- Email: [email protected]
FIB-induced electro-optical alterations in a DFB InP laser diode
Microelectronics Reliability 54 (2014) 2151–2153
Contents lists available at ScienceDirect
Microelectronics Reliability journal homepage: www.elsevier.com/locate/microrel
FIB-induced electro-optical alterations in a DFB InP laser diode G. Mura, M. Vanzi ⇑, G. Marcello DIEE, University of Cagliari, Piazza d’Armi, 09123 Cagliari, Italy
a r t i c l e
i n f o
Article history: Received 27 June 2014 Accepted 8 July 2014 Available online 10 August 2014 Keywords: Laser diode XEBIC Focussed ion beam
a b s t r a c t A recent experiment used FIB to induce local modification in a single-mode edge emitting laser operating at 1310 nm, and led to a method for estimating gain parameters. In this experiment, the final FIB modification introduced large variation in electrical characteristics that were not analysed in detail, and even seemed to contradict the basic laser model that the experiment itself aimed to confirm. This paper focuses on this puzzling point, and solves it with a circuital hypothesis, a circuit simulation and a direct inspection by XEBIC. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction
gT, both obtained by the L–I curves by interpolating the experimental data, with the well-known linear formula [3]
In a recent paper [1] authors studied the variation of the gain coefficients in some kinds of single-mode laser diodes. One kind was made of a 1310 nm edge emitting DFB InP-based device, progressively modified by FIB erosion at one side of the optical cavity. The evolution of both the threshold current and the optical efficiency were monitored, and shown to be due to the sole variation of the internal optical losses. This consideration led to draw a gain/ loss curve and then to predict the zero-loss coefficients, that are related to the transparency condition for the given material and rule over all real-loss cases. Anyway, when looking at the full set of available data, that also included the evaluation of the series resistance RS and the internal threshold voltage Vth (corresponding to the separation of the quasiFermi levels required to achieve lasing), the last FIB-induced state showed really puzzling figures, including a reduction of qVth, as for a reduced optical loss. Moreover, such last experimental result seemed to destroy the underlying DC model progressively developed and applied during the last few years. The paper analyses this puzzling point, proposes its solution and shows an experimental XEBIC [2] observation that validates the conclusions.
2. Experimentals The measurement of gain coefficients [1] in single-mode laser diodes involves the threshold current Ith and the total efficiency ⇑ Corresponding author. Tel.: +39 (070) 675 5775; fax: +39 (070) 675 5900. E-mail address: [email protected] (M. Vanzi). http://dx.doi.org/10.1016/j.microrel.2014.07.049 0026-2714/Ó 2014 Elsevier Ltd. All rights reserved.
L ¼ gT ðI Ith Þ and using them, for different couples Ith, gT, in the empiric relationship [3]
Ith ¼ Ith0 exp
c aT ¼ Ith0 exp gT g0
where Ith0 is the extrapolated zero-loss threshold current, that should coincide with the value of the total current I measured at qV = hm, where m is the nominal single-mode operating frequency of the given device, aT is the total loss coefficient, g0 is the gain coefficient corresponding to the pure gain at transparency and c is a suitable coupling constant. The alignment of the experimental data along a straight line confirms that only internal losses lead to changes in aT, and data fitting leads to the ratios aT/g0, to the coupling constant c and to Ith0. Such analysis led to the curve in Fig. 1, that seems to fully confirm all expectations. The result of the FIB experiment perfectly reproduce the figures obtained in the completely different case of an external-cavity laser, whose internal losses were directly modulated by means of a tunable mirror, as reported in detail in [1]. The conclusion was that really FIB was able to perturb the sole internal losses, without any significant effect on the other characteristics. The last point is important because many results obtained insofar have been based on an equivalent circuit [4] that, after a common series resistance, links in parallel several elements representing, separately, the light emitting ‘‘diode’’, the non-radiative currents, the lateral currents flowing at low injection at the sides of the active region. Removal
2152
G. Mura et al. / Microelectronics Reliability 54 (2014) 2151–2153
Fig. 3. The simple circuit for describing Fig. 2.
