252 ACI Structural Journal/March-April 2005

ACI Structural Journal, V. 102, No. 2, March-April 2005.MS No. 03-438 received October 20, 2003, and reviewed under Institute publication

policies. Copyright © 2005, American Concrete Institute. All rights reserved, includingthe making of copies unless permission is obtained from the copyright proprietors. Pertinentdiscussion including author’s closure, if any, will be published in the January-February2006 ACI Structural Journal if the discussion is received by September 1, 2005.

ACI STRUCTURAL JOURNAL TECHNICAL PAPER

Reinforced concrete (RC) jacketing is most frequently used tostrengthen columns. The common practice to prepare the interfaceis empirically based and consists on increasing the surface roughness,applying a bonding agent and eventually steel connectors. Anexperimental study was performed to analyze the influence of theinterface treatment on the structural behavior of columns strengthenedby RC jacketing. Seven column-footing, full-scale models werebuilt. Three months later, the columns were strengthened by RCjacketing after their surface had been prepared considering differenttechniques. Later, the models were tested under monotonic loading. Itwas concluded that, for current undamaged columns (that is,where a bending moment-shear force ratio is greater than 1.0 m), amonolithic behavior of the composite element can be achievedeven without increasing their surface roughness, using bondingagents, or applying steel connectors before strengthening it byRC jacketing.

Keywords: concrete; strengthening; surface.

INTRODUCTIONJacketing is one of the most frequently used techniques to

strengthen reinforced concrete (RC) columns. With thismethod, axial strength, bending strength, and stiffness of theoriginal column are increased. It is well known that thesuccess of this procedure is dependent on the monolithicbehavior of the composite element. To achieve this purpose,the treatment of the interface must be carefully chosen. Thecommon practice consists of increasing the roughness of theinterface surface and applying a bonding agent, normally anepoxy resin. Steel connectors are also occasionally applied.These steps involve specialized workmanship, time, and cost.Concerning the added concrete mixture and due to the reducedthickness of the jacket, the option is usually a grout withcharacteristics of self-compacting concrete (SCC) and high-strength concrete (HSC).

In published experimental studies on this subject, thepreparation of the column surface before jacketing is alwaysreferred to. Ramírez and Bárcena1 increased the roughnessof the columns by chipping; Bett, Klingner, and Jirsa2

submitted models to light sandblasting; Alcocer and Jirsa,3

Gomes,4 and Gomes and Appleton5 used a chipping hammerto expose the outermost concrete aggregate; Rodriguez andPark6 had the surface of columns lightly roughened bychipping before jacketing; and Stoppenhagen, Jirsa, andWyllie7 used an electric concrete hammer to roughen thespandrels. Although researchers refer to the importance ofthe interface behavior, a quantitative analysis of its influenceis never reported.8

An initial experimental study was performed by theauthors9,10 to quantify in small specimens the influence of:1) the roughness of the interface surface; 2) using a bondingagent; 3) the added concrete mixture; and 4) applying steel

connectors on the strength of concrete against concretejoints. Pull-off tests, slant shear tests, and push-off tests wereperformed and it was concluded that: 1) sandblasting is thebest roughness treatment between those adopted; 2) the useof epoxy resins does not improve the interface strength ifsandblasting is used; 3) adding an HSC increases the inter-face strength; and 4) the use of steel connectors does notsignificantly increases the interface debonding stress,although, after that, shear stress is highly dependent on therelation between the cross section area of steel connectorsand the area of the interface.

Based on these conclusions, seven column-footing, full-scale models were built. Three months later, each column wasencased, considering different surface treatments. Twenty-eight days later the models were tested under monotonicloading. The objective of these tests was to analyze theinfluence of the interface treatment on the structural behaviorof the strengthened column under monotonic loading.

RESEARCH SIGNIFICANCEThe common practice to prepare a RC column to be

strengthened by jacketing is empirically based and consists onincreasing the roughness of the interface surface, applying abonding agent and eventually steel connectors. The mainobjective of these procedures is to achieve a monolithicbehavior of the composite element. The research studypresented in this paper allows engineers to choose thebest treatment based on experimental results instead ofempirical judgment.

