IEEE ELECTRON DEVICE LETTERS, VOL. 27, NO. 5, MAY 2006 309

Piezoelectric GaN Sensor StructuresT. Zimmermann, M. Neuburger, P. Benkart, F. J. Hernández-Guillén, C. Pietzka, M. Kunze,

I. Daumiller, A. Dadgar, A. Krost, and E. Kohn

Abstract—Free-standing GaN and AlGaN/GaN cantilevers havebeen fabricated on (111) silicon substrate using dry etching. Onthese cantilevers, a piezoresistor and a high-electron-mobilitytransistor (HEMT) structure have been realized, and the piezore-sponse has been characterized. Cantilever bending experimentsresulted in a Young’s modulus of approximately 250 GPa, a sen-sitivity of K ∼ 90, and a modulation of the HEMT current ofup to 50%. It is seen that the piezoresponse could be related toboth the bulk properties and the properties of the heterostructureinterface.

Index Terms—AlGaN/GaN, cantilever, microelectromechanicalsystem (MEMS), piezoresistor, polarization, sensor.

I. INTRODUCTION

THE STRONG spontaneous polarization and piezopolar-ization of the GaN-based material system has been widely

used in the design of field-effect transistor (FET) and light-emitting diode (LED) structures. Also being attractive forsensor applications, some interesting piezoeffects have beeninvestigated [1]–[3], but only a few basic studies have been pub-lished in respect to cantilever devices [4]–[7], mainly due to thedifficulty to fabricate membranes and free-standing cantileverstructures on sapphire [8], SiC [9], Si [10], and LiNbO3 [11].In this investigation, GaN-based device structures have beengrown on (111)-oriented Si by metal–organic chemical vapordeposition (MOCVD) [12], where the Si can be removed fromthe back or the front side by dry etching. As a highly polar andwide bandgap ceramic semiconductor, GaN has the advantageto operate in harsh environment and at high temperature.

In contrast to former piezoceramics, GaN is a semiconduct-ing piezomaterial and combines the piezoresistive and piezo-electric behaviors. Therefore, electromechanical investigationsresult in a so-called piezoresponse. With its large spontaneouspolarization, several effects can contribute to the piezoresponse.First, bending a bulk GaN cantilever induces a vertical stressgradient, which in turn induces polarization doping (similar to

Manuscript received November 28, 2005. The review of this letter wasarranged by Editor J. del Alamo.

T. Zimmermann, P. Benkart, F. J. Hernández-Guillén, C. Pietzka, andE. Kohn are with the Department of Electron Devices and Circuits, Universityof Ulm, Ulm 89081, Germany (e-mail: Tom.Zimmermann@ebs.e-technik.uni-ulm.de).

M. Neuburger was with the Department of Electron Devices and Circuits,University of Ulm, Ulm 89081, Germany. He is now with Robert Bosch GmbH,AE/EDP3, Postfach 1342, Reutlingen 72703, Germany.

M. Kunze and I. Daumiller are with MicroGaN GmbH, Albert Einstein Allee45, Ulm 89081, Germany.

A. Dadgar and A. Krost are with the Institute of Experimental Physics, Ottovon Guericke University Magdeburg, Magdeburg 39016, Germany.

Digital Object Identifier 10.1109/LED.2006.872918

the case of a graded AlGaN layer [13]). For Ga-face material,this will result in p-type (bulk) polarization doping for upwardbending and n-type (bulk) polarization doping for downwardbending. Second, there is the energy band structure realignmentwith its associated carrier redistribution in the transversal andlongitudinal valleys [14]. In a heterostructure, there is addition-ally the modulation of the interfacial two-dimensional electron-gas (2DEG) or two-dimensional hole-gas (2DHG) density bystress modulation of the heterostructure interface.

In the following sections, single-anchored cantilevers havebeen fabricated and investigated to demonstrate a highpiezosensitivity.