Fig. 1. The alignment of ln (Ith) vs. 1/gT pairs when pure internal losses are involved.
of such a series resistance is then a starting step for any deeper analysis of the electrical end electro-optical characteristics, including that relationship between laser voltage and monitor current IM (V) that, named Trans characteristics in a recent paper [5], lead to reconstruct the pure radiative component Iph of the total laser current I. When one performs such step for the initial state, after the first FIB and after the second FIB, one obtains the result shown in Fig. 2. Here both the laser current I and the relative monitor diode current IM are plotted vs. the separately calculated reduced voltage V RSI, at the initial step (suffix ‘‘0’’), the first FIB (‘‘1’’) and the second FIB (‘‘2’’). Also the corresponding threshold voltages Vth are put into evidence. The surprising result was that after the last FIB, the clamp voltage Vth2 appeared to decrease, while the threshold current increased, in evident conflict with the statement about their relationship. The evidence, in current I2, for a parasitic shunting ohmic path, roughly corresponding to some 700 X, addresses the qualitative interpretation. Anyway, the preservation of alignment of the highest point in Fig. 1 with the others remains puzzling, if relevant extra-currents are invoked.
correspond to the initial state, and the resistance itself can be neglected. FIB is likely to introduce both optical effects, by modifying the lateral losses, and electrical effects in terms of a parasitic shunting path along the walls of the grooved trench. Such electrical effects can be neglected when confined to the sub-mA range, as for current I1 in Fig. 2. In this case, in Fig. 3 the lateral resistance R0 can be omitted, and the dV/dI characteristics correctly give the overall series resistance RS when measured at I > Ith. On the contrary, the last curve I2 shows a dominant leakage current, that does not physically involve the optically active region, but drains current enough to perturb the I(V) relationship in the whole device. It is matter of trivial circuit analysis to see that, when the element L in Fig. 3 clamps at its internal clamp voltage Vth, not only Ith but also gT and the externally measured value of itself Vth change depending on the critical ratio R0/(RS + R0). More precisely, if the following relationships hold for the initial state (no shunting path) for I > Ith.
V ¼ V th þ RS I IM ¼ gðI Ith Þ
ð1Þ
the appearance of the parallel path changes them into
8 <
0 V ¼ RSRþR ðV th þ RS IÞ 0 : IM ¼ R0 g I RS þR0 Ith þ V th RS þR0 R0 R0
ð2Þ
When one considers the configuration in Fig. 3, where the block L represents the laser diode, large values of the shunt resistance R0
The result is that the functional relationships between the measurable quantities V, I, IM remain the same, and that the four constants Vth, Ith, RS, g seem to change their value, mostly depending on the value of the ratio R0/(RS + R0) that for the given case amounts to about 0.985. The correspondence between old and new parameters is:
Fig. 2. I and IM characteristics at the initial state (0), after the 1st FIB (1) and after 2nd FIB (2).
Fig. 4. The last curve IM2 can be brought to recover the previous state, IM1, by adjusting the electrical parameters of the shunting path.
3. Analysis and discussion
G. Mura et al. / Microelectronics Reliability 54 (2014) 2151–2153
2153
It should be noticed that this states that the third point in Fig. 1 is possibly unreliable, and that the FIB method can be safely applied only when evidence exists that no electrical alterations took place at currents higher that the mA range. Finally, Fig. 5 shows a XEBIC map superimposed onto the corresponding SEM image on the polished facet of the device under test after the final FIB. The edges of the floating blocking layer disappear on the left side, that is just the side where FIB created its trenches [1], confirming with full experimental evidence the hypothesis of a more and more important shunting path, shorting the lateral junction, growing during FIB action. Fig. 5. XEBIC confirms the expected shunted path by cancelling the upper blocking junction signal on the left side, where FIB grooved a trench.
8 0 V th ! RSRþR V th > > 0 > > > R < RS ! 0 RS RS þR0 > g ! RSRþR0 0 g > > > > : 0 Ith ! RSRþR Ith þ VRth0 0
ð3Þ
Here is the explanation of the apparent reduction of the clamp voltage Vth: the lateral path kids the electrical model, and emulates the original simple pure serial scheme with false values for the key parameters. Not only the clamp voltage, but even the series resistance RS seems to reduce, challenging any physical interpretation, upon FIB milling of the laser surface. But what is possibly more intriguing is that the qualitative effect of the parasitic path is that it simultaneously increases the threshold current Ith and decreases the total optical efficiency g. This is just what one expects from an increase of optical losses, the foundation of Fig. 1 and a key point in Ref. [1]. It is then natural to wonder if the third point in Fig. 3 is a true effect of pure optical losses, or if it is somewhat affected by electrical effects, or even if it is only an electrical effect. One way at least to infer the answer is to check if, supposing a tentative Rx for the parasitic shunt resistance, one can reproduce or approach the former state IM1 starting from the measured values of IM2, by means of the inverse of Eq. (3). Fig. 4 shows an intermediate situation, that confirms that the variations from state 1 to state 2 can be reproduced by mere electrical means. The choice of a suitable value for Rx allows to continuously recover all intermediate states between 1 and 2. Such a possibility does not say, in principle, that nothing optical took place after the second FIB, but prevents from separating the optical from the electrical effects.
