵ which is in good agreement with experimental findings on the track diameter and the hillock size on the surface. The track conductivity strongly depends on the initial sp³ content of the carbon films and on the ion species, i.e. the electronic energy loss.¹⁶ The onset of conducting track formation occurs around 20-30 keV/nm and the highest track conductivities are obtained for maximum achievable electronic energy loss of about 40 keV/nm for 1GeV ²³⁸U ions or 72 keV/nm for 30 MeV C₆₀ cluster ions.¹⁷ Until now, beside ta-C conductive ion track formation was only observed in fullerene films.¹⁸ For non-overlapping ion tracks in polymers only a very low conductivity was found.^19,20

After irradiation of crystalline CuIr₂S₄, a sulphospinel which exhibits a temperature-induced metal-insulator transition at 235 K,^21,22 with MeV H⁺ and He ions below T_C a slight metastable increase in conductivity was observed and was interpreted as the formation of conductive nanowires produced by the impinging ions.²³ However, a conductivity change only occurred for ion fluences exceeding 10¹³ ions/cm², which points to a macroscopic effect rather than individual conducting filaments. Irradiation of the insulating phase of CuIr₂S₄ at low temperatures of 8.5 K with X-rays also causes a metastable increase of the conductivity accompanied by a transformation to a tetragonal crystal structure.²⁴ 

As mentioned before, conductivity changes and changes of the metal-semiconductor transition temperature were observed for swift heavy ion irradiated transition metal oxide Cr-doped V₂O₃ and were interpreted in terms of ion-induced stresses around the ion tracks, stabilizing the conductive phase of the material.^8,10

Transition metal oxides showing correlated electron behavior, such as temperature-driven reversible metal-insulator transition (MIT) and colossal magneto-resistance (CMR)²⁵ are of interest for novel all-oxide electronic devices.^26–28 These materials can also be integrated with other oxide materials with exceptional properties, e.g. ferroelectrics and high-k dielectrics.²⁹ Whereas manganites, magnetite (Fe₃O₄)^30,31 and the CuIr₂S₄ have MIT transition temperatures below room temperature, vanadium dioxide (VO₂) has a transition temperature of about 67 °C^32–35 and the rare earth nickelate SmNiO₃ of about 130 °C.³⁶ The latter materials possess a high potential for future applications in logic or memory devices and possibly integration into CMOS technology.^27–39 

An interesting MIT system as candidate to generate conducting ion tracks or generally to produce ion-induced conductivity changes is VO₂. VO₂ exhibits a reversible temperature driven MIT, which markedly changes its electronic and optical properties.^33,40 Below the transition temperature VO₂ behaves as an insulator or semiconductor with a monoclinic crystal structure and a band gap of about 0.7 eV, whereas for temperatures higher than 67 °C, it transforms to a metallic state with a tetragonal rutile structure.⁴¹ The electrical resistivity decreases by 3 to 5 orders of magnitude depending on the crystalline quality of the deposited films,⁴² stoichiometry and doping,⁴³ while the optical reflectivity markedly increases.^44,45 In VO2 thin films, this transition can be triggered by thermal,^33,40,46 electrical (charge injection, bias or Joule heating)^40,46–48 or optical excitation (photon excitation),^44,45 and even by external pressure or strain.^49,50

The MIT temperature of VO₂ can be reduced by ion implantation of oxygen, producing implantation defects which are stable up to at least 120 °C annealing temperature.⁵¹ Also ion implantation of W ions and subsequent annealing lead to a significant reduction of the MIT temperature down to about 30 °C.⁵² It is also possible to tune the hysteresis of the MIT transition. For VO₂ precipitates formed by ion implantation into SiO₂ and subsequent annealing, a hysteresis width of more than 30 K was seen in optical transmission, correlated to a structural transformation seen with X-ray diffraction (XRD).⁵³ 

This motivated us to investigate vanadium dioxide regarding its electrical behavior after swift heavy ion irradiation with high electronic energy loss along the ion path. The extremely rapid annealing and quenching during thermal spikes may freeze-in a conductive high temperature phase, may lead to defect generation, slightly modify the stoichiometry and bond structure within the ion track or may induce mechanical stresses. All these processes could result in a metastable or persistent increase of the conductivity along the ion track.

In this work we present experimental evidence that swift heavy ion irradiation causes a persistent and dramatic increase of the conductivity of vanadium dioxide films on silicon substrates.