II. EXPERIMENTAL DETAILS

The key for the fabrication of free-standing GaN cantileversis the growth of GaN on silicon. However, several problemsarise in this case. For seeding and suppression of meltback etch-ing of Si, GaN is nucleated on a low-temperature-grown AlN(LT-AlN) nucleation layer [15]. Furthermore, the difference inthermal expansion leads to tensile stress in the GaN layer aftercooling from the growth temperature (approximately 1000 ◦C)and cracking for a layer thickness above 1 µm. To release theaccumulated stress, LT-AlN stress release interlayers are intro-duced periodically, enabling the growth of thick GaN layers inexcess of 6 µm [16]. These LT-AlN interlayers introduce a highcompressive stress, which may, by overcompensation, lead toconvex bowing of the Si wafer. Therefore, full strain compensa-tion may be difficult to obtain. In our case, the GaN piezoresis-tor material is 3.0 µm thick and contains one stress relieve layer[see cross section in Fig. 1(a)]. The AlGaN/GaN cantilever ma-terial stack has a thickness of 1.5 µm also containing one stressrelief layer. The Si-doped layer of the piezoresistor displayed asheet charge density of Ns = 3 × 1011 cm−2 and a conductivityof 1.7 S/cm. The AlGaN heterostructure contained a 25-nm-thick nominally undoped AlGaN barrier layer with 26% Alcontent and an AlGaN/GaN interfacial sheet charge density(measured before cantilever etching) of Ns = 8 × 1012 cm−2.The semi-insulating GaN buffer and support layer was compen-sated by Fe doping throughout its thickness [17].

For the piezodevices located at the pivot point of thecantilever, a commonly used FET fabrication process,including mesa etching by Ar dry etching, Ti/Al/Ni/Au ohmiccontact formation, and Ni/Au Schottky gate contact patterning,has been applied [18]. The GaN cantilever pattern was etchedin an Ar plasma, and the Si was removed by anisotropic siliconetching (ASE) from the rear side (“Bosch” process) or in a CF4

plasma by reactive ion etching (RIE) from the front side (by

0741-3106/$20.00 © 2006 IEEE

310 IEEE ELECTRON DEVICE LETTERS, VOL. 27, NO. 5, MAY 2006

Fig. 1. (a) Schematic of the cross section of used GaN structures. (b) GaNcantilever structures fabricated by membrane etching of the Si substrate by ASEfrom the rear side. (c) Piezoresistor located at the pivot point of the cantilever.

undercut), patterning being obtained by a 1-µm-thick Al mask.Fig. 1(b) shows a cantilever structure obtained by backsideetching. The cantilever length and width of the largest beam are200 and 100 µm, respectively. Fig. 1(c) shows a piezoresistorstructure with 80 µm width and 30 µm contact spacing atthe pivot point of the cantilever. In the high-electron-mobilitytransistor (HEMT) structure also placed at the pivot point at thecantilever, the identical geometry was used with a 2-µm gatecentered symmetrically between the source and the drain. Afterremoving of the Si substrate, the residual stress that is caused bythe thermal mismatch of the epitaxial layer with the substrateand by an internal stress distribution due to inhomogeneousoutgrowth of the layer is changed. Therefore, in most cases, theunstressed beams were bent downward, indicating a verticalbuilt-in stress component. The bending experiments weretherefore, in essence, restricted to upward bending.

Cantilever bending was performed using a needle driven bya high-precision stepping motor. The change in current wasmeasured in the dark after 10 min settling time to avoid theinfluence of deep traps.

Fig. 2. (a) Load–displacement characteristics for a double-anchored GaN can-tilever and fit to simulation resulting in a Young’s modulus of E ∼ 250 GPa.Fracture occurred at P = 350 MPa, stressing the beam in the opposite directionof the built-in prestress. (b) Stress experiment of a double-anchored GaNcantilever structure etched out of the 1.5-µm-thick GaN film, resulting in agauge factor of K = (dR/R)/(dL/L) ∼ 90.

III. MEASUREMENT RESULTS

The mechanical properties have been analyzed for the1.5-µm-thick GaN film by an iterative process of experimentaland simulation data due to the difficulty of its analytical calcula-tion (using FEM ANSYS software). First, a load–displacementexperiment was performed on a double-anchored cantilever.Values of the load–displacement experiments were comparedwith the theoretical data obtained by simulation. Each of thetheoretical curves was simulated for a different value of theYoung’s modulus. Then, an estimated Young’s modulus, in ourcase approximately 250 GPa (similar to [19]), can be extractedfrom the theoretical curve that matches the experimental data[Fig. 2(a)]. The simulation took into account a constant verticalbuilt-in stress of 125 MPa, extracted from the vertical dis-placement of single-anchored cantilever measured by scanningelectron microscopy (SEM), to adjust the Young’s modulus andto obtain an accurate fracture strength. The resimulation of thestructures and the matching with the experimental data resultedin an unaltered value of the Young’s modulus and in a fracturestrength of 350 MPa. Double-anchored beams were used forthe mechanical measurements because, in the case of single-anchored cantilevers, the needle would slip off the surface at ahigh bending angle.