4. Conclusions A puzzling set of measurements in a FIB-modified laser diode has been shown to be explained by a modified equivalent circuit, and confirmed by EBIC. This raises some warnings in applying the FIB method for gain measurement, as reported in [1], when evidence for electrical alterations appear in the mA range of the laser current, which also defines a criterion for validating data in such FIB-based experiments.
Acknowledgments The authors would like to thank Simona Podda, PhD (Telemicroscopy Laboratory – Sardegna Ricerche – Pula) for FIB device preparation. This work is partially supported by Sardinia Regional Government (P.O.R. Sardegna F.S.E. Operational Programme of the Autonomous Region of Sardinia, European Social Fund 2007–2013 – Axis IV Human Resources, Objective l.3, Line of Activity l.3.1 ‘‘Avviso di chiamata per il finanziamento di Assegni di Ricerca’’).
References [1] Vanzi et al. Optical losses in single-mode laser diodes. Microelectron Reliab 2013;53(9–11):1529–33. [2] Vanzi et al. XEBIC at the dual beam. Microelectron Reliab 2013;53(9– 11):1399–402. [3] Coldren LA, Corzine SW, Mašanovic´ ML. Diode lasers and photonic integrated circuits. 2nd ed. Hoboken, New Jersey: John Wiley & Sons, Inc.; 2012. [4] Mura G, Vanzi M. The interpretation of the DC characteristics of LED and laser diodes to address their failure analysis. Microelectron Reliab 2010;50(4):471–8. [5] Mura G, Vanzi M, Marcello G, Cao R. The role of the optical trans-characteristics in laser diode analysis. Microelectron Reliab 2013;53(9–11):1538–42.
Contents lists available at ScienceDirect
Microelectronics Reliability journal homepage: www.elsevier.com/locate/microrel
FIB-induced electro-optical alterations in a DFB InP laser diode G. Mura, M. Vanzi ⇑, G. Marcello DIEE, University of Cagliari, Piazza d’Armi, 09123 Cagliari, Italy
a r t i c l e
i n f o
Article history: Received 27 June 2014 Accepted 8 July 2014 Available online 10 August 2014 Keywords: Laser diode XEBIC Focussed ion beam
a b s t r a c t A recent experiment used FIB to induce local modification in a single-mode edge emitting laser operating at 1310 nm, and led to a method for estimating gain parameters. In this experiment, the final FIB modification introduced large variation in electrical characteristics that were not analysed in detail, and even seemed to contradict the basic laser model that the experiment itself aimed to confirm. This paper focuses on this puzzling point, and solves it with a circuital hypothesis, a circuit simulation and a direct inspection by XEBIC. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction
gT, both obtained by the L–I curves by interpolating the experimental data, with the well-known linear formula [3]
In a recent paper [1] authors studied the variation of the gain coefficients in some kinds of single-mode laser diodes. One kind was made of a 1310 nm edge emitting DFB InP-based device, progressively modified by FIB erosion at one side of the optical cavity. The evolution of both the threshold current and the optical efficiency were monitored, and shown to be due to the sole variation of the internal optical losses. This consideration led to draw a gain/ loss curve and then to predict the zero-loss coefficients, that are related to the transparency condition for the given material and rule over all real-loss cases. Anyway, when looking at the full set of available data, that also included the evaluation of the series resistance RS and the internal threshold voltage Vth (corresponding to the separation of the quasiFermi levels required to achieve lasing), the last FIB-induced state showed really puzzling figures, including a reduction of qVth, as for a reduced optical loss. Moreover, such last experimental result seemed to destroy the underlying DC model progressively developed and applied during the last few years. The paper analyses this puzzling point, proposes its solution and shows an experimental XEBIC [2] observation that validates the conclusions.