EXPERIMENTAL INVESTIGATIONAll models were built at the same time. The materials

chosen were concrete with 20 MPa characteristiccompressive cylinder strength at 28 days and steel with400 MPa characteristic yielding stress. The dimensionsadopted for the original column cross section and for thereinforced concrete jacket thickness were 0.20 x 0.20 m2 and35 mm, respectively. The column height was 1.35 m and thecorresponding jacket height was 0.90 m. The column wassymmetrically reinforced with three bars with 10 mm diameterat each face. The longitudinal reinforcement of the jacketwas the same and it was anchored to the footing in apredrilled hole of 250 mm depth, with a commercial epoxyresin. The transverse reinforcement of the column consistedof 6 mm diameter stirrups spaced 150 mm and the transverse

Title no. 102-S25

Reinforced Concrete Jacketing—Interface Influence on Monotonic Loading Responseby Eduardo N. B. S. Júlio, Fernando A. B. Branco, and Vítor D. Silva

253ACI Structural Journal/March-April 2005

reinforcement of the added jacket consisted of 6 mm diameterstirrups spaced 75 mm and out of phase with those of thecolumn (Fig. 1), since this is the most effective geometry toobtain a monolithic behavior of the strengthened column.4

The loading system consisted of an increasing horizontalforce and a constant axial force of 170 kN. This was achievedwith a hydraulic jack, positioned horizontally at 1.0 m fromthe column footing, with both ends hinged to avoidsecondary efforts, and a tubular system of two sets of twowelded U profiles, connected with two prestressing tendons,tensioned with a hydraulic jack (Fig. 1).

The models’ footing was fixed to the laboratory slab bymeans of a tubular system of two sets of two welded Uprofiles, positioned at the footing ends. Each tube wasconnected to the slab with two DYWIDAG bars (Fig. 1). Atension force of at least 50 kN was installed in each of thesebars in order to resist footing slipping and rotation.

The axial force was measured with a load cell placedbetween the top set of the welded U profiles and thehydraulic jack used to apply the axial force (Fig. 1,Element A). The horizontal force was obtained from thedifference between the values read in two load cells, placedon opposite sites of the column top (Fig. 1, Elements B andC). The tension installed in the DYWIDAG bars wasmeasured with four load cells (Fig. 1, Elements D throughG). The imposed horizontal displacement was measured bya displacement transducer (Fig. 1, Element H).

Strain gauges were bonded to longitudinal and transversereinforcing bars of the column and of the added concretejacket (Fig. 1). On each central bar, close to the footing, werebonded strain gauges (Fig. 1, Elements 1 to 4 and 7 to 10).On the second stirrup from the bottom, of both the columnand the concrete jacket, were also bonded strain gauges, inopposite branches (Fig. 1, Elements 5, 6, 11, and 12).

The characteristics of the seven models tested weredefined according to the conclusions of the study on thejoints behavior performed by the authors.9,10 The first model(M1) was left unstrengthened to serve as the reference specimen.A second model (M2) was strengthened with a nonadherentjacket, materialized with a thin, hard, greased layer placed onthe interface. The objective of using this second model wasto reach the lower limit of the structural behavior of thecomposite model. A third model (M3) was producedmonolithically to reach the upper limit of that behavior. Afourth model (M4) was strengthened by jacketing, withoutany interface treatment. A fifth model (M5) was strengthened byjacketing after its interface surface had been treated by sand-blasting. This surface preparation method was chosenbecause it originated the highest values of shear strength andtension strength obtained with slant shear tests and pull-offtests, respectively.9 For the same reason, a bonding agentwas not used. In fact, the value of the ultimate shear strength

obtained with the slant shear specimens and the value of theultimate tension strength obtained with the pull-off specimenswere higher with the interface surface prepared with sand-blasting than with the interface surface prepared with sand-blasting followed by epoxy resin application.10 A sixth model(M6) was strengthened by jacketing after its interface surfacehad been prepared by sandblasting and steel connectors hadbeen applied. Although the values of the debonding shearstrength of the push-off specimens with none, two, four, orsix steel connectors were identical, the corresponding values ofthe maximum shear strength, after that point, were almostdirectly proportional to the number of steel connectorsemployed.10 For this reason, it was decided to consider thismodel. Finally, a seventh model (M7) was strengthened byjacketing after sandblasting its interface surface and after theaxial force had been applied. Here the purpose was toanalyze the difference between strengthening columns withand without an axial force already applied. The second situationimplies, in most practical situations, an active shoring of thecolumn with additional costs. The characteristics of theseseven models are summarized in Table 1.