Vanadium dioxide films with about 100 nm in thickness were grown on heavily-doped n-type Si (0.002-0.005 Ωcm) by rf magnetron sputtering with a V₂O₅ target at 550°C. During synthesis, the gas pressure and sputtering gun power were maintained at 10 mTorr and 120 W respectively. The process gas was a 99.22 SCCM argon and 0.78 SCCM oxygen gas mixture. The films are polycrystalline and have a typical rms roughness of 10 nm as determined by atomic force microscopy (AFM) (Figure 1). Also shown in figure 1 is the current mapping which represents a good conducting region of the sample. For other sample regions the conductivity can be lower and even strongly varies from grain to grain. The polycrystalline surface structure will complicate the analysis of possible ion track formation using conductive AFM and limits the possibility to detect ion tracks essentially to the central parts of the crystallite regions. During film synthesis at elevated substrate temperatures a few nm thick silica interfacial layer with high resistivity is formed,²⁹ which may influence the I-V measurements between a top contact on the films and a bottom contact to the Si substrate.

The films were irradiated at the GSI/Darmstadt with 1 GeV ²³⁸U ions at an ion fluence of 5·10⁹ U/cm² and 1·10¹⁰ U/cm². The ion fluence was determined from the area density of hillocks on ta-C thin films irradiated together with the VO₂ samples. The GSI LINAC provided ²³⁸U²⁸⁺ ions with 11.4 MeV/u, which were decelerated to the desired energy by passing through an Al degrader foil. The ions leave the degrader with their equilibrium charge state of about U⁴⁰⁺. For 1 GeV ²³⁸U the electronic energy loss in VO₂ reaches the maximum value of 50 keV/nm (99.9 %), the nuclear energy loss is about 50 eV/nm (0.1 %) and within a 100 nm film about 70 vacancies per ion are created.⁵⁴ Most of the collisional defects are located in the vicinity of the ion path and will be eliminated during the subsequent rapid melting and quenching in the thermal spike.

Temperature dependent I-V measurements were recorded with a Keithley 237 source measure unit connected to an automated temperature control system. Evaporated Au contacts of 0.9 mm diameter were used as top contacts. The effective spacing of the top contacts is 0.75 mm. Back contacts to the Si substrate were made with Ag paste. Prior to gluing the sample onto the holder, the backside was cleaned with diluted HF to remove the oxide. For measurements at low temperature down to 15 K the sample was mounted on a sapphire holder on a closed cycle refrigerator. Irradiated and as-grown samples were analyzed with conductive atomic force microscopy using a PSIA XE-100 microscope.

I-V curves were measured in-plane to the film, i.e. through two top contacts but also perpendicular by Ag-paste contact to the Si substrate. Figure 2 shows the resistance versus temperature curves measured in-plane for a bias voltage of 0.5 V, comparing the un-irradiated film and the films irradiated with 1 GeV U at different ion fluences. The un-irradiated film exhibits MIT at a temperature of 68 °C during heating and 56 °C for cool-down with a ratio of R(30 °C)/R(90 °C) ≈ 400. After irradiation with 5·10⁹ U/cm² the room temperature resistance decreases by a factor of 3. The high temperature resistance remains almost unchanged. At the same time the MIT decreases slightly to 65 °C. The temperature hysteresis width decreases from initially 12 K to 9 K. The change in resistance at room temperature is persistent, i.e. it does not anneal out by heating the sample to 90 °C. The same hysteresis of the MIT is observed after multiple heating and cooling cycles. Irradiation with a higher fluence of 1·10¹⁰ U/cm² causes a further persistent decrease of the room temperature resistance by a factor of 3. Again, the high temperature resistance remains nearly unchanged and increases only by about 30%. The MIT further decreases to 63 °C with a hysteresis width of 7 K. The MIT for the cooling down cycles remains nearly constant at about 56 °C. The downward shift of the MIT transition temperature and changes in the hysteresis are similar to previous observations for Cr-doped V₂O₃ upon swift heavy ion irradiation^8,10 and are indications for an ion-induced modification of the VO₂ material, possibly due to stress around the ion tracks.

I-V curves measured at different temperatures are shown in figure 3. The shape of the symmetric curves is rather well described by Ohmic conduction for T = 90 °C and also for the un-irradiated sample at 30°C. The I-V curves of the irradiated samples measured at 30°C exhibit a non-linear increase of current with increasing voltage, which is an indication for Poole-Frenkel conduction.^55,56 Resistance versus temperature measurements were carried out for both the as-grown and irradiated films. The samples were mounted onto the cold head of a closed cycle refrigerator in vacuum. R-T curves obtained for different bias voltages are shown in figure 4. For the whole temperature range below the MIT down to 15 K the resistance of the irradiated films remains significantly lower compared to the un-irradiated ones. Therefore, the resistivity increase due to ion irradiation is persistent and a MIT shifted below room temperature does not occur. On the other hand one should note that the resistance above the MIT increases by a factor of 2 upon ion irradiation. As can be seen in figure 3 the behavior is Ohmic at 90 °C and for un-irradiated films also at 30 °C. At 30 °C, I-V curves of the irradiated films become s-shaped indicating that thermally activated non-linear contributions to the conductivity, such as Poole-Frenkel conduction, come into play.