2. Experimentals The measurement of gain coefficients [1] in single-mode laser diodes involves the threshold current Ith and the total efficiency ⇑ Corresponding author. Tel.: +39 (070) 675 5775; fax: +39 (070) 675 5900. E-mail address: [email protected] (M. Vanzi). http://dx.doi.org/10.1016/j.microrel.2014.07.049 0026-2714/Ó 2014 Elsevier Ltd. All rights reserved.
L ¼ gT ðI Ith Þ and using them, for different couples Ith, gT, in the empiric relationship [3]
Ith ¼ Ith0 exp
c aT ¼ Ith0 exp gT g0
where Ith0 is the extrapolated zero-loss threshold current, that should coincide with the value of the total current I measured at qV = hm, where m is the nominal single-mode operating frequency of the given device, aT is the total loss coefficient, g0 is the gain coefficient corresponding to the pure gain at transparency and c is a suitable coupling constant. The alignment of the experimental data along a straight line confirms that only internal losses lead to changes in aT, and data fitting leads to the ratios aT/g0, to the coupling constant c and to Ith0. Such analysis led to the curve in Fig. 1, that seems to fully confirm all expectations. The result of the FIB experiment perfectly reproduce the figures obtained in the completely different case of an external-cavity laser, whose internal losses were directly modulated by means of a tunable mirror, as reported in detail in [1]. The conclusion was that really FIB was able to perturb the sole internal losses, without any significant effect on the other characteristics. The last point is important because many results obtained insofar have been based on an equivalent circuit [4] that, after a common series resistance, links in parallel several elements representing, separately, the light emitting ‘‘diode’’, the non-radiative currents, the lateral currents flowing at low injection at the sides of the active region. Removal
2152
G. Mura et al. / Microelectronics Reliability 54 (2014) 2151–2153
Fig. 3. The simple circuit for describing Fig. 2.
Fig. 1. The alignment of ln (Ith) vs. 1/gT pairs when pure internal losses are involved.
of such a series resistance is then a starting step for any deeper analysis of the electrical end electro-optical characteristics, including that relationship between laser voltage and monitor current IM (V) that, named Trans characteristics in a recent paper [5], lead to reconstruct the pure radiative component Iph of the total laser current I. When one performs such step for the initial state, after the first FIB and after the second FIB, one obtains the result shown in Fig. 2. Here both the laser current I and the relative monitor diode current IM are plotted vs. the separately calculated reduced voltage V RSI, at the initial step (suffix ‘‘0’’), the first FIB (‘‘1’’) and the second FIB (‘‘2’’). Also the corresponding threshold voltages Vth are put into evidence. The surprising result was that after the last FIB, the clamp voltage Vth2 appeared to decrease, while the threshold current increased, in evident conflict with the statement about their relationship. The evidence, in current I2, for a parasitic shunting ohmic path, roughly corresponding to some 700 X, addresses the qualitative interpretation. Anyway, the preservation of alignment of the highest point in Fig. 1 with the others remains puzzling, if relevant extra-currents are invoked.
correspond to the initial state, and the resistance itself can be neglected. FIB is likely to introduce both optical effects, by modifying the lateral losses, and electrical effects in terms of a parasitic shunting path along the walls of the grooved trench. Such electrical effects can be neglected when confined to the sub-mA range, as for current I1 in Fig. 2. In this case, in Fig. 3 the lateral resistance R0 can be omitted, and the dV/dI characteristics correctly give the overall series resistance RS when measured at I > Ith. On the contrary, the last curve I2 shows a dominant leakage current, that does not physically involve the optically active region, but drains current enough to perturb the I(V) relationship in the whole device. It is matter of trivial circuit analysis to see that, when the element L in Fig. 3 clamps at its internal clamp voltage Vth, not only Ith but also gT and the externally measured value of itself Vth change depending on the critical ratio R0/(RS + R0). More precisely, if the following relationships hold for the initial state (no shunting path) for I > Ith.
V ¼ V th þ RS I IM ¼ gðI Ith Þ
ð1Þ
the appearance of the parallel path changes them into
8 <
0 V ¼ RSRþR ðV th þ RS IÞ 0 : IM ¼ R0 g I RS þR0 Ith þ V th RS þR0 R0 R0
ð2Þ
When one considers the configuration in Fig. 3, where the block L represents the laser diode, large values of the shunt resistance R0
The result is that the functional relationships between the measurable quantities V, I, IM remain the same, and that the four constants Vth, Ith, RS, g seem to change their value, mostly depending on the value of the ratio R0/(RS + R0) that for the given case amounts to about 0.985. The correspondence between old and new parameters is:
Fig. 2. I and IM characteristics at the initial state (0), after the 1st FIB (1) and after 2nd FIB (2).