Due to the positive results obtained with high-strengthconcrete,10 all added jackets were cast with a commercialgrout with characteristics of SCC and HSC, with the exception

Eduardo N. B. S. Júlio is an assistant professor at the University of Coimbra, Portugal.He received his PhD from the University of Coimbra in 2001. His research interestsinclude structural strengthening and rehabilitation of buildings and monuments.

ACI member Fernando A. B. Branco is a professor at IST (Technical University ofLisbon) and head of the Construction Sector. He is a member of ACI Committee 342,Evaluation of Concrete Bridges and Bridge Elements. His research interests includedesign, rehabilitation, and construction technology of concrete structures.

Vítor D. Silva is an associate professor at the University of Coimbra. He received hisPhD from the Faculty of Aerospace Engineering at the University of Stuttgart, Germany.His research interests include rheological modeling of structural materials, mainlyvisco-elasticity and elasto-visco-plasticity, nonlinear structural analysis includinggeometrical nonlinearity, and rehabilitation of civil engineering structures.

Fig. 1—Testing installation, instrumentation, and cross section.

Table 1—Description of models and compressive strength of concrete

Models Description

Compressive strength of concrete, MPa

Original column

Added jacket

M1 Nonstrengthened column 34.60 —

M2 Column with nonadherent jacket 35.48 83.58

M3 Column with monolithic jacket(cast simultaneously) 34.75 34.75

M4 Column jacketed withoutsurface preparation 34.64 79.79

M5 Column jacketed after surface preparation with sandblasting 34.79 82.76

M6Column jacketed after surface preparation with sandblasting and application of steel

connectors35.13 81.68

M7Column jacketed after surface preparation with sandblasting and after loading of axial

force35.36 80.51

254 ACI Structural Journal/March-April 2005

of the concrete used in the jacket of Model M3, which wasthe same as that used in the original column, since they werecast at the same time.

RESULTSThe results of the monotonic tests performed with the

seven described models were analyzed taking severalparameters into consideration: 1) cracking pattern; 2)measured yielding load, including a comparison withtheoretical values; 3) maximum load, also compared withtheoretical values; 4) initial stiffness and secant stiffness;5) axial load stability because it was decided to keep itbetween 160 and 180 kN; 6) strain analysis of the columnand added jacket bars; and 7) computation of the column andadded jacket contributions to the global strength.

The analysis of the cracking pattern observed in each ofthe seven models was the only parameter available on site tocompare their structural behavior. On the jackets’ top, theonly cross section where the interface boundary was visible,no cracking was registered, excluding the M2 model wherenonadherence between the original column and the jacketwas produced. All models with added RC jacketing showeda similar cracking pattern (Fig. 2) except Model M2 (Fig. 3).The crushing level of concrete was significantly lower in themodels which were strengthened with HSC jackets than in

the nonstrengthened model (M1) and in the monolithicmodel (M3), both casted with NSC (normal strengthconcrete) only.

The yielding load was determined from the differencebetween the values measured on the two load cells of theloading system (Fig. 1, Elements B and C), when themeasured strains on the longitudinal bars reached yielding.An analytical approach was also performed to predict theyielding load, assuming two hypotheses: total nonadherenceand perfect bonding of the jacket. For the first case, it wasassumed that the curvature radius of the original column andof the added jacket were the same at the support crosssection. For the second case, compatible strain diagrams ofthe original column and of the added jacket at the supportcross section were assumed. The experimentally determinedvalue of the steel yielding strain was fixated at the mosttensioned bars of the added jacket. The strain diagram wasestablished iteratively until the corresponding stress diagrampresented a resultant force of the same value of the measuredaxial force. It was adopted the parabola-rectangle stressdiagram for concrete. With the resultant bending moment,the yielding force could be easily determined. For Model M7,strengthened after the axial force had been applied, theprocedure adopted to determine that the theoretical yieldingforce was adapted to take into account an initial straindiagram due to that load. Considering perfect bonding, therelative error between the experimental and the theoreticalvalue varied from –5.2% to +6.7% (Table 2), except forModel M2. Considering total nonadherence, the relativeerror between the experimental and the theoretical valuevaried from –31.8% to –22.7% (Table 2), for the samemodels. This leads to the conclusion, proven by visualinspection, that there was no jacket debonding in any model,excluding M2. For this model, the relative errors referred towere +14.9% and –14.5%, respectively, (Table 2), indicatingthat the desired nonadherence was not totally achieved. InFig. 4 the theoretical (nonadherent and monolithic) valuesand the experimental values of the yielding load of eachmodel are plotted.