ZIMMERMANN et al.: PIEZOELECTRIC GaN SENSOR STRUCTURES 311

Fig. 3. (a) Response of an n-type doped channel piezoresistor (Ns = 3 ×1011 cm−2) for vertical displacement of the cantilever (L = 200 µm, W =100 µm) at 1 V applied voltage. The initial current level I0 was 24 µA, andthe resistor dimensions were L = 80 µm and W = 30 µm. Current decreaseswith upward bending as indicated. (b) Relative change in HEMT output currentfor two gate bias points for a displacement of up to 30 µm. The inset showsthe change of output characteristic by 30 µm cantilever bending. Cantileverlength L = 200 µm. HEMT structure data: W = 80 µm, LSD = 30 µm,Lg = 2.0 µm, IDsat = 80 mA/mm for VGS = 0 V, VDS = 3.0 V, andVth = −4.0 V.

Taking the value of the Young’s modulus estimatedfrom the previously described mechanical measurement, thestress–displacement characteristic yields in a gauge factor ofK ∼ 90 [Fig. 2(b)]. This is very comparable to previous mea-surements [20].

The electrical piezoresponses have been measured on single-anchored beams and were realized by a vertical displacementof the cantilever with a needle at the free end. Electrical mea-surements on single-anchored beams were reproducible and nota feature of degradation by cracks.

In the case of the GaN beam with integrated doped-channelresistor at the pivot point of the cantilever, the channel resis-tance could be modulated by approximately 20% for an upwardbending of 40 µm [Fig. 3(a)]. Upward bending resulted in acurrent decrease. This is indeed expected inasmuch as p-typemodulation starts to be generated in this case.

Fig. 3(b) shows the piezoresponse of the AlGaN/GaN HEMTstructure with piezoinduced 2DEG channel. In [21], pressure-induced conductivity changes of a membrane were illustrated.Here, the piezoresponse of a gated AlGaN/GaN heterostructure(HFET) structure on the pivot point of the cantilever havebeen measured. By upward bending of the cantilever, the chan-

nel current is again decreased; the output characteristics arecompressed, the relatively low saturated output current being80 mA/mm at VGS = 0 V for the unstressed case and is causedby the large transistor geometry. Changes in IDS for two dif-ferent gate bias points VGS = 0 V and VGS = −3 V are plotted.The highest relative change in current is obtained near pinch-off. At VGS = −3 V this is 50% (conductivity change as a func-tion of strain in AlGaN/GaN structure also investigated in [22]).

IV. DISCUSSION

As already mentioned previously, several effects could con-tribute to a cantilever bending such as carrier redistribution inthe longitudinal and transversal energy valleys in the bulk onone hand and the piezopolarization effects in the bulk and atinterfaces on the other hand. Especially the generation of po-larization doping in the bulk is a novel semiconductor conduc-tion phenomenon [13]. Bulk conduction is usually considereddetrimental in electroceramics because it will shield the surfacepiezocharges [23]. It should therefore be shortly discussed.

Bending a cantilever will induce a vertical stress gradientacross the beam, generating a polarization gradient, whichin turn will generate a distribution of free carriers (holes orelectrons) [13] across the beam. The corresponding donors oracceptors are represented by the stress-induced change in thebonded polarization charge distribution. Upward bending of theGa-face-oriented GaN cantilever, which is usually the case forMOCVD-grown layers on an LT-AlN buffer, will show p-typepolarization doping in the bulk. Correspondingly, downwardbending will result in n-type polarization doping, both gen-erating bulk conduction paths in parallel to the piezoresistivechannel.

The conduction of an n-type piezoresistor will therefore bemodulated in two ways, namely 1) the channel conduction willbe modulated by the piezo-induced change in mobility and2) upward bending will result in p-type backgate and downwardbending in an n-type buffer layer bypass.

In the undoped AlGaN/GaN HEMT structure, the channelcurrent is entirely generated by the heterointerface polarizationdiscontinuity. Therefore, in addition to the aforementioned ef-fects, the piezoresponse of the interface needs to be considered.Downward bending will stress the already prestressed AlGaNbarrier layer further and enhance the piezo-induced 2DEGdensity, whereas upward bending will have the opposite effect.

However, both cases discussed previously are ideal and donot reflect the technical material stack. The bulk of the GaNbuffer layer is semi-insulating by Fe doping. This doping con-centration may well overcompensate any polarization dopinggenerated. Furthermore, the AlN stress release layers are alsohighly polarized forming dipoles. Their magnitude may bemodulated [16]. Finally, the films are intrinsically prestressedalready, generating an intrinsic stress gradient. It is thereforeat present not yet possible to separate the various effectsqualitatively.