Fig. 4. The last curve IM2 can be brought to recover the previous state, IM1, by adjusting the electrical parameters of the shunting path.
3. Analysis and discussion
G. Mura et al. / Microelectronics Reliability 54 (2014) 2151–2153
2153
It should be noticed that this states that the third point in Fig. 1 is possibly unreliable, and that the FIB method can be safely applied only when evidence exists that no electrical alterations took place at currents higher that the mA range. Finally, Fig. 5 shows a XEBIC map superimposed onto the corresponding SEM image on the polished facet of the device under test after the final FIB. The edges of the floating blocking layer disappear on the left side, that is just the side where FIB created its trenches [1], confirming with full experimental evidence the hypothesis of a more and more important shunting path, shorting the lateral junction, growing during FIB action. Fig. 5. XEBIC confirms the expected shunted path by cancelling the upper blocking junction signal on the left side, where FIB grooved a trench.
8 0 V th ! RSRþR V th > > 0 > > > R < RS ! 0 RS RS þR0 > g ! RSRþR0 0 g > > > > : 0 Ith ! RSRþR Ith þ VRth0 0
ð3Þ
Here is the explanation of the apparent reduction of the clamp voltage Vth: the lateral path kids the electrical model, and emulates the original simple pure serial scheme with false values for the key parameters. Not only the clamp voltage, but even the series resistance RS seems to reduce, challenging any physical interpretation, upon FIB milling of the laser surface. But what is possibly more intriguing is that the qualitative effect of the parasitic path is that it simultaneously increases the threshold current Ith and decreases the total optical efficiency g. This is just what one expects from an increase of optical losses, the foundation of Fig. 1 and a key point in Ref. [1]. It is then natural to wonder if the third point in Fig. 3 is a true effect of pure optical losses, or if it is somewhat affected by electrical effects, or even if it is only an electrical effect. One way at least to infer the answer is to check if, supposing a tentative Rx for the parasitic shunt resistance, one can reproduce or approach the former state IM1 starting from the measured values of IM2, by means of the inverse of Eq. (3). Fig. 4 shows an intermediate situation, that confirms that the variations from state 1 to state 2 can be reproduced by mere electrical means. The choice of a suitable value for Rx allows to continuously recover all intermediate states between 1 and 2. Such a possibility does not say, in principle, that nothing optical took place after the second FIB, but prevents from separating the optical from the electrical effects.
4. Conclusions A puzzling set of measurements in a FIB-modified laser diode has been shown to be explained by a modified equivalent circuit, and confirmed by EBIC. This raises some warnings in applying the FIB method for gain measurement, as reported in [1], when evidence for electrical alterations appear in the mA range of the laser current, which also defines a criterion for validating data in such FIB-based experiments.
Acknowledgments The authors would like to thank Simona Podda, PhD (Telemicroscopy Laboratory – Sardegna Ricerche – Pula) for FIB device preparation. This work is partially supported by Sardinia Regional Government (P.O.R. Sardegna F.S.E. Operational Programme of the Autonomous Region of Sardinia, European Social Fund 2007–2013 – Axis IV Human Resources, Objective l.3, Line of Activity l.3.1 ‘‘Avviso di chiamata per il finanziamento di Assegni di Ricerca’’).
References [1] Vanzi et al. Optical losses in single-mode laser diodes. Microelectron Reliab 2013;53(9–11):1529–33. [2] Vanzi et al. XEBIC at the dual beam. Microelectron Reliab 2013;53(9– 11):1399–402. [3] Coldren LA, Corzine SW, Mašanovic´ ML. Diode lasers and photonic integrated circuits. 2nd ed. Hoboken, New Jersey: John Wiley & Sons, Inc.; 2012. [4] Mura G, Vanzi M. The interpretation of the DC characteristics of LED and laser diodes to address their failure analysis. Microelectron Reliab 2010;50(4):471–8. [5] Mura G, Vanzi M, Marcello G, Cao R. The role of the optical trans-characteristics in laser diode analysis. Microelectron Reliab 2013;53(9–11):1538–42.