The experimental value of the maximum load was alsoobtained from the maximum difference measured betweenthe two load cells of the loading system (Fig. 1, Elements Band C). An analytical study was also developed to predict themaximum load, assuming the same conditions referred tototal nonadherence and perfect bonding of the jacket. Thealgorithm previously referred to was modified in order to

Table 2—Experimental and theoretical values (assuming nonadherent jacket and monolithic cross section) of yielding load of each model

Experimental values

Theoretical values

Nonadherent jacketMonolithic

cross section

ModelsAxial

force, kNYielding load, kN

Yielding load, kN Error, %

Yielding load, kN Error, %

M1 168.9 29.9 — — 31.4 +4.8

M2 172.5 57.5 50.2 –14.5 67.6 +14.9

M3 173.2 66.8 50.9 –23.8 63.5 –5.2

M4 170.8 66.2 50.5 –31.1 67.9 +2.5

M5 170.9 64.5 50.6 –27.5 68.1 +5.3

M6 171.6 66.7 50.6 –31.8 68.1 +2.1

M7 170.5 61.1 49.8 –22.7 65.5* +6.7

*This value was determined considering column was jacketed after loading axial force.

Fig. 2—Cracking pattern of Model M4.

Fig. 3—Cracking pattern of Model M2.

ACI Structural Journal/March-April 2005 255

fixate the ultimate concrete strain at the most stressedconcrete fibers, instead of fixating the steel yielding stress atthe most tensioned bars. Considering perfect bonding, therelative error between the experimental and the theoreticalvalue varied from –1.9% to +6.7% (Table 3), except for M2and M5. Considering total nonadherence, the relative errorbetween the experimental and the theoretical value variedfrom –28.1% to –15.4% (Table 3), for the same models,confirming the conclusion that no slipping between the originalcolumn and the added jacket occurred in these models. InModel M2, the experimental and theoretical value confirmedthat its behavior was between the theoretical behavior ofperfect bonding and absolute absence of friction. For Model M5the experimental value was 16.7% higher than the theoreticalvalue, assuming perfect bonding (Table 3), and a deficiencyin the experimental procedures was found, so this test wasnot considered. The models with HSC totally-bonded jacketspresented similar horizontal force versus displacementcurves (Fig. 5, Curve M6). The curve of Model M2 wassimilar but with a lower resistance (Fig. 5, Curve M2). It canalso be concluded that the resistance of the strengthenedmodels, including Model M2, is much greater than that of thenonstrengthened model (Fig. 5, Curve M1).

It was noticed that the resistance of the monolithic ModelM3 was slightly inferior to that of Model M6, probably dueto the fact that the concrete of M3 model jacket was a normalstrength concrete and the concrete of M6 model jacket was acommercial grout with average compressive strength ofapproximately 80 MPa. Model M7, identical to Model M5except for the fact that it was strengthened after the axialforce had been applied, showed a similar behavior in relationto the other models with HSC perfectly bonded jackets. Itcan be concluded that, for the adopted conditions, the fact ofthe strengthening operation being performed with or withoutaxial loading has no influence on the structural compositebehavior. In Fig. 6, the theoretical (nonadherent and monolithic)values and the experimental values of the maximum load ofeach model are plotted.

The initial stiffness determination of the seven monoton-ically tested models was performed based on the horizontalforce versus displacement curves (Fig. 5). The initial stiffnessof the models was obtained by an interpolator polynomial(considering all values up until the yielding strain) of themost tensioned reinforcing bars. The secant stiffness wasobtained by dividing the experimental value of the horizontalyielding force by the corresponding displacement (Fig. 7).The only relevant conclusions from this parameter are that

the initial stiffness and secant stiffness of the strengthenedmodels are much higher than that of the original column andthat the secant stiffness of the models with a perfectlybonded jacket is slightly higher than that of the model with anonadherent jacket.

The adopted reduced axial force design value was 0.4,giving an applied axial force value of 170 kN. With thepreliminary tests it was verified that keeping this valueconstant presented some difficulties. For this reason it waskept between 160 and 180 kN. Due to the interaction between

Table 3—Experimental and theoretical values (assuming nonadherent jacket and monolithic cross section) of maximum load of each model

Experimental values

Theoretical values

Nonadherent jacketMonolithic

cross section

ModelAxial

force, kNMaximum load, kN

Maximum load, kN Error, %

Maximum load, kN Error, %

M1 175.7 33.3 — — 33.0 –0.9

M2 173.5 71.5 64.8 –10.3 82.0 +12.8

M3 173.2 73.5 63.7 –15.4 74.9 +1.9

M4 177.6 77.5 65.5 –18.3 83.1 +6.7

M5 175.6 96.9 65.5 –47.9 83.0 –16.7

M6 174.7 83.8 65.4 –28.1 82.9 –1.1

M7 175.6 80.7 64.6 –24.9 82.0 +1.6Fig. 4—Theoretical and experimental values of yieldingloading for each model.

Fig. 5—Horizontal force versus displacement curves ofModels M1, M2, and M6.

Fig. 6—Theoretical and experimental values of maximumload of each model.

256 ACI Structural Journal/March-April 2005

the axial force and the bending moment it was important toensure that its value would not go out of range, or else thebehavior of the different models could not be compared. Insome tests, the axial force value was controlled manually andin the others it was controlled automatically. In all models theaxial force level was kept within the defined range.

The effects that mobilize the transverse reinforcingstirrups are: 1) horizontal tension by Poisson effect due tocompression combined with bending moment; 2) inclinedstrut tension due to shear; and 3) redistribution of stressesdue to concrete cracking. This complex system gave rise to

significant differences between the strain values measuredwith strain gauges bonded on opposite branches of the samestirrup. In spite of this, it could be concluded that the strainvalue of the original column stirrup was considerably higheron the nonstrengthened model, M1, than on the strengthenedmodels. On the monolithic model and on the model with thenonadherent jacket, M3 and M2, respectively, the originalcolumn stirrup and the jacket stirrup were equally strained.On the models with a perfectly bonded jacket, the strainvalue of the jacket stirrup (Fig. 8) was significantly higherthan that of the original column stirrup (Fig. 9). This indi-cates that, in these models, concrete confinement is mainlydue to the jacket stirrups.

The strain diagram on the base cross section of each modelwas another parameter adopted to analyze and compare themodels’ behavior. The jacket and column analyticaldiagrams were plotted with the analytically determinedlongitudinal reinforcing bar strains. The correspondingexperimental diagrams were superimposed on these forcomparison (Fig. 10 and 11). For the situation of yieldingforce applied, the agreement between the experimental andthe analytical diagrams, assuming perfect bonding of thejacket, was good for all models (Fig. 10), except for Model M2,confirming that all these models behaved monolithicallyindependent of the interface surface treatment. For Model M2,although the superimposition of the experimental diagram andthe analytical diagram (assuming a completely nonadherent

Fig. 7—Secant stiffness of each model.

Fig. 8—Strain evolution on jacket stirrup of Model M6.

Fig. 9—Strain evolution on original column strip of Model M6.

Fig. 10—M7 model theoretical and experimental straindiagrams.

Fig. 11—M2 model theoretical and experimental straindiagrams.

257ACI Structural Journal/March-April 2005

jacket) was not perfect (Fig. 11), it clearly showed that slippingof the added jacket had occurred.

Due to the good results obtained with the analyticalapproach, the latter was used to analyze the contribution ofthe original column and of the added jacket to the axial forceand horizontal load resistance.

It was concluded that, relative to the axial force, the addedjacket was subjected to: 1) a null compression force in thecase of perfectly nonadherent concrete encasing; 2) acompression force equal to the axial force in the case of themonolithic model; and 3) a compression considerably higherthan the axial force in the case of all the other strengthenedmodels; being the original column subjected to a tensionforce for these situations.

It was also concluded that the yielding force was resisted:1) 41% by the original column and 59% by the jacket in thecase of perfectly nonadherent concrete encasing; 2) 23% bythe original column and 77% by the jacket in the case of themonolithic model; and 3) between 10 and 14% by theoriginal column and between 90 to 86% by the jacket in thecase of all the other strengthened models.

CONCLUSIONSThe analysis of results of this experimental study led to the

following statements:1. All models behaved monolithically independent of the

adopted interface preparation method, with the exception ofModel M2, in which the nonadherence of the jacket wasprovoked;

2. Even Model M2 presented a structural behaviorbetween the theoretical perfectly frictionless model and thetheoretical perfectly adherent model;

3. Whether the strengthening operation was carried outwith or without an axial load applied had no significantinfluence for the adopted conditions;

4. The resistance of the strengthened models wasconsiderably higher than that of the original column andslightly higher than that of the monolithic model;

5. The stiffness of the strengthened models was considerablyhigher than that of the original column;

6. The transverse reinforcement strain of the originalcolumn was significantly higher in the nonstrengthenedmodel than in the strengthened models, although the horizontalforce applied in the first case was less than half the corre-sponding value in the other cases; and

7. The contribution of the adherent jacket to the horizontalforce resistance varied between 86 and 90%.

This means that for current undamaged columns (namelywith a bending moment) with a shear force ratio greater than1.0 m, a monolithic behavior of the composite element canbe achieved even without increasing their surface roughness,using bonding agents, or applying steel connectors beforestrengthening it by RC jacketing. It should be noted,however, that for other conditions, such as RC short columnsand deteriorated or damaged RC columns, these conclusionsmay not apply.

With this study, it was also confirmed that RC jacketing isa very effective strengthening technique, leading to values ofresistance and stiffness of the strengthened column considerablyhigher than those of the original column.

ACKOWLEDGMENTSThe authors are grateful to Sika, Hilti, Betão Liz, Fivinte, DYWIDAG,

Pregaia, Cimpor, and Secil for their collaboration in this research project.

REFERENCES1. Ramírez Ortiz, J. L., and Bárcena Diaz, J. M., “Strengthening

Effectiveness of Low Quality Reinforced Concrete Columns Strengthenedby Two Different Procedures,” Informes de la Construcción, No. 272, July1975, pp. 89-98. (in Spanish)

2. Bett, B. J.; Klingner, R. E.; and Jirsa, J. O., “Lateral Load Response ofStrengthened and Repaired Reinforced Concrete Columns,” ACI StructuralJournal, V. 85, No. 5, Sept.-Oct. 1988, pp. 499-508.

3. Alcocer, S., and Jirsa, J., “Assessment of the Response of ReinforcedConcrete Frame Connections Redesigned by Jacketing,” Proceedings of theFourth U.S. National Conference on Earthquake Engineering, V. 3, May1990, pp. 295-304.

4. Gomes, A., “Behavior and Strengthening of Reinforced ConcreteElements Subjected to Cyclic Loading,” PhD thesis, Instituto SuperiorTécnico, 1992, 333 pp. (in Portuguese)

5. Gomes, A., and Appleton, J., “Experimental Tests of StrengthenedReinforced Concrete Columns Subjected to Cyclic Loading,” RevistaPortuguesa de Engenharia de Estruturas, No. 38, 1994, pp. 19-29. (inPortuguese)

6. Rodriguez, M., and Park, R., “Seismic Load Tests on ReinforcedConcrete Columns Strengthened by Jacketing,” ACI Structural Journal,V. 91, No. 2, Mar.-Apr. 1994, pp. 150-159.

7. Stoppenhagen, D. R.; Jirsa, J. O.; and Wyllie, L. A., Jr., “SeismicRepair and Strengthening of a Severely Damaged Concrete Frame,” ACIStructural Journal, V. 92, No. 2, Mar.-Apr. 1995, pp. 177-187.

8. Júlio, E. S.; Branco, F.; and Silva, V. D., “Structural Rehabilitation ofColumns using Reinforced Concrete Jacketing,” Progress in StructuralEngineering and Materials, V. 5, No. 1, Jan.-Mar. 2003, pp. 29-37.

9. Júlio, E. S.; Branco, F.; and Silva, V. D., “Concrete-to-ConcreteBond Strength—Influence of the Roughness of the Substrate Surface,”Construction and Building Materials, V. 18, No. 9, pp. 675-681.

10. Júlio, E. S., “Influence of the Interface on the Behavior of ColumnsStrengthened by Reinforced Concrete Jacketing,” PhD thesis, Universidade deCoimbra, 2001, 274 pp. (in Portuguese)