V. CONCLUSION AND OUTLOOK

A technology has been developed to fabricate free-standingGaN membrane and cantilever structures on Si substrate. For

312 IEEE ELECTRON DEVICE LETTERS, VOL. 27, NO. 5, MAY 2006

cantilever structures etched from these membranes, a Young’smodulus of approximately 250 GPa and a fracture strength of350 MPa for bending in the opposite direction of the prestresshave been measured. First electrical experiments have enabledus to verify the piezoresponse of n-doped GaN piezoresistorsand AlGaN/GaN piezo-HEMTs. A gauge factor of approxi-mately K ∼ 90 could be extracted. Current modulation hasbeen as high as 20% for the piezoresistor and 50% for thepiezo-HEMT for a displacement of 30 µm of a 200-µm-longbeam. Especially the piezoresponse of the piezoresistor onan n-type Si-doped GaN layer clarifies that the polarizationdoping generated within the stressed films is a new importantcontribution but could not yet be determined qualitatively. Theresults also show that the response of the piezo-HEMT maynot entirely be dominated by the piezomodulation of its 2DEGdensity.

REFERENCES

[1] M. Eickhoff, J. Schalwig, G. Steinhoff, O. Weidemann, L. Görgens,R. Neuberger, M. Hermann, B. Baur, G. Müller, O. Ambacher, andM. Stutzmann, “Electronics and sensors based pyroelectric AlGaN/GaNheterostructures, Part B: Sensor applications,” Phys. Stat. Sol. C, vol. 0,no. 6, pp. 1908–1918, 2003.

[2] Y. Takagaki, Y. J. Sun, O. Brandt, and K. H. Ploog, “Strain relaxation inAlN/GaN bilayer films grown on γ-LiAlO2 (100) for nanoelectromechan-ical systems,” Appl. Phys. Lett., vol. 84, no. 23, pp. 4756–4758, Jun. 2004.

[3] R. Gaska, J. W. Yang, A. D. Bykhovski, M. S. Shur, V. V. Kaminskii, andS. Soloviov, “Piezoresistive effect in GaN–AlN–GaN structures,” Appl.Phys. Lett., vol. 71, no. 26, pp. 3817–3819, Dec. 1997.

[4] S. Davies, T. S. Huang, M. H. Gass, A. J. Papworth, T. B. Joyce, andP. R. Chalker, “Fabrication of GaN cantilevers on silicon substratesfor microelectromechanical devices,” Appl. Phys. Lett., vol. 84, no. 14,pp. 2566–2568, Apr. 2004.

[5] R. P. Strittmatter, R. A. Beach, G. S. Picus, and T. C. McGill, “Piezoelec-trically enhanced capacitive strain sensors using GaN metal–insulator–semiconductor diodes,” J. Appl. Phys., vol. 94, no. 9, pp. 5958–5963,Nov. 2003.

[6] Y.-R. Wu and J. Singh, “Metal piezoelectric semiconductor field effecttransistors for piezoelectric strain sensors,” Appl. Phys. Lett., vol. 85,no. 7, pp. 1223–1225, Aug. 2004.

[7] B. S. Kang, S. Kim, J. Kim, F. Ren, K. Baik, S. J. Pearton, B. P. Gila,C. R. Abernathy, C.-C. Pan, G.-T. Chen, J.-I. Chyi, V. Chandrasekaran,M. Sheplak, T. Nishida, and S. N. G. Chu, “Effect of external strain onthe conductivity of AlGaN/GaN high-electron-mobility transistors,” Appl.Phys. Lett., vol. 83, no. 23, pp. 4845–4847, Dec. 2003.

[8] A. D. Bykhovski, V. V. Kaminski, M. S. Shur, Q. C. Chen, and M. A.Khan, “Piezoresistive effect in wurtzite n-type GaN,” Appl. Phys. Lett.,vol. 68, no. 6, pp. 818–819, Feb. 1996.

[9] L. Jiang, M. Hassan, A. J. Harris, J. S. Burdess, C. A. Zorman,M. Mehregany, and R. Cheung, “Fabrication of SiC microelectromechan-

ical systems using one-step dry etching,” J. Vac. Sci. Technol. B, Micro-electron. Process. Phenom., vol. 21, no. 6, pp. 2998–3001, Nov. 2003.

[10] B. S. Kang, J. Kim, S. Jang, F. Ren, J. W. Johnson, R. J. Therrien,P. Rajagopal, J. C. Roberts, E. L. Piner, K. J. Linthicum, S. N. G. Chu,K. Baik, B. P. Gila, C. R. Abernathy, and S. J. Pearton, “Capacitancepressure sensor based on GaN high-electron-mobility transistor-on-Simembrane,” Appl. Phys. Lett., vol. 86, no. 25, pp. 253502-1–253502-3,Jun. 2005.