Values for the room temperature resistivity of VO₂ single crystals and thin films, mainly measured for films on insulating substrates, are between 0.4 Ωcm and 10 Ωcm and typical reported resistivity for T = 70 - 90 °C are between 300 μΩcm – 10 mΩcm.^37,57,58 Depending on the film quality, the conductivity increases by 3 – 4 orders of magnitude when heating the samples above the MIT temperature. Adopting a resistivity of 1 Ωcm for VO₂ at room temperature and 1 mΩcm at 90°C we can analyze the R-T curves of the un-irradiated film (Figure 2) and find that the measured resistance is due to both, an in-plane current flow in the VO₂ film between two top contacts and a current flow through the n+ Si substrate. Current flow through the highly conducting n-Si substrates seems to be limited by a rather high resistance, which could be a resistance due to a thin insulating interface silica layer. In this case the R-T curves measured with a top and bottom contact should then essentially reflect the resistance of this interface resistance. However, as can be seen in figure 4, there is a MIT visible as pronounced as for the in-plane measurement. A high serial interface resistance is therefore unlikely. If one assumes that VO₂ interconnects through the silica interface to the Si substrate at a few nm sized spots, then the total area of such interconnections, rather than the top contact area, represents the contact area for the top-bottom geometry. The measured high resistance values for the top-bottom geometry is then explained by the rather small contact area given by the interconnected silica interface.

Upon SHI irradiation, the room temperature resistance strongly decreases with increasing ion fluence. On the other hand, the high temperature resistance slightly increases, in the case of top-bottom measurements (Figure 4) even by a factor of two.

One possible explanation for the persistent decrease in resistivity below the MIT could be the formation of conducting ion tracks generated in the interfacial layer, thus increasing the effective contact area to the n-Si substrate. Such amorphous conducting paths would also explain the observed non-linear conduction similar to Poole-Frenkel conduction (Figure 3). We would expect a lower resistivity above the MIT temperature as well, however, this resistance value could be limited by a contact resistance. A simplified circuit diagram for the measurement setup used in figure 2 and figure 4 is shown in figure 5. For a contact resistance of ≤ 100 Ω and negligible resistance of the Si substrate we find a perpendicular resistance R_⊥ = 25 kΩ of the non-irradiated system at room temperature (figure 4). After irradiation with 10¹⁰ U/cm² the perpendicular resistance drops to 2 - 3 kΩ (bias ∼ 0.25 V). For the top-top measurement setup with 0.5 V bias (Figure 2) the total resistance at room temperature is about 45 kΩ and drops to 4 kΩ after 10¹⁰ U/cm² irradiation. Using these values we can calculate a parallel resistance of R_∥ ∼ 300 kΩ, corresponding to a resistivity of ρ ≈ 3 - 4 Ωcm of VO₂ at room temperature. For a VO₂ film on sapphire we determined a resistivity of ρ ≈ 4 - 5 Ωcm for the same contact geometry. This simple analysis suggests that the persistent resistivity changes may be explained by formation of conducting ion tracks in the silica interface layer, which allows tuning of the perpendicular resistance R_⊥ through ion irradiation and leaving the much higher parallel resistance R_∥ of the VO₂ film unchanged.

Another explanation would be a persistent ion-induced increase of the conductivity of the VO₂ film itself, either in an isotropic way or through conducting ion tracks similar to tetrahedral amorphous carbon.

An example of an AFM image and current mapping of an identical sample region of the sample irradiated with 5·10⁹ U/cm² is shown in Figure 6. The fluence corresponds to about 12±4 ion tracks within the image area of 0.25 μm². Compared to other materials like tetrahedral amorphous carbon¹³ or ZrO₂,⁵⁹ there are no indications of hillock structures in the topography image. It is therefore difficult to identify the location of possible ion tracks. However, in the current map there appear small features of about 20 nm in diameter with low conductivity surrounded by a halo with higher conductivity. Such features are preferentially seen in the irradiated films but the identification is complicated by the rather complex current distribution even for un-irradiated films (Figure 1).

To get an idea of the possible track structure and its diameter we adopt results obtained for swift heavy ion irradiation of vanadium sesquioxide V₂O₃ by Kokabi et al..⁸ From an XRD analysis of relative intensity changes of diffraction peaks an average track diameter of about 5.4 nm was calculated and discontinuous tracks with damage regions were seen in TEM images. Also for swift heavy ion irradiated ZrO₂ thin films, Moll et al. and Sattonnay et al..^59,60 observed tracks consisting of damaged rather than amorphous material with dislocation loops along the ion trajectories. Using Rutherford Backscattering/Channeling and XRD an average track diameter of about 5 - 6 nm was determined. AFM also showed hillock formation for SHI irradiated ZrO₂ with a hillock height of 1.5 nm.

Adopting a track diameter of 6 nm in V₂O₃ and ZrO₂ also for VO₂, the total ion track cross sectional area is only 0.14 % of the surface area for 5·10⁹ U/cm² fluence and 0.3 % for 10¹⁰ U/cm². Therefore, we expect well separated individual ion tracks. If the features seen in the current mapping in figure 6 are due to ion tracks, then these ion tracks in VO₂ would not exhibit a high conductivity. However, if the track structure is a damage track similar to that reported for V₂O₃^8,10 and ZrO₂,^59,60 the damage regions may generate stress in the region around the tracks. The conducting ‘halo’ indicated in the AFM images (Fig. 6) could then be generated by stress induced by the ion track, which may stabilize the conductive state of VO₂ in the ‘halo’ region. This behavior would be in agreement with the results of Kokabi et al. obtained for swift heavy ion irradiation of Cr-doped V₂O₃,^8,10 and also for magnetoelastic effects seen in magnetite and ferrites,^9,11 as discussed in the introduction.

Since no conducting ion tracks seem to exist in VO₂ we try to explain the persistent conductivity increase assuming a rather isotropic ion-induced conductivity increase of VO₂. Again we adopt the circuit diagrams of figure 5 and simply reduce all resistances by the same factor to obtain the measured total resistances for different irradiation conditions. In this picture R_∥ decreases from 300 kΩ for the non irradiated film down to 30 kΩ after irradiation with 10¹⁰ U/cm². In the same way R_⊥ decreases from 25 kΩ to 2.5 kΩ. However, such a prominent conductivity increase is not seen in the conductive AFM measurements, which could be an indication that indeed conducting tracks through the silica interface may be the dominating ion irradiation effect. Further ion irradiation experiments of VO₂ films on single crystal oxide substrates are scheduled to clarify the origin of the ion-induced conductivity changes.

We have demonstrated that swift heavy ion irradiation of VO₂ thin films on Si substrates with 1 GeV U ions at very low ion fluence up to 10¹⁰ U/cm² allows efficient tuning of the conductivity of the samples in particular for temperatures below the MIT. Irradiation with only 10¹⁰ U/cm² increases the room temperature conductivity by almost an order of magnitude, whereas the high temperature conductivity is only slightly decreased. The conductivity increase below the MIT is proportional to the ion fluence and is persistent. Only a small shift of the MIT temperature toward lower values is observed. There is no indication of hillock formation, so that the mass density of ion tracks in VO₂ is not significantly changed compared to the virgin film density. The absence of hillocks makes it difficult to identify the tracks with AFM topography images. Conductive AFM measurements give no indication of good conducting ion tracks in VO₂ similar to tracks formed in tetrahedral amorphous carbon, however there might be tracks with poor conductivity surrounded by a conducting halo. An analysis of the resistance versus temperature curves for different irradiation fluences currently allow two alternative explanations for the observed conductivity changes. In one model conducting ion tracks are formed in an initially high resistive silica interface layer. The existence of such an interface layer was previously observed in TEM²⁹ and is required to explain the unusual high resistance of the samples for measurements with top and bottom contact. Conducting ion tracks in the interface layer could explain the nonlinear I-V curves seen after irradiation. In the other model the conductivity of VO₂ is increased isotropically, possibly due to stress generated by damage regions along the ion tracks, similar to earlier observations for V₂O₃.^8,10 In particular, the shift in transition temperature towards lower temperatures and the changes in temperature hysteresis are similar to the behavior seen for V₂O₃ upon ion irradiation and are an indication for ion-induced effects on VO₂. Swift heavy ion irradiation of VO₂ on different insulating and conducting substrates are scheduled to distinguish between both models and will be the focus of future studies.

This work was financially supported by the Deutsche Forschungsgemeinschaft under grant HO 1125/17-1 and by BMBF under grant 05KK7MG2.