[11] C. Sones, S. Mailis, V. Apostolopoulos, I. E. Barry, C. Gawith, P. G. R.Smith, and R. W. Eason, “Fabrication of piezoelectric micro-cantileversin domain-engineered LiNbO3 single crystals,” J. Micromech. Microeng.,vol. 12, no. 1, pp. 53–57, Jan. 2002.

[12] A. Krost and A. Dadgar, “GaN based devices on Si,” Phys. Stat. Sol. A,vol. 194, no. 2, pp. 361–375, 2002.

[13] D. Jena, S. Heikman, D. Green, D. Buttari, R. Coffie, H. Xing, S. Keller,S. Denbaars, J. S. Speck, I. Smorchkova, and U. K. Mishra, “Realizationof wide electron slabs by polarization bulk doping in graded III–V nitridesemiconductor alloys,” Appl. Phys. Lett., vol. 81, no. 23, pp. 4395–4397,Dec. 2002.

[14] J.-Q. Wang, Z.-Q. Gu, M.-F. Li, and W.-Y. Lai, “Intervalley Γ−X de-formation potentials in III–V zinc-blende semiconductors by ab initiopseudopotential calculations,” Phys. Rev. B, Condens. Matter, vol. 46,no. 19, pp. 12358–12364, Nov. 1992.

[15] A. Dadgar, A. Strittmatter, J. Bläsing, M. Poschenrieder, O. Contreras,P. Veit, T. Riemann, F. Bertram, A. Reiher, A. Krtschil, A. Diez,T. Hempel, T. Finger, A. Kasic, M. Schubert, D. Bimberg, F. A. Ponce,J. Christen, and A. Krost, “Metalorganic chemical vapor phase epi-taxy of gallium-nitride on silicon,” Phys. Stat. Sol. C, vol. 0, no. 6,pp. 1583–1606, 2003.

[16] A. Krost, A. Dadgar, G. Strassburger, and R. Clos, “GaN-based epitaxyon silicon: Stress measurements,” Phys. Stat. Sol. A, vol. 200, no. 1,pp. 26–35, 2003.

[17] S. Heikman, S. Keller, S. P. DenBaars, and U. K. Mishra, “Growth of Fedoped semi-insulating GaN by metalorganic chemical vapor deposition,”Appl. Phys. Lett., vol. 81, no. 3, pp. 439–441, Jul. 2002.

[18] I. Daumiller, P. Schmid, E. Kohn, C. Kirchner, M. Kamp, K. J. Ebeling,L. L. Pond, and C. Weitzel, “DC and RF characteristics of AlN/GaN dopedchannel heterostructure field effect transistor,” Electron. Lett., vol. 35,no. 18, pp. 1588–1590, Sep. 1999.

[19] I. Vurgaftman, J. R. Meyer, and L. R. Ram-Mohan, “Band parameters forIII–V compound semiconductors and their alloys,” J. Appl. Phys., vol. 89,no. 11, pp. 5815–5875, Jun. 2001.

[20] M. Stutzmann, G. Steinhoff, M. Eickhoff, O. Ambacher, C. E. Nebel,J. Schalwig, R. Neuberger, and G. Müller, “GaN-based heterostruc-tures for sensor applications,” Diamond Relat. Mater., vol. 11, no. 3–6,pp. 886-891, 2002.

[21] B. S. Kang, J. Kim, F. Ren, J. W. Johnson, R. J. Therrien, P. Rajagopal,J. C. Roberts, E. L. Piner, K. J. Linthicum, S. N. G. Chu, K. Baik, B. P.Gila, C. R. Abernathy, and S. J. Pearton, “Pressure-induced changes inthe conductivity of AlGaN/GaN high-electron mobility-transistor mem-branes,” Appl. Phys. Lett., vol. 85, no. 14, pp. 2962–2964, Oct. 2004.

[22] M. Eickhoff, O. Ambacher, G. Krötz, and M. Stutzmann, “Piezoresistivityof AlxGa1−xN layers and AlxGa1−xN/GaN heterostructures,” J. Appl.Phys., vol. 90, no. 7, pp. 3383–3386, Oct. 2001.

[23] A. D. Carlo, F. D. Sala, P. Lugli, V. Fiorentini, and F. Bernadini, “Dopingscreening of polarization fields in nitride heterostructures,” Appl. Phys.Lett., vol. 76, no. 26, pp. 3950–3952, Jun. 